{mode.title}
{mode.description}
- {mode.tags.map(tag =>
- {tag} )}
diff --git a/CHANGELOG.md b/CHANGELOG.md index 61338315..bfa8339c 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -2,6 +2,83 @@ All notable changes to this project will be documented in this file. +## [0.8.3] - 2026-10-14 + +Changes since v0.8.2, through upstream main `31c605568` (2026-10-13; 282 commits). The plugin-system series was reverted and is not part of this release. [Website release notes](./website-docs/07-releases/v0.8.3.md) · [Full commit comparison](https://github.com/Tencent/WeKnora/compare/v0.8.2...31c6055687d33f9ad7a92cf453a264ce13a1700d). + +### New Features + +- **Image gallery** — browse document images in an infinite masonry view, search captions/OCR, filter observed attributes and multiple source documents, and inspect images with their source metadata (#3694, #4130, #4156, #4159, #4190). +- **Multimodal Embedding and retrieval** — support for multimodal embedding models that encode text and images into a shared vector space, including Tencent WeMM-Embedding and Qwen3-VL-Embedding; opt-in image-vector retrieval; image-aware reranking and image context for vision chat models. Existing knowledge bases remain opted out; enable multimodal processing and `image_vector_enabled`, then reparse existing documents. Scanned PDF pages are excluded (#3683, #4013, #4080). +- **Source citation navigation** — open cited passages in original PDF, Office, Markdown, TXT and EPUB files; audio references seek to segment timestamps where available. Older documents fall back to text matching until reparsed (#3700, #3794, #4139). +- **On-demand chat attachments** — upload persists the source and queues background parsing; Quick Answer waits for pending parses after the stream starts, while agents use `read_file` to read `attachment://` documents and images when needed, with or without a sandbox. Small ready files may be inlined; successful image observations can be reused for the same model and reading request. Attachments referenced by live conversations or forks survive TTL cleanup (#4203). +- **Platform user management** — list users, create accounts, enable/disable accounts, reset passwords and manage system administrators in System Settings. Disabling revokes sessions; self-disable and disabling the last active administrator are rejected (#4153). +- **Invitation registration mode** — `invite_register` allows new accounts only through valid invitations. The legacy `invite_only` value now disables invitation registration as well as public password registration; existing users can still accept invitations (#3900). +- **Administrator model catalog** — inspect built-in/deployment/effective catalogs, preview and publish an overlay, and restore historical configurations with optimistic version and deployment-baseline checks (#3662). +- **Model integrations** — Hugging Face TEI, Pinecone Inference and Cohere v2 reranking; WeMM-Embedding 2B/4B/9B presets and a Transformers serving script; Claude Opus/Sonnet/Haiku 5.5 catalog entries. The catalog now includes 30 built-in vendors (#3956, #3957, #4158, #4238). +- **MCP interoperability** — support `2026-07-28` discovery alongside `2025-11-25` initialization; external peers negotiate independently. OAuth supports preregistered clients and JWT Bearer assertions for callers without an interactive browser when the upstream supports that grant (#4095, #4229). +- **Agent tool catalog** — `GET /agents/tools` exposes selectable tools and capability requirements to the editor; empty and unset tool lists have distinct semantics, and new unknown tool names are rejected (#4211). +- **Docker sandbox terminal** — interactive terminal support joins Cube/E2B; Docker disconnect ends the shell, and reconnect starts a new shell (#4023). +- **Lite host skills** — install and run skills on the macOS desktop host, with version directories and activation separate from the workspace catalog (#3661). +- **Seafile connector** — synchronize selected library files/directories with hierarchy, incremental checkpoints, bounded downloads and optional deletion reconciliation; encrypted libraries and SeaDoc are excluded (#3616). +- **Data source capabilities** — Confluence directory hierarchy and private image assets (#3319), Yuque TOC hierarchy (#3476), and uploaded Office/PDF files from DingTalk Docs (#3786). +- **Chat UI** — global search within the @ mention menu, unread-reply markers, clearer rewind confirmation, and screenshot paste in embedded chat (#4249, #4247). +- **Cross-KB Wiki search** — `POST /wiki-search` searches accessible Wiki scopes (#3535). +- **Traditional Chinese UI** — add `zh-TW`, including Seafile translations (#3916). + +### Improvements + +- **Retrieval APIs** share rerank behavior and diagnostics; `knowledge-search` exposes thresholds, retrieval switches and rerank overrides. Knowledge-base and explicit-document scopes are a union; document-only search must omit the enclosing KB IDs (#3675, #4191). +- **Retrieval quality and cost** — correct recall, score scales, RRF rank handling and tool behavior; accelerate MMR with sparse postings; skip KB retrieval for summarize intent; show FILTER_TOP_K and graph-query truncation explicitly to the model (#3687, #3758, #3885). +- **Multimodal models** — batch Qwen3-VL-Embedding message inputs, keep text and images in the same embedding space, and expose image-input and rerank capabilities in the editor (#4126, #4080). Image/OCR processing records distinct validation outcomes instead of indexing empty, repetitive or truncated output (#4132, #4135). +- **Wiki** — improve graph layout, theme captions and group focus (#4147); reject machine whole-page rewrites that lose table rows (#3793); drain follow-up work with bounded concurrency; resolve titles for unloaded backlinks (#3931). +- **Parser integrations** — AnyDoc 0.2.4 parses assets once; self-hosted MinerU 4.0 V1 API is detected alongside legacy endpoints. MinerU/cloud parse timeouts follow configurable budgets (#3891). +- **Description generation** — `KNOWLEDGE_BASE_PROFILE_MAX_TOKENS` raises the default knowledge-base description completion budget to 2048 and makes it configurable (#4230). +- **Agent runtime** — stream-establishment waits obey idle timeouts; LLM retries honor `Retry-After`; file-read budgets count paths; tool validation preserves numeric precision, counts Unicode string lengths correctly and compares enums without string coercion (#3936, #3937). +- **MCP UX/runtime** — service and skill list search (#3731), tool approval can be followed by OAuth authorization (#4248), structured and embedded resource results reach agents (#3766), and slow connection cleanup no longer blocks unrelated services. +- **Chat/editor UX** — smoother citation streaming/highlights; preserve Markdown list content, caret and indentation selection boundaries; preserve external-link targets and literal single tildes; fewer redundant startup requests and false unsaved-change prompts (#3562, #3818, #3963, #4195, #4224). +- **Model usage** — agent completion events carry the actual turn chat-model ID, distinct from the originally requested message model when fallback occurs (#4252). +- **Datasource robustness** — bounded pagination and response sizes across Feishu, Notion, IMA, Yuque and GitLab; Feishu Wiki request pacing and rate-limit retries (#3645, #3770, #3840, #3841, #3889, #3890, #3985, #4039). +- **Observability** — distinguish retrieval diagnostic spans and close terminal/superseded document-processing spans (#3855, #4128). +- **UI consistency** — refine sidebar, list pages, toolbox, settings, knowledge-base hints and citation views (#4154, #4193). + +### Bug Fixes + +- **Index persistence** — Elasticsearch CopyIndices preserves vectors/enabled state (#3678); bulk item errors surface (#3842); Qdrant waits for writes, handles concurrent collection creation and uses returned pagination offsets (#3882, #3999); Milvus copies vectors without offset-window loss and preserves connection/resolver behavior (#3654, #3883); Weaviate handles concurrent class creation, bounded cursors and GraphQL failures (#3689, #3881); SQLite index writes roll back instead of silently leaving partial rows (#3768, #3843). Qdrant/Milvus keyword failures are no longer reported as empty matches (#3847). +- **Document processing** — avoid stranded processing/finalizing rows (#3696), retry transient indexing failures (#3915), remove written chunks after quota rejection (#3656), reject invalid JSON before asynchronous parsing (#3360), reject malformed or truncated summaries (#3769), skip deleted-document/empty table-summary work, and clean FAQ payloads while preserving cloned entry body/index state (#3993). +- **Parsing fidelity** — repair malformed Excel fill styles and unsupported validation ranges (#3608, #3638); preserve PPTX fallback image associations (#3973); resolve EPUB images relative to spine-ordered chapters and preserve inline previews (#3924, #3977, #4226); preserve whitespace-only inline elements (#3978), Markdown fences, escaped table pipes and table child source offsets (#3969, #3979); align AnyDoc EMF/WMF extensions (#3938); bound remote image downloads (#3906). +- **Datasource consistency** — preserve retryable cursors after Notion/Yuque detail failures (#3682, #3693, #3702), stop deletion reconciliation after truncated Notion discovery (#3845), expose block-limit truncation (#3771), remove old searchable bodies when Notion/IMA/GitLab sources are emptied (#4008), compare GitLab commit snapshots and bound downloads (#3887), retry RSS full-text fallback and update unchanged article titles/links (#3824). DingTalk skips and unsupported blocks are visible, and partial workspace access no longer invalidates the entire source (#3782, #3822, #4079). Manual-only sync and cron validation persist correctly (#3653). +- **Agent correctness** — preserve adjacent assistant tool calls (#3574); recover tool and MCP-prewarm panics (#3846); reject broken/cancelled compaction streams as checkpoints (#3994); preserve steering order through persistence failures (#4078) and attach queued follow-up answers to the right question; preserve SQL string literals and adjacent prefixed Python literals (#3812, #4006); keep failed table loads retryable and skip style-only worksheets (#3660, #3745). +- **Authorization and credentials** — database analysis is scoped to authorized shared KBs (#4136); failed token persistence cannot report a successful login; canonical storage references work in bare providers (#3166); cloned/released images acquire resource catalog bindings (#3752); access logs stop buffering full request bodies (#3927), and LLM debug logs stay inside their directory. +- **MCP** — retain dynamic-registration client secrets (#3763), reconnect forgotten sessions (#3765), report advertised directory failures (#3905), include shared KBs in endpoint scopes (#3848), bound resource pagination (#4007), preserve other callers’ current connections during authorization while recycling stale registrations (#4138), fix OAuth expiry across process time zones (#4207), preserve disabled state at creation, close legacy error responses (#3930), and mark read-only tools non-destructive. +- **Models** — report only thinking levels the wire format can send (#3552), honor legacy OpenRouter thinking (#3807), send explicit empty assistant content for tool-call turns (#3605), avoid unsupported Claude temperature (#4155), treat HTTP 400 as reachable in chat connection tests (#3929), and preserve copied model names (#3830). +- **Chat/integrations** — restore valid file/tool/KB selections, preserve reference custom metadata (#4087), allow image-only embedded messages, fix host-context follow-up suggestions, correct desktop browser pairing against the local listener (#3677), and allow shared agents to use the caller's local browser. Tenant administrators can manage channel sessions (#3939). +- **IM** — preserve WeCom images alongside text (#3901), validate custom WeCom WSS endpoints, handle unresolved Feishu resource images (#2761), and retry Yunzhijia attachment downloads (#3809). +- **Skills/sandbox** — skip vendored dependency trees during verification (#3943), do not fail installs on non-blocking observations (#4246), heal sandbox idle-sweep races (#3980), pass desktop slot timing into renewal, recognize E2B `in_progress` builds (#4108), and omit environment overrides on Aliyun E2B snapshot creation (#4054). +- **Wiki** — retain pending work when source lookup fails (#3715); propagate per-page link-rebuild failures (#3884). +- **Knowledge graph** — preserve entity-relationship direction and document provenance (#3950). +- **CLI/Go SDK** — preserve recursive upload folders (#3753), apply list limits after filters/sorts (#3990), render paginated text results (#3984), propagate pagination read errors (#3983), classify wrapped HTTP errors (#3992), preserve jq integer precision (#3991), replay multipart uploads (#3989), and parse multiline SSE with optional field spacing (#2122, #3975). +- **Miscellaneous** — Metaso non-webpage scopes return results (#4010), web-search count honors its configured maximum (#4011), dataset generation accepts a selected answer model and a supported default (#3944, #3965), COS uses virtual-hosted S3 addressing (#3722), dead-letter messages truncate at rune boundaries (#3964), and graceful shutdown runs resource cleanup (#3804). + +### Upgrade Notes + +- `invite_only` now blocks password signup even with an invitation; choose `invite_register` to retain invitation signup. `DISABLE_REGISTRATION=true` still maps to `invite_only`. OIDC provisioning and administrator-created accounts are separate. +- `allowed_tools: []` now means no selected built-ins; missing/null selects defaults. Migrations rewrite existing smart-reasoning empty arrays to explicit defaults. An agent using all KBs then follows the same capability filter as an explicitly selected default list, excluding Wiki-only KBs from that retrieval scope. Switch-managed tools retain their independent controls. +- PostgreSQL migrations **000111–000118** add model-catalog overrides, embeddings lookup indexes, gallery projections/backfill, source locators, extension upgrades, explicit agent defaults, attachment reference indexing and timezone-aware OAuth timestamps. SQLite migrations **000031–000036** cover the applicable changes and fill the previously missing MCP metadata schema. +- ParadeDB now uses **`paradedb/paradedb:v0.25.11-pg17`**. Migration 000115 upgrades eligible installed pg_search 0.22.0–0.25.10 and vector 0.8.0–0.8.3 when the target extension packages are available. Follow the [database upgrade guide](./website-docs/01-getting-started/06-paradedb-upgrade.md); do not run the old 0.22.6 upgrade command against the new image. Allow time for image projection backfill and ordinary index creation; very large embeddings tables may prebuild the same indexes concurrently. +- Existing image vectors are not generated retroactively. Reparse selected documents after enabling image retrieval, and rebuild vectors created with incompatible old text-input encoding when switching to Qwen3-VL-Embedding/WeMM message-based templates. +- OAuth migration treats existing naive timestamps as UTC; tokens previously written with a shifted non-UTC value need reauthorization or refresh. Live conversation attachments now persist beyond their upload TTL, increasing retention-related storage use. +- MCP tool calls outside the endpoint allowlist return JSON-RPC `-32602` rather than an `isError` tool result. Docker terminal reconnect opens a new shell. + +### Infrastructure, Tests & Documentation + +- Helm pins app/frontend/docreader images to appVersion by default, exposes Service-name overrides and rollout strategy, and documents shared-storage requirements (#3669, #4109, #4180). +- Build real arm64 sandbox/Cube images; support app/docreader download mirrors (#4073); install CA certificates before BrowserSkill mirror substitution (#4104); fix dev-script exit propagation and MinIO image retrieval. +- Make outbound dialing configurable with `WEKNORA_OUTBOUND_DIAL_TIMEOUT` (default 5s), including Compose forwarding (#4122, #4157). Keep existing migration objects tolerated where appropriate (#4164). +- Regenerate Swagger from current handlers, document system model-catalog endpoints, and correct the processing-trace route base path; retain reproducible generation (#3751); clarify frontend proxy configuration, upload folder paths, model copy, retrieval-scope semantics and API quickstart tenant extraction; refresh license notices (#4178). +- Refresh website feature/API/deployment documentation, add this release page and real product screenshots, update homepage/navigation and the project version to 0.8.3. Preserve historical release notes. +- Add or strengthen regressions for parsing/splitting, SSE, attachments, tool selection, multimodal protocols, connectors, index failures, OAuth, user management and platform configuration; synchronize host-skill and concurrent-schema tests to avoid timing-only failures. + ## [0.8.2] - 2026-09-24 ### New Features diff --git a/README.md b/README.md index 120e233c..258fd240 100644 --- a/README.md +++ b/README.md @@ -10,7 +10,7 @@
-
+
@@ -138,13 +138,13 @@ docker compose down # Stop services
If you already have WeKnora running and downloaded a newer release:
```bash
-# Set WEKNORA_VERSION in .env to the target release (e.g. 0.8.2), or keep latest
+# Set WEKNORA_VERSION in .env to the target release (e.g. 0.8.3), or keep latest
docker compose pull # Pull images matching WEKNORA_VERSION
docker compose up -d # Recreate containers with new images
```
> [!NOTE]
-> `docker compose up -d` alone reuses locally cached images and may leave the UI version out of sync with the release you downloaded. Read the [upgrade notes](https://weknora.weixin.qq.com/docs/07-releases/v0.8.2#upgrade-notes) before moving from v0.8.0.
+> `docker compose up -d` alone reuses locally cached images and may leave the UI version out of sync with the release you downloaded. Read the [upgrade notes](https://weknora.weixin.qq.com/docs/07-releases/v0.8.3#upgrade-notes) before upgrading. Deployments older than v0.8.2 should also read the [v0.8.2 notes](https://weknora.weixin.qq.com/docs/07-releases/v0.8.2#upgrade-notes).
### Other ways to deploy
@@ -166,6 +166,16 @@ All options, hardware requirements and deployment topologies: [Installation guid
## What's New
+### v0.8.3 · 2026-10-14 · [release notes](https://weknora.weixin.qq.com/docs/07-releases/v0.8.3)
+
+- **Image gallery**: browse document images and filter by source, caption, OCR and image attributes.
+- **Multimodal Embedding**: support for multimodal embedding models that encode text and images into a shared vector space, including Tencent WeMM-Embedding and Qwen3-VL-Embedding, with opt-in image-vector retrieval and compatible image-aware reranking.
+- **Wiki graph**: clearer group layouts and theme captions, with group focus, switching and return to the full graph.
+- **Source citations**: navigate from an answer to the cited passage in the original document.
+- **Chat and tools**: on-demand attachment reading without a sandbox, unread replies, @ mention search, Docker terminals, Lite host skills and MCP interoperability updates.
+- **Administration and integrations**: platform user management, invitation-only signup mode, administrator model catalog, WeMM embeddings, TEI/Pinecone/Cohere reranking, Seafile and Traditional Chinese.
+- **Reliability**: broad retrieval, parsing, datasource, Wiki, OAuth and CLI fixes. Review [upgrade requirements](https://weknora.weixin.qq.com/docs/07-releases/v0.8.3#upgrade-notes) for registration modes, empty tool lists, database migrations and reindexing.
+
### v0.8.2 · 2026-09-24 · [release notes](https://weknora.weixin.qq.com/docs/07-releases/v0.8.2)
Agents can operate the browser on your computer, knowledge bases can be published to other AI tools over MCP, and a running conversation can be steered, forked or rewound.
@@ -364,7 +374,7 @@ The full product documentation lives at **[weknora.weixin.qq.com/docs](https://w
| [Configuration](https://weknora.weixin.qq.com/docs/01-getting-started/04-configuration) | Environment variables and models |
| [Troubleshooting FAQ](https://weknora.weixin.qq.com/docs/01-getting-started/05-troubleshooting) | Common problems and fixes |
| [API reference](https://weknora.weixin.qq.com/docs/04-api/01-api-overview) | REST API overview |
-| [Release notes](https://weknora.weixin.qq.com/docs/07-releases/v0.8.2) | What changed in each release |
+| [Release notes](https://weknora.weixin.qq.com/docs/07-releases/v0.8.3) | What changed in each release |
## Development
diff --git a/README_CN.md b/README_CN.md
index 0ae6a3e5..5438f923 100644
--- a/README_CN.md
+++ b/README_CN.md
@@ -10,7 +10,7 @@
-
+
@@ -138,13 +138,13 @@ docker compose down # 停止服务
若已有部署并下载了更新的 release:
```bash
-# 在 .env 中将 WEKNORA_VERSION 设为目标版本(如 0.8.2),或保持 latest
+# 在 .env 中将 WEKNORA_VERSION 设为目标版本(如 0.8.3),或保持 latest
docker compose pull # 拉取与 WEKNORA_VERSION 匹配的镜像
docker compose up -d # 用新镜像重建容器
```
> [!NOTE]
-> 仅执行 `docker compose up -d` 会复用本地缓存镜像,可能导致 Web UI 显示版本与下载的 release 不一致。从 v0.8.0 升级前请先阅读[升级须知](https://weknora.weixin.qq.com/docs/07-releases/v0.8.2#upgrade-notes)。
+> 仅执行 `docker compose up -d` 会复用本地缓存镜像,可能导致 Web UI 显示版本与下载的 release 不一致。升级前请先阅读[升级须知](https://weknora.weixin.qq.com/docs/07-releases/v0.8.3#upgrade-notes)。
### 其他部署方式
@@ -166,6 +166,16 @@ docker compose up -d # 用新镜像重建容器
## 最新更新
+### v0.8.3 · 2026-10-14 · [版本说明](https://weknora.weixin.qq.com/docs/07-releases/v0.8.3)
+
+- **画廊**:集中浏览文档图片,按来源、描述、OCR 与图片属性筛选,查看原图和出处。
+- **多模态 Embedding**:支持将文字与图片编码到同一向量空间的多模态向量模型,例如腾讯 WeMM-Embedding、Qwen3-VL-Embedding 等;可按需开启图片向量检索,并搭配支持图片输入的重排模型。
+- **Wiki 图谱**:重新组织分组布局与主题说明,支持聚焦分组、切换分组和返回全图。
+- **原文引用**:从回答直接定位原始文档中的引用片段。
+- **对话与工具**:无需沙箱即可按需读取附件,未读回答标记、@ 全局搜索、Docker 终端、Lite 本机技能与 MCP 兼容性更新。
+- **管理与集成**:平台用户管理、仅限邀请注册、管理员模型目录、WeMM 向量、TEI/Pinecone/Cohere 重排、Seafile 数据源与繁体中文。
+- **稳定性**:集中修复检索、解析、数据源、Wiki、OAuth 和 CLI。注册模式、空工具列表、数据库迁移与向量重建要求见[升级须知](https://weknora.weixin.qq.com/docs/07-releases/v0.8.3#upgrade-notes)。
+
### v0.8.2 · 2026-09-24 · [版本说明](https://weknora.weixin.qq.com/docs/07-releases/v0.8.2)
智能体可以操作你电脑上的浏览器,知识库可以通过 MCP 接入其他 AI 工具,进行中的对话可以补充要求、分叉或回滚。
@@ -364,7 +374,7 @@ weknora chat "总结一下设计文档"
| [配置说明](https://weknora.weixin.qq.com/docs/01-getting-started/04-configuration) | 环境变量与模型配置 |
| [常见问题排查](https://weknora.weixin.qq.com/docs/01-getting-started/05-troubleshooting) | 常见问题与解决办法 |
| [API 文档](https://weknora.weixin.qq.com/docs/04-api/01-api-overview) | REST API 总览 |
-| [版本说明](https://weknora.weixin.qq.com/docs/07-releases/v0.8.2) | 各版本变更详解 |
+| [版本说明](https://weknora.weixin.qq.com/docs/07-releases/v0.8.3) | 各版本变更详解 |
## 开发指南
diff --git a/README_JA.md b/README_JA.md
index d12df80a..a7dc5c50 100644
--- a/README_JA.md
+++ b/README_JA.md
@@ -10,7 +10,7 @@
-
+
@@ -138,13 +138,13 @@ docker compose down # サービスを停止
既存のデプロイがあり、新しい release をダウンロードした場合:
```bash
-# .env の WEKNORA_VERSION を対象バージョン(例: 0.8.2)に設定、または latest のまま
+# .env の WEKNORA_VERSION を対象バージョン(例: 0.8.3)に設定、または latest のまま
docker compose pull # WEKNORA_VERSION に一致するイメージを取得
docker compose up -d # 新しいイメージでコンテナを再作成
```
> [!NOTE]
-> `docker compose up -d` のみではローカルキャッシュのイメージが再利用され、Web UI の表示バージョンがダウンロードした release と一致しない場合があります。v0.8.0 からアップグレードする前に[アップグレードの注意事項](https://weknora.weixin.qq.com/docs/07-releases/v0.8.2#upgrade-notes)を確認してください。
+> `docker compose up -d` のみではローカルキャッシュのイメージが再利用され、Web UI の表示バージョンがダウンロードした release と一致しない場合があります。アップグレードする前に[アップグレードの注意事項](https://weknora.weixin.qq.com/docs/07-releases/v0.8.3#upgrade-notes)を確認してください。
### その他のデプロイ方法
@@ -166,6 +166,16 @@ docker compose up -d # 新しいイメージでコンテナを再作成
## 最新アップデート
+### v0.8.3 · 2026-10-14 · [リリースノート](https://weknora.weixin.qq.com/docs/07-releases/v0.8.3)
+
+- **画像ギャラリー**:文書内の画像を一覧表示し、出典・説明・OCR・画像属性で絞り込み。
+- **マルチモーダル Embedding**:テキストと画像を同じベクトル空間で扱うモデルに対応。Tencent WeMM-Embedding や Qwen3-VL-Embedding などを利用でき、画像ベクトル検索と対応モデルによる画像リランキングを任意で有効化。
+- **Wiki グラフ**:グループ配置とテーマ説明を改善。グループへのフォーカス、切り替え、全体表示への復帰に対応。
+- **原文引用**:回答から原文の引用箇所へ直接移動。
+- **会話とツール**:サンドボックス不要の添付ファイルのオンデマンド読み取り、未読返信、@ 検索、Docker ターミナル、Lite ローカルスキル、MCP 互換性の改善。
+- **管理と連携**:ユーザー管理、招待による登録、管理者モデルカタログ、WeMM、TEI/Pinecone/Cohere、Seafile、繁体字中国語。
+- **安定性**:検索、解析、同期、Wiki、OAuth、CLI を修正。登録モード、空のツール一覧、DB 移行と再インデックスについては[アップグレード案内](https://weknora.weixin.qq.com/docs/07-releases/v0.8.3#upgrade-notes)を確認してください。
+
### v0.8.2 · 2026-09-24 · [リリースノート](https://weknora.weixin.qq.com/docs/07-releases/v0.8.2)
エージェントがあなたのコンピューターのブラウザを操作できるようになり、ナレッジベースを MCP で他の AI ツールに公開でき、実行中の会話に要件を追加したり分岐・巻き戻ししたりできます。
@@ -364,7 +374,7 @@ weknora chat "設計ドキュメントを要約して"
| [設定](https://weknora.weixin.qq.com/docs/01-getting-started/04-configuration) | 環境変数とモデル設定 |
| [トラブルシューティング](https://weknora.weixin.qq.com/docs/01-getting-started/05-troubleshooting) | よくある問題と解決方法 |
| [API ドキュメント](https://weknora.weixin.qq.com/docs/04-api/01-api-overview) | REST API の概要 |
-| [リリースノート](https://weknora.weixin.qq.com/docs/07-releases/v0.8.2) | 各バージョンの変更点 |
+| [リリースノート](https://weknora.weixin.qq.com/docs/07-releases/v0.8.3) | 各バージョンの変更点 |
## 開発ガイド
diff --git a/README_KO.md b/README_KO.md
index f33eb8e5..55c4f6fb 100644
--- a/README_KO.md
+++ b/README_KO.md
@@ -10,7 +10,7 @@
01 / KNOWLEDGE AT WORK 查询资料用 RAG,处理多步任务用 Agent, {mode.description} 02 / INTRODUCING v0.8.2 {item} 03 / SKILLS & SANDBOX 02 / INTRODUCING v0.8.3 03 / YOUR LOCAL BROWSER 04 / SKILLS & SANDBOX 05 / INTEGRATIONS 同步飞书、Confluence、GitLab 等平台的资料, 06 / INTEGRATIONS 同步飞书、Confluence、GitLab 等平台的资料, 06 / BUILT FOR YOUR TEAM 配置数据存储与成员权限, 07 / BUILT FOR YOUR TEAM 配置数据存储与成员权限, 支持 Docker、Kubernetes 与 Helm,也提供 Lite 单机版和桌面应用。模型、向量数据库和存储后端可按需替换,支持本地推理。 多空间隔离与四级角色矩阵。API Key 按能力和知识库限定范围,支持 OIDC 身份集成,可限制服务只访问白名单域名。 空间审计日志、运行时任务队列面板与 Langfuse 追踪,帮助团队定位问题、管理运行状态。 GET STARTED
-
+
@@ -148,13 +148,13 @@ docker compose down # 서비스 중지
기존 배포가 있고 새 release를 다운로드한 경우:
```bash
-# .env에서 WEKNORA_VERSION을 대상 버전(예: 0.8.2)으로 설정하거나 latest 유지
+# .env에서 WEKNORA_VERSION을 대상 버전(예: 0.8.3)으로 설정하거나 latest 유지
docker compose pull # WEKNORA_VERSION에 맞는 이미지 가져오기
docker compose up -d # 새 이미지로 컨테이너 재생성
```
> [!NOTE]
-> `docker compose up -d`만 실행하면 로컬 캐시 이미지가 재사용되어 Web UI 버전이 다운로드한 release와 일치하지 않을 수 있습니다. v0.8.0에서 업그레이드하기 전에 [업그레이드 안내](https://weknora.weixin.qq.com/docs/07-releases/v0.8.2#upgrade-notes)를 확인하세요.
+> `docker compose up -d`만 실행하면 로컬 캐시 이미지가 재사용되어 Web UI 버전이 다운로드한 release와 일치하지 않을 수 있습니다. 업그레이드하기 전에 [업그레이드 안내](https://weknora.weixin.qq.com/docs/07-releases/v0.8.3#upgrade-notes)를 확인하세요.
### 기타 배포 방식
@@ -176,6 +176,16 @@ docker compose up -d # 새 이미지로 컨테이너 재생성
## 최신 업데이트
+### v0.8.3 · 2026-10-14 · [릴리스 노트](https://weknora.weixin.qq.com/docs/07-releases/v0.8.3)
+
+- **이미지 갤러리**: 문서 이미지를 모아 보고 출처, 설명, OCR 및 이미지 속성으로 필터링.
+- **멀티모달 Embedding**: 텍스트와 이미지를 같은 벡터 공간으로 인코딩하는 모델 지원. Tencent WeMM-Embedding, Qwen3-VL-Embedding 등을 사용할 수 있으며 이미지 벡터 검색과 호환 모델의 이미지 리랭킹을 선택적으로 활성화.
+- **Wiki 그래프**: 그룹 배치와 주제 설명을 개선하고 그룹 집중 보기, 그룹 전환 및 전체 그래프 복귀를 지원.
+- **원문 인용**: 답변에서 원문의 인용 부분으로 바로 이동.
+- **대화와 도구**: 샌드박스 없이 첨부 파일을 필요할 때 읽기, 읽지 않은 답변 표시, @ 검색, Docker 터미널, Lite 로컬 스킬과 MCP 호환성 개선.
+- **관리와 연동**: 플랫폼 사용자 관리, 초대 등록 모드, 관리자 모델 카탈로그, WeMM, TEI/Pinecone/Cohere, Seafile과 번체 중국어.
+- **안정성**: 검색, 파싱, 동기화, Wiki, OAuth와 CLI 수정. 등록 모드, 빈 도구 목록, DB 마이그레이션과 재색인은 [업그레이드 안내](https://weknora.weixin.qq.com/docs/07-releases/v0.8.3#upgrade-notes)를 확인하세요.
+
### v0.8.2 · 2026-09-24 · [릴리스 노트](https://weknora.weixin.qq.com/docs/07-releases/v0.8.2)
에이전트가 사용자 컴퓨터의 브라우저를 조작할 수 있고, 지식베이스를 MCP로 다른 AI 도구에 공개할 수 있으며, 진행 중인 대화에 요구 사항을 추가하거나 분기·되감기할 수 있습니다.
@@ -374,7 +384,7 @@ weknora chat "설계 문서를 요약해 줘"
| [설정](https://weknora.weixin.qq.com/docs/01-getting-started/04-configuration) | 환경 변수와 모델 설정 |
| [문제 해결 FAQ](https://weknora.weixin.qq.com/docs/01-getting-started/05-troubleshooting) | 자주 발생하는 문제와 해결 방법 |
| [API 문서](https://weknora.weixin.qq.com/docs/04-api/01-api-overview) | REST API 개요 |
-| [릴리스 노트](https://weknora.weixin.qq.com/docs/07-releases/v0.8.2) | 버전별 변경 사항 |
+| [릴리스 노트](https://weknora.weixin.qq.com/docs/07-releases/v0.8.3) | 버전별 변경 사항 |
## 개발자 가이드
diff --git a/VERSION b/VERSION
index 100435be..ee94dd83 100644
--- a/VERSION
+++ b/VERSION
@@ -1 +1 @@
-0.8.2
+0.8.3
diff --git a/docs/docs.go b/docs/docs.go
index 156d80be..5c5bd210 100644
--- a/docs/docs.go
+++ b/docs/docs.go
@@ -782,31 +782,47 @@ const docTemplate = `{
}
}
},
- "/api/v1/knowledge/{id}/spans": {
+ "/artifacts": {
"get": {
- "description": "返回该知识在解析流水线的 trace tree(root → stage → subspan):每段状态、耗时、input/output、错误码、langfuse_trace_id。支持 ?attempt=N 查看历史尝试;不传则返回最新尝试。前端用于渲染时间线 + 多模态/embedding 子节点 + 一键跳转 Langfuse。",
- "consumes": [
- "application/json"
+ "security": [
+ {
+ "Bearer": []
+ },
+ {
+ "ApiKeyAuth": []
+ }
],
+ "description": "跨会话列出当前用户可见会话中技能生成的文件,每个文件只返回最新版本(不含存储 URL)",
"produces": [
"application/json"
],
"tags": [
- "知识管理"
+ "会话"
],
- "summary": "获取知识文档解析的 Span 树(含历史尝试)",
+ "summary": "列出我的产物",
"parameters": [
{
"type": "string",
- "description": "知识ID",
- "name": "id",
- "in": "path",
- "required": true
+ "description": "按文件名过滤",
+ "name": "keyword",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "逗号分隔的扩展名,如 .pdf,.pptx",
+ "name": "file_types",
+ "in": "query"
},
{
"type": "integer",
- "description": "指定尝试号;省略=最新",
- "name": "attempt",
+ "description": "页码",
+ "name": "page",
+ "in": "query"
+ },
+ {
+ "type": "integer",
+ "description": "每页数量(最大 100)",
+ "name": "page_size",
"in": "query"
}
],
@@ -819,6 +835,65 @@ const docTemplate = `{
}
}
}
+ },
+ "delete": {
+ "security": [
+ {
+ "Bearer": []
+ }
+ ],
+ "description": "删除产物库中的一个文件。产物库一行代表一个文件的最新版本,默认连同该文件在本会话中的所有历史版本一并删除",
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "会话"
+ ],
+ "summary": "删除产物库中的文件",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "会话ID",
+ "name": "session_id",
+ "in": "query",
+ "required": true
+ },
+ {
+ "type": "string",
+ "description": "消息ID",
+ "name": "message_id",
+ "in": "query",
+ "required": true
+ },
+ {
+ "type": "integer",
+ "description": "产物在消息中的位置",
+ "name": "index",
+ "in": "query",
+ "required": true
+ },
+ {
+ "type": "boolean",
+ "description": "是否删除所有版本,默认 true",
+ "name": "all_versions",
+ "in": "query"
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "OK",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "404": {
+ "description": "Not Found",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ }
+ }
}
},
"/auth/auto-setup": {
@@ -1345,7 +1420,7 @@ const docTemplate = `{
},
"/auth/register-by-invite": {
"post": {
- "description": "通过 Owner 生成的共享邀请链接 token 完成注册,绕过 invite_only 模式拦截。\n注册者自填邮箱(与 token 不绑定);注册成功后自动加入对应空间。",
+ "description": "通过共享邀请链接 token 完成注册,仅 self_serve / invite_register 模式允许。\n注册者自填邮箱(与 token 不绑定);注册成功后自动加入对应空间。",
"consumes": [
"application/json"
],
@@ -1380,6 +1455,12 @@ const docTemplate = `{
"$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
}
},
+ "403": {
+ "description": "注册功能已禁用",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ },
"409": {
"description": "邮箱已注册",
"schema": {
@@ -2666,7 +2747,7 @@ const docTemplate = `{
"ApiKeyAuth": []
}
],
- "description": "更新指定 IM 渠道的名称、模式、知识库、凭证或启用状态",
+ "description": "更新指定 IM 渠道的名称、模式、回复语言、知识库、凭证或启用状态",
"consumes": [
"application/json"
],
@@ -2686,7 +2767,7 @@ const docTemplate = `{
"required": true
},
{
- "description": "更新字段(name/mode/output_mode/knowledge_base_id/credentials/enabled)",
+ "description": "更新字段(含 locale)",
"name": "request",
"in": "body",
"required": true,
@@ -4939,7 +5020,7 @@ const docTemplate = `{
"ApiKeyAuth": []
}
],
- "description": "在知识库中执行向量和关键词混合搜索。推荐使用 POST;GET 携带 JSON 请求体仍受支持(兼容旧客户端)。",
+ "description": "底层召回:向量+关键词混合检索,默认不 rerank(可用 rerank 字段开启);一般检索请用 /knowledge-search。推荐 POST,GET 带 JSON 体仅兼容旧客户端。",
"consumes": [
"application/json"
],
@@ -5004,7 +5085,7 @@ const docTemplate = `{
"ApiKeyAuth": []
}
],
- "description": "在知识库中执行向量和关键词混合搜索。推荐使用 POST;GET 携带 JSON 请求体仍受支持(兼容旧客户端)。",
+ "description": "底层召回:向量+关键词混合检索,默认不 rerank(可用 rerank 字段开启);一般检索请用 /knowledge-search。推荐 POST,GET 带 JSON 体仅兼容旧客户端。",
"consumes": [
"application/json"
],
@@ -5061,6 +5142,166 @@ const docTemplate = `{
}
}
},
+ "/knowledge-bases/{id}/images": {
+ "get": {
+ "security": [
+ {
+ "Bearer": []
+ },
+ {
+ "ApiKeyAuth": []
+ }
+ ],
+ "description": "浏览某知识库内的全部图片资产,支持关键字搜索(caption/OCR 文本)、按图片属性筛选(attr_filters / attr_rules)与排序分页。",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "knowledge"
+ ],
+ "summary": "列出知识库图片资产",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "知识库 ID",
+ "name": "id",
+ "in": "path",
+ "required": true
+ },
+ {
+ "type": "integer",
+ "description": "页码,默认 1",
+ "name": "page",
+ "in": "query"
+ },
+ {
+ "type": "integer",
+ "description": "每页数量,默认 20,最大 1000",
+ "name": "page_size",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "关键字,对 search_in 指定字段做不区分大小写的子串匹配",
+ "name": "keyword",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "参与搜索的属性 ID(逗号分隔);须为 in_searchfield=true",
+ "name": "search_in",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "排序属性 ID;须为 in_sortfield=true,默认 created_at",
+ "name": "sort_by",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "排序方向:asc / desc,默认 desc",
+ "name": "sort_order",
+ "in": "query"
+ },
+ {
+ "type": "boolean",
+ "description": "仅包含启用状态的图片",
+ "name": "is_enabled",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "来源文档 ID 列表(逗号分隔,最多 50 个),仅返回这些文档的图片",
+ "name": "knowledge_ids",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "属性筛选 JSON,同一属性多值 OR、属性间 AND",
+ "name": "attr_filters",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "属性逐值裁决 JSON(off 隐藏 / on 强制显示,on 优先)",
+ "name": "attr_rules",
+ "in": "query"
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "图片资产分页列表",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "400": {
+ "description": "请求参数错误",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ }
+ }
+ },
+ "post": {
+ "security": [
+ {
+ "Bearer": []
+ },
+ {
+ "ApiKeyAuth": []
+ }
+ ],
+ "description": "与 GET /images 同义,筛选参数改经 JSON body 传递:knowledge_ids 不再受 URL 长度限制(上限 1000),供来源文档全选使用。",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "knowledge"
+ ],
+ "summary": "列出知识库图片资产(body 版)",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "知识库 ID",
+ "name": "id",
+ "in": "path",
+ "required": true
+ },
+ {
+ "description": "筛选参数(与 GET /images 同名参数同义)",
+ "name": "body",
+ "in": "body",
+ "schema": {
+ "$ref": "#/definitions/internal_handler.imageListRequestBody"
+ }
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "图片资产分页列表",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "400": {
+ "description": "请求参数错误",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ }
+ }
+ }
+ },
"/knowledge-bases/{id}/knowledge": {
"get": {
"security": [
@@ -5155,6 +5396,18 @@ const docTemplate = `{
"description": "为 true 时同时返回子文件夹内的文档",
"name": "folder_recursive",
"in": "query"
+ },
+ {
+ "type": "string",
+ "description": "排序字段: updated_at/created_at/file_name,默认 created_at",
+ "name": "sort_by",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "排序方向: asc/desc,默认 desc",
+ "name": "sort_order",
+ "in": "query"
}
],
"responses": {
@@ -5747,6 +6000,50 @@ const docTemplate = `{
}
}
},
+ "/knowledge-bases/{id}/profile/generate": {
+ "post": {
+ "security": [
+ {
+ "Bearer": []
+ },
+ {
+ "ApiKeyAuth": []
+ }
+ ],
+ "description": "基于文档画像聚合,立即重新生成知识库的 AI 描述(不覆盖手写描述)",
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "知识库"
+ ],
+ "summary": "生成知识库描述",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "知识库ID",
+ "name": "id",
+ "in": "path",
+ "required": true
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "生成的知识库画像",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "400": {
+ "description": "知识库类型不支持或未配置模型",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ }
+ }
+ }
+ },
"/knowledge-bases/{id}/shares": {
"get": {
"security": [
@@ -6277,7 +6574,7 @@ const docTemplate = `{
"ApiKeyAuth": []
}
],
- "description": "在知识库中搜索(不使用LLM总结)",
+ "description": "在知识库中搜索(不使用LLM总结)。与产品内问答使用同一检索流程(召回、rerank、合并),外部检索首选;可覆盖召回参数与 rerank",
"consumes": [
"application/json"
],
@@ -7265,6 +7562,53 @@ const docTemplate = `{
}
}
},
+ "/knowledge/{id}/spans": {
+ "get": {
+ "security": [
+ {
+ "Bearer": []
+ },
+ {
+ "ApiKeyAuth": []
+ }
+ ],
+ "description": "返回该知识在解析流水线的 trace tree(root → stage → subspan):每段状态、耗时、input/output、错误码、langfuse_trace_id。支持 ?attempt=N 查看历史尝试;不传则返回最新尝试。前端用于渲染时间线 + 多模态/embedding 子节点 + 一键跳转 Langfuse。",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "知识管理"
+ ],
+ "summary": "获取知识文档解析的 Span 树(含历史尝试)",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "知识ID",
+ "name": "id",
+ "in": "path",
+ "required": true
+ },
+ {
+ "type": "integer",
+ "description": "指定尝试号;省略=最新",
+ "name": "attempt",
+ "in": "query"
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "OK",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ }
+ }
+ }
+ },
"/knowledgebase/{kb_id}/wiki/auto-fix": {
"post": {
"security": [
@@ -8683,6 +9027,20 @@ const docTemplate = `{
"MCP服务"
],
"summary": "获取MCP服务列表",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "Agent ID; needs agent_source_tenant_id",
+ "name": "agent_id",
+ "in": "query"
+ },
+ {
+ "type": "integer",
+ "description": "Shared agent source workspace",
+ "name": "agent_source_tenant_id",
+ "in": "query"
+ }
+ ],
"responses": {
"200": {
"description": "MCP服务列表",
@@ -8696,6 +9054,12 @@ const docTemplate = `{
"schema": {
"$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
}
+ },
+ "403": {
+ "description": "无权使用该共享智能体",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
}
}
},
@@ -10824,6 +11188,65 @@ const docTemplate = `{
}
}
},
+ "/models/catalog/resolve": {
+ "get": {
+ "security": [
+ {
+ "Bearer": []
+ },
+ {
+ "ApiKeyAuth": []
+ }
+ ],
+ "description": "根据厂商、模型名、Base URL 与 extra_config 返回目录解析结果(协议、思考等级、上下文等),供模型编辑器实时展示",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "模型管理"
+ ],
+ "summary": "解析模型的有效接入配置",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "厂商标识",
+ "name": "provider",
+ "in": "query",
+ "required": true
+ },
+ {
+ "type": "string",
+ "description": "模型名",
+ "name": "model",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "Base URL",
+ "name": "base_url",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "模型类型",
+ "name": "model_type",
+ "in": "query"
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "解析结果",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ }
+ }
+ }
+ },
"/models/providers": {
"get": {
"security": [
@@ -10834,7 +11257,7 @@ const docTemplate = `{
"ApiKeyAuth": []
}
],
- "description": "根据模型类型获取支持的厂商列表及配置信息",
+ "description": "根据模型类型获取支持的厂商定义(含图标、默认地址、额外字段、内置模型目录与思考能力)",
"consumes": [
"application/json"
],
@@ -10848,7 +11271,7 @@ const docTemplate = `{
"parameters": [
{
"type": "string",
- "description": "模型类型 (chat, embedding, rerank, vllm)",
+ "description": "模型类型 (chat, embedding, rerank, vllm, asr)",
"name": "model_type",
"in": "query"
}
@@ -11052,6 +11475,68 @@ const docTemplate = `{
}
}
},
+ "/models/{id}/copy": {
+ "post": {
+ "security": [
+ {
+ "Bearer": []
+ },
+ {
+ "ApiKeyAuth": []
+ }
+ ],
+ "description": "复制模型配置。name 与源模型一致,凭证从已存记录复制,不接受客户端重传。",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "模型管理"
+ ],
+ "summary": "复制模型",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "源模型ID",
+ "name": "id",
+ "in": "path",
+ "required": true
+ },
+ {
+ "description": "副本展示名",
+ "name": "request",
+ "in": "body",
+ "required": true,
+ "schema": {
+ "$ref": "#/definitions/internal_handler.CopyModelRequest"
+ }
+ }
+ ],
+ "responses": {
+ "201": {
+ "description": "复制的模型",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "400": {
+ "description": "请求参数错误",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ },
+ "404": {
+ "description": "模型不存在",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ }
+ }
+ }
+ },
"/organizations": {
"get": {
"security": [
@@ -13800,6 +14285,67 @@ const docTemplate = `{
"responses": {}
}
},
+ "/sessions/{session_id}/messages/{message_id}/artifacts/{index}": {
+ "delete": {
+ "security": [
+ {
+ "Bearer": []
+ }
+ ],
+ "description": "删除某条 assistant 消息生成的单个文件;all_versions=true 时连同本会话中该文件的所有历史版本一并删除",
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "会话"
+ ],
+ "summary": "删除会话中的产物文件",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "会话ID",
+ "name": "session_id",
+ "in": "path",
+ "required": true
+ },
+ {
+ "type": "string",
+ "description": "消息ID",
+ "name": "message_id",
+ "in": "path",
+ "required": true
+ },
+ {
+ "type": "integer",
+ "description": "产物在消息中的位置",
+ "name": "index",
+ "in": "path",
+ "required": true
+ },
+ {
+ "type": "boolean",
+ "description": "是否删除该文件的所有版本",
+ "name": "all_versions",
+ "in": "query"
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "OK",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "404": {
+ "description": "Not Found",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ }
+ }
+ }
+ },
"/sessions/{session_id}/messages/{message_id}/artifacts/{index}/download": {
"get": {
"responses": {}
@@ -13954,6 +14500,80 @@ const docTemplate = `{
}
}
},
+ "/sessions/{session_id}/rewind": {
+ "post": {
+ "security": [
+ {
+ "Bearer": []
+ },
+ {
+ "ApiKeyAuth": []
+ }
+ ],
+ "description": "回滚当前会话到指定用户或助手消息:删除其后消息,可到达时 git reset 工作区。",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "会话"
+ ],
+ "summary": "回滚会话",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "会话 ID",
+ "name": "session_id",
+ "in": "path",
+ "required": true
+ },
+ {
+ "description": "回滚请求",
+ "name": "request",
+ "in": "body",
+ "required": true,
+ "schema": {
+ "$ref": "#/definitions/internal_handler_session.RewindSessionRequest"
+ }
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "回滚结果",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "400": {
+ "description": "请求参数错误 / 回滚点角色不支持",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ },
+ "404": {
+ "description": "会话或消息不存在",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ },
+ "409": {
+ "description": "会话正在生成中 / 无检查点 / 沙箱已更换",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ },
+ "500": {
+ "description": "工作区回滚失败",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ }
+ }
+ }
+ },
"/sessions/{session_id}/steer": {
"post": {
"security": [
@@ -14325,9 +14945,21 @@ const docTemplate = `{
"parameters": [
{
"type": "string",
- "description": "Sandbox config ID",
+ "description": "Sandbox config ID; ignored for a shared agent",
"name": "sandbox_config_id",
"in": "query"
+ },
+ {
+ "type": "string",
+ "description": "Agent ID; needs agent_source_tenant_id",
+ "name": "agent_id",
+ "in": "query"
+ },
+ {
+ "type": "integer",
+ "description": "Shared agent source workspace",
+ "name": "agent_source_tenant_id",
+ "in": "query"
}
],
"responses": {
@@ -14337,6 +14969,12 @@ const docTemplate = `{
"type": "object",
"additionalProperties": true
}
+ },
+ "403": {
+ "description": "无权使用该共享智能体",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
}
}
}
@@ -15189,6 +15827,198 @@ const docTemplate = `{
}
}
},
+ "/system/admin/model-catalog": {
+ "get": {
+ "security": [
+ {
+ "Bearer": []
+ }
+ ],
+ "description": "System administrator JWT only; API keys are not accepted. Returns built-in, deployment and effective\ncatalogs, the current overlay, version/baseline tokens, and the latest 20 published revisions.",
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "系统管理"
+ ],
+ "summary": "Get the system model catalog",
+ "responses": {
+ "200": {
+ "description": "OK",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogState"
+ }
+ },
+ "401": {
+ "description": "Unauthorized",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "403": {
+ "description": "Forbidden",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "503": {
+ "description": "Service Unavailable",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ }
+ }
+ },
+ "put": {
+ "security": [
+ {
+ "Bearer": []
+ }
+ ],
+ "description": "System administrator JWT only; API keys are not accepted. Validates and publishes a new overlay\nrevision using the current version and baseline. Stale values return 409. To restore a historical\nconfiguration, submit its overlay with the current version and baseline.",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "系统管理"
+ ],
+ "summary": "Publish a model catalog overlay",
+ "parameters": [
+ {
+ "description": "Version, baseline and overlay object (max 1 MiB)",
+ "name": "request",
+ "in": "body",
+ "required": true,
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogUpdate"
+ }
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "OK",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogState"
+ }
+ },
+ "400": {
+ "description": "Bad Request",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "401": {
+ "description": "Unauthorized",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "403": {
+ "description": "Forbidden",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "409": {
+ "description": "Conflict",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "500": {
+ "description": "Internal Server Error",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ }
+ }
+ }
+ },
+ "/system/admin/model-catalog/preview": {
+ "post": {
+ "security": [
+ {
+ "Bearer": []
+ }
+ ],
+ "description": "System administrator JWT only; API keys are not accepted. Validates the overlay without publishing\nit. Supply the current version and baseline; stale values return 409. The response contains the\ncandidate effective catalog and normalized overlay; history, builtin and deployment are null.",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "系统管理"
+ ],
+ "summary": "Preview a model catalog overlay",
+ "parameters": [
+ {
+ "description": "Version, baseline and overlay object (max 1 MiB)",
+ "name": "request",
+ "in": "body",
+ "required": true,
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogUpdate"
+ }
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "OK",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogState"
+ }
+ },
+ "400": {
+ "description": "Bad Request",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "401": {
+ "description": "Unauthorized",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "403": {
+ "description": "Forbidden",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "409": {
+ "description": "Conflict",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "500": {
+ "description": "Internal Server Error",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ }
+ }
+ }
+ },
"/system/admin/promote": {
"post": {
"description": "Grant system administrator privileges to a user (SystemAdmin only).\nIdempotent: re-promoting an existing system admin returns 200 with no DB write.\nIdentify the user by email (preferred for human operators) or user_id (UUID, for API clients).",
@@ -15617,6 +16447,49 @@ const docTemplate = `{
}
}
},
+ "/system/admin/users": {
+ "get": {
+ "description": "Retrieve a paginated list of every user across tenants (SystemAdmin only).\nSupports ` + "`" + `offset` + "`" + ` (default 0) and ` + "`" + `limit` + "`" + ` (default 50, max 200).\nPassword/hashes are never serialised — only the redacted UserInfo shape.",
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "System Admin"
+ ],
+ "summary": "List all platform users",
+ "parameters": [
+ {
+ "type": "integer",
+ "default": 0,
+ "description": "Page offset",
+ "name": "offset",
+ "in": "query"
+ },
+ {
+ "type": "integer",
+ "default": 50,
+ "description": "Page size (max 200)",
+ "name": "limit",
+ "in": "query"
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "Users retrieved successfully",
+ "schema": {
+ "$ref": "#/definitions/internal_handler.ListAllUsersResponse"
+ }
+ },
+ "403": {
+ "description": "Forbidden: not a system admin",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ }
+ }
+ }
+ },
"/system/admin/users/create": {
"post": {
"description": "Provision a new local user account (SystemAdmin only).\nWhen ` + "`" + `password` + "`" + ` is omitted or null, a cryptographically random\npassword is generated (OIDC-style crypto/rand + base64url)\nand returned once in the response body. Any provided value,\nincluding empty string, is policy-checked. Tenant provisioning\nfollows the shared auth.default_tenant_mode policy.",
@@ -15741,6 +16614,55 @@ const docTemplate = `{
}
}
},
+ "/system/admin/users/status": {
+ "post": {
+ "description": "Disable revokes the target's active sessions immediately.\nSelf-disable and disabling the last active system administrator are rejected.",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "System Admin"
+ ],
+ "summary": "Enable or disable a user account",
+ "parameters": [
+ {
+ "description": "Account status request",
+ "name": "request",
+ "in": "body",
+ "required": true,
+ "schema": {
+ "$ref": "#/definitions/internal_handler.SetUserStatusRequest"
+ }
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "Status updated",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "400": {
+ "description": "Self-disable / last-admin guard",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "404": {
+ "description": "User not found",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ }
+ }
+ }
+ },
"/system/capabilities": {
"get": {
"description": "返回当前部署版本及实际注册的后端路由所对应的功能能力;仅 supported=false 表示入口应隐藏",
@@ -18046,9 +18968,183 @@ const docTemplate = `{
}
}
}
+ },
+ "/wiki-search": {
+ "post": {
+ "security": [
+ {
+ "Bearer": []
+ },
+ {
+ "ApiKeyAuth": []
+ }
+ ],
+ "description": "Cross-KB wiki search using the same POSIX regex ranking as single-KB wiki search",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "Wiki"
+ ],
+ "summary": "Cross-KB wiki search",
+ "parameters": [
+ {
+ "description": "Wiki search request",
+ "name": "request",
+ "in": "body",
+ "required": true,
+ "schema": {
+ "$ref": "#/definitions/internal_handler.SearchWikiRequest"
+ }
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "OK",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "400": {
+ "description": "Bad Request",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ }
+ }
+ }
}
},
"definitions": {
+ "github_com_Tencent_WeKnora_internal_application_service.CatalogProviderView": {
+ "type": "object",
+ "properties": {
+ "api": {
+ "type": "string"
+ },
+ "id": {
+ "type": "string"
+ },
+ "model_thinking_levels": {
+ "description": "ModelThinkingLevels lists, per chat model id, the levels offered once\nthinking is enabled, resolved against the vendor map and protocol so\nthe console does not re-implement level inheritance.",
+ "type": "object",
+ "additionalProperties": {
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_models_api.ReasoningEffort"
+ }
+ }
+ },
+ "model_types": {
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ModelType"
+ }
+ },
+ "models": {
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_models.ModelSpec"
+ }
+ },
+ "name": {
+ "type": "string"
+ },
+ "settings": {
+ "type": "object",
+ "additionalProperties": {}
+ },
+ "vendor_thinking_levels": {
+ "description": "VendorThinkingLevels is what a chat model without its own map offers.",
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_models_api.ReasoningEffort"
+ }
+ }
+ }
+ },
+ "github_com_Tencent_WeKnora_internal_application_service.CatalogRevision": {
+ "type": "object",
+ "properties": {
+ "overlay": {
+ "type": "object"
+ },
+ "updated_at": {
+ "type": "string"
+ },
+ "updated_by": {
+ "type": "string"
+ },
+ "version": {
+ "type": "integer"
+ }
+ }
+ },
+ "github_com_Tencent_WeKnora_internal_application_service.CatalogState": {
+ "type": "object",
+ "properties": {
+ "applied_version": {
+ "type": "integer"
+ },
+ "baseline": {
+ "type": "string"
+ },
+ "builtin": {
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogProviderView"
+ }
+ },
+ "deployment": {
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogProviderView"
+ }
+ },
+ "deployment_error": {
+ "type": "string"
+ },
+ "effective": {
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogProviderView"
+ }
+ },
+ "history": {
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogRevision"
+ }
+ },
+ "overlay": {
+ "type": "object"
+ },
+ "sync_error": {
+ "type": "string"
+ },
+ "version": {
+ "type": "integer"
+ }
+ }
+ },
+ "github_com_Tencent_WeKnora_internal_application_service.CatalogUpdate": {
+ "type": "object",
+ "properties": {
+ "baseline": {
+ "type": "string"
+ },
+ "overlay": {
+ "type": "object"
+ },
+ "version": {
+ "type": "integer"
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_application_service.WikiLintIssue": {
"type": "object",
"properties": {
@@ -18324,6 +19420,152 @@ const docTemplate = `{
}
}
},
+ "github_com_Tencent_WeKnora_internal_models.ModelCost": {
+ "type": "object",
+ "properties": {
+ "cache_read": {
+ "type": "number"
+ },
+ "cache_write": {
+ "type": "number"
+ },
+ "currency": {
+ "type": "string"
+ },
+ "input": {
+ "type": "number"
+ },
+ "output": {
+ "type": "number"
+ }
+ }
+ },
+ "github_com_Tencent_WeKnora_internal_models.ModelSpec": {
+ "type": "object",
+ "properties": {
+ "aliases": {
+ "description": "Aliases are additional ids that resolve to this entry (dated\nsnapshots, vendor-prefixed names on gateways).",
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ },
+ "api": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_models_api.API"
+ },
+ "compat": {
+ "description": "Compat is the protocol-specific overlay for this model (flat object,\nfields depend on API). Decoded lazily by Resolve.",
+ "type": "array",
+ "items": {
+ "type": "integer"
+ }
+ },
+ "context_window": {
+ "type": "integer"
+ },
+ "cost": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_models.ModelCost"
+ },
+ "deprecated": {
+ "description": "Deprecated hides the entry from pickers but keeps resolution working.",
+ "type": "boolean"
+ },
+ "dimension": {
+ "description": "Dimension is the embedding vector width (embedding entries only).",
+ "type": "integer"
+ },
+ "id": {
+ "type": "string"
+ },
+ "input": {
+ "description": "Input lists accepted modalities: text, image, audio, video.",
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ },
+ "match": {
+ "description": "Match is a case-insensitive glob (\"gpt-5*\", \"*deepseek-r1*\") that lets\none entry describe a family. Exact ids and aliases win over patterns;\namong patterns the longest literal prefix wins.",
+ "type": "string"
+ },
+ "max_output_tokens": {
+ "type": "integer"
+ },
+ "name": {
+ "type": "string"
+ },
+ "reasoning": {
+ "description": "Reasoning reports whether the model can emit thinking at all.",
+ "type": "boolean"
+ },
+ "source": {
+ "description": "Source is the documentation URL the facts were taken from.",
+ "type": "string"
+ },
+ "thinking_levels": {
+ "description": "ThinkingLevels maps protocol-neutral levels to vendor values; missing\nkeys inherit the vendor map. See api.ThinkingLevelMap.",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_models_api.ThinkingLevelMap"
+ }
+ ]
+ },
+ "type": {
+ "description": "Type is the WeKnora model type this entry describes. Chat models with\nimage input are still KnowledgeQA; the UI derives VLM eligibility from\nInput. Embedding / Rerank / ASR entries exist for the model picker.",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ModelType"
+ }
+ ]
+ }
+ }
+ },
+ "github_com_Tencent_WeKnora_internal_models_api.API": {
+ "type": "string",
+ "enum": [
+ "openai-completions",
+ "openai-responses",
+ "anthropic-messages",
+ "google-generative-ai",
+ "ollama"
+ ],
+ "x-enum-varnames": [
+ "APIOpenAICompletions",
+ "APIOpenAIResponses",
+ "APIAnthropicMessages",
+ "APIGoogleGenerativeAI",
+ "APIOllama"
+ ]
+ },
+ "github_com_Tencent_WeKnora_internal_models_api.ReasoningEffort": {
+ "type": "string",
+ "enum": [
+ "off",
+ "auto",
+ "minimal",
+ "low",
+ "medium",
+ "high",
+ "xhigh",
+ "max"
+ ],
+ "x-enum-varnames": [
+ "ReasoningOff",
+ "ReasoningAuto",
+ "ReasoningMinimal",
+ "ReasoningLow",
+ "ReasoningMedium",
+ "ReasoningHigh",
+ "ReasoningXHigh",
+ "ReasoningMax"
+ ]
+ },
+ "github_com_Tencent_WeKnora_internal_models_api.ThinkingLevelMap": {
+ "type": "object",
+ "additionalProperties": {
+ "type": "string"
+ }
+ },
"github_com_Tencent_WeKnora_internal_models_limiter.RuntimeStat": {
"type": "object",
"properties": {
@@ -18407,6 +19649,13 @@ const docTemplate = `{
"description": "ReasoningContent stores the OpenAI-protocol reasoning_content emitted by the\nmodel in this round. Persisted on AgentStep so cross-turn replay can put it\nback on the assistant message — required by MiMo / DeepSeek V3.2+ thinking\nmode, ignored by providers that don't recognize the field.",
"type": "string"
},
+ "reasoning_metadata": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ProviderMetadata"
+ },
+ "reasoning_signature": {
+ "description": "ReasoningSignature / ReasoningMetadata are the provider artifacts that\nmust accompany ReasoningContent on replay (Anthropic signatures,\nOpenAI Responses encrypted reasoning items).",
+ "type": "string"
+ },
"thought": {
"description": "LLM's reasoning/thinking (Think phase)",
"type": "string"
@@ -18422,6 +19671,10 @@ const docTemplate = `{
"$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ToolCall"
}
},
+ "truncated": {
+ "description": "Truncated marks the round the completion-token cap cut off. It rides in\nthe agent_steps JSON so a reloaded, shared or re-opened conversation can\nstill show that the answer stops mid-sentence by design, rather than\nlooking finished. Live streaming carries the same fact on the answer\nevent; this is what survives the round trip.",
+ "type": "boolean"
+ },
"user_messages_before": {
"description": "UserMessagesBefore records consumed steer rows in delivery order, before\nthis model response. Unlike timestamps, this remains unambiguous on replay.",
"type": "array",
@@ -19170,6 +20423,10 @@ const docTemplate = `{
}
]
},
+ "reasoning_effort": {
+ "description": "ReasoningEffort selects the thinking intensity (off | auto | minimal |\nlow | medium | high | xhigh | max). Empty keeps the boolean Thinking\nsemantics: true means \"auto\". See internal/models/api.ReasoningEffort.",
+ "type": "string"
+ },
"rerank_model_id": {
"description": "ReRank model ID for retrieval",
"type": "string"
@@ -19736,6 +20993,22 @@ const docTemplate = `{
}
}
},
+ "github_com_Tencent_WeKnora_internal_types.GalleryUserPrefs": {
+ "type": "object",
+ "properties": {
+ "mode": {
+ "description": "Mode is \"all\" (every search-eligible field active) or \"custom\"\n(per-field Status governs). Empty = \"all\".",
+ "type": "string"
+ },
+ "status": {
+ "description": "Status maps a namespaced attribute id (\"builtin:caption\") to\n\"on\"/\"off\". Only consulted in custom mode; an unrecorded field is\noff. Source-declared on/off does not exist — activation is always\nthe user's own record.",
+ "type": "object",
+ "additionalProperties": {
+ "type": "string"
+ }
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_types.GraphNode": {
"type": "object",
"properties": {
@@ -19770,9 +21043,60 @@ const docTemplate = `{
}
}
},
+ "github_com_Tencent_WeKnora_internal_types.ImageActionsConfig": {
+ "type": "object",
+ "properties": {
+ "ocr": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ImageOCRAction"
+ }
+ }
+ },
+ "github_com_Tencent_WeKnora_internal_types.ImageAttrCondition": {
+ "type": "object",
+ "properties": {
+ "is": {
+ "type": "string"
+ },
+ "prop": {
+ "type": "string"
+ }
+ }
+ },
+ "github_com_Tencent_WeKnora_internal_types.ImageOCRAction": {
+ "type": "object",
+ "properties": {
+ "on": {
+ "description": "On is the list of attribute conditions that enable OCR. Matching any one\nturns OCR on.",
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ImageAttrCondition"
+ }
+ },
+ "on_unobserved": {
+ "description": "OnUnobserved is the OCR decision used when the observation does not\nanswer for one of the attributes the policy reads — the attribute is\nabsent because the model skipped the line, or because the value it wrote\nwas not one of the allowed ones. The default is true: an incomplete\nobservation keeps extracting, so a missed block of text costs one extra\ncall instead of being lost. Turn it off only with a model reliable enough\nto answer every attribute line.",
+ "type": "boolean"
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_types.ImageProcessingConfig": {
"type": "object",
"properties": {
+ "image_actions": {
+ "description": "ImageActions overrides the built-in attribute→work table (see\nDefaultImageActions). A nil value keeps the conservative default: run OCR\nfor block text, data visuals, and unobserved images.",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ImageActionsConfig"
+ }
+ ]
+ },
+ "image_attrs_enabled": {
+ "description": "ImageAttrsEnabled turns the attribute-observed image pipeline on for this\nknowledge base. It is off by default on purpose, so upgrading an\nexisting deployment never changes what happens to documents already\nbeing ingested: with it off every image is described and OCR'd exactly\nas before. When on, the describe round also observes image attributes and\nthe attribute policy decides whether OCR runs for it. A single upload can\noverride it per document through\nKnowledgeProcessOverrides.ImageAttrsEnabled.",
+ "type": "boolean"
+ },
+ "image_vector_enabled": {
+ "description": "ImageVectorEnabled lets a knowledge base whose embedding model takes\nimages index each image by its own vector and recall it with one. It\nis off by default on purpose: a model that takes images is also used\nfor plain text, and turning this on costs an embedding call per image\nand widens vector recall. Existing knowledge bases stay off after an\nupgrade because the stored JSON has no such key.",
+ "type": "boolean"
+ },
"model_id": {
"description": "Model ID",
"type": "string"
@@ -19876,7 +21200,7 @@ const docTemplate = `{
],
"properties": {
"representative_user_id": {
- "description": "RepresentativeUserID identifies the user attached to the OTM row for\ndisplay/audit. Optional: when unset, the handler picks a stable default\n(the user from the legacy UserID field, or the workspace's owner).",
+ "description": "RepresentativeUserID is accepted for compatibility and ignored: a\ndirect add attaches no user of the enrolled workspace, since the\ninviter must not choose whose details the roster shows.",
"type": "string"
},
"role": {
@@ -19892,7 +21216,7 @@ const docTemplate = `{
"type": "integer"
},
"user_id": {
- "description": "UserID is retained for backward compatibility. When set without\nTenantID, the handler resolves the user's TenantID and uses this\nuser as the representative.",
+ "description": "UserID is retained for backward compatibility. When set without\nTenantID, the handler resolves the user's TenantID.",
"type": "string"
}
}
@@ -20047,6 +21371,14 @@ const docTemplate = `{
}
]
},
+ "generated_profile": {
+ "description": "GeneratedProfile is the machine-generated description: a gist, merged\ntopics, typical questions and the aggregate snapshot they came from. It\nnever overwrites the user-authored Description; both are shown to agents.",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfile"
+ }
+ ]
+ },
"id": {
"description": "Unique identifier of the knowledge base",
"type": "string"
@@ -20095,6 +21427,14 @@ const docTemplate = `{
"description": "ProcessingCount indicates the number of knowledge items being processed (for document type knowledge bases)",
"type": "integer"
},
+ "profile_config": {
+ "description": "ProfileConfig controls automatic generation of the knowledge-base\ndescription from per-document profiles (document knowledge bases only).",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfileConfig"
+ }
+ ]
+ },
"question_generation_config": {
"description": "QuestionGenerationConfig stores question generation configuration for document knowledge bases",
"allOf": [
@@ -20193,7 +21533,7 @@ const docTemplate = `{
]
},
"image_processing_config": {
- "description": "Image processing configuration",
+ "description": "Image processing configuration.\n\nnil means \"no change\" when updating, the same contract IndexingStrategy\nuses. A request that does not mention this field must leave the image\nsettings alone: the attribute-observation switch and the attribute policy\nare things a knowledge base accumulates, and a client that predates them\nwould otherwise reset the lot on any save. Sending an object — empty\nincluded — replaces the whole configuration.",
"allOf": [
{
"$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ImageProcessingConfig"
@@ -20208,6 +21548,14 @@ const docTemplate = `{
}
]
},
+ "profile_config": {
+ "description": "ProfileConfig controls optional automatic knowledge-base description\ngeneration. nil means \"no change\" when updating.",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfileConfig"
+ }
+ ]
+ },
"wiki_config": {
"description": "Wiki configuration (only for wiki-enabled knowledge bases)",
"allOf": [
@@ -20218,6 +21566,128 @@ const docTemplate = `{
}
}
},
+ "github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfile": {
+ "type": "object",
+ "properties": {
+ "aggregate_hash": {
+ "description": "AggregateHash identifies the aggregate the gist was generated from. A\nmismatch with the live aggregate means the text is stale.",
+ "type": "string"
+ },
+ "error": {
+ "description": "Error holds the last generation error when Status is failed.",
+ "type": "string"
+ },
+ "generated_at": {
+ "description": "GeneratedAt records when the text was produced.",
+ "type": "string"
+ },
+ "gist": {
+ "description": "Gist is a short statement of what the knowledge base covers.",
+ "type": "string"
+ },
+ "model_id": {
+ "description": "ModelID records which chat model produced the text.",
+ "type": "string"
+ },
+ "stats": {
+ "description": "Stats is the aggregate snapshot the text was generated from.",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfileStats"
+ }
+ ]
+ },
+ "status": {
+ "description": "Status is one of the KnowledgeBaseProfileStatus* values.",
+ "type": "string"
+ },
+ "topics": {
+ "description": "Topics is the merged topic list (at most KnowledgeBaseProfileMaxTopics).",
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ },
+ "typical_questions": {
+ "description": "TypicalQuestions lists questions this knowledge base can answer.",
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ }
+ }
+ },
+ "github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfileConfig": {
+ "type": "object",
+ "properties": {
+ "custom_instructions": {
+ "description": "CustomInstructions is appended to the stable system prompt, e.g. the\naudience the description should address or terms to keep.",
+ "type": "string"
+ },
+ "enabled": {
+ "type": "boolean"
+ },
+ "model_id": {
+ "description": "ModelID overrides the chat model; empty falls back to the KB summary model.",
+ "type": "string"
+ }
+ }
+ },
+ "github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfileStats": {
+ "type": "object",
+ "properties": {
+ "doc_types": {
+ "description": "DocTypes counts documents per model-assigned document type.",
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.NamedCount"
+ }
+ },
+ "document_count": {
+ "description": "DocumentCount is the number of eligible documents at generation time.",
+ "type": "integer"
+ },
+ "earliest_at": {
+ "description": "EarliestAt and LatestAt bound document creation times.",
+ "type": "string"
+ },
+ "file_types": {
+ "description": "FileTypes counts documents per file extension.",
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.NamedCount"
+ }
+ },
+ "folders": {
+ "description": "Folders lists the top-level folder names present.",
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ },
+ "latest_at": {
+ "type": "string"
+ },
+ "profiled_count": {
+ "description": "ProfiledCount is how many of them carried a document profile.",
+ "type": "integer"
+ },
+ "raw_topics": {
+ "description": "RawTopics counts documents per normalized topic keyword before merging.",
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.NamedCount"
+ }
+ },
+ "tags": {
+ "description": "Tags counts documents per knowledge-base tag.",
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.NamedCount"
+ }
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_types.KnowledgeBaseShareResponse": {
"type": "object",
"properties": {
@@ -20341,6 +21811,22 @@ const docTemplate = `{
"graph_enabled": {
"type": "boolean"
},
+ "image_actions": {
+ "description": "ImageActions overrides rows of the attribute-\u003ework table for this document\n(merged per action key on top of the knowledge base's table).\nnil keeps the knowledge base's setting.",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ImageActionsConfig"
+ }
+ ]
+ },
+ "image_attrs_enabled": {
+ "description": "ImageAttrsEnabled overrides the knowledge base's attribute-observed image\npipeline switch for this one document. nil keeps the knowledge base's\nsetting.",
+ "type": "boolean"
+ },
+ "image_vector_enabled": {
+ "description": "ImageVectorEnabled overrides image vector indexing for this document.\nIt does not change the KB-wide retrieval setting or embedding model.",
+ "type": "boolean"
+ },
"parser_engine_overrides": {
"description": "ParserEngineOverrides passes key-value configuration to docreader parsers\n(e.g. pdf_force_scanned=true). Merged with workspace-level overrides in the\nparse pipeline; per-upload values take priority on conflict.",
"type": "object",
@@ -20747,6 +22233,20 @@ const docTemplate = `{
"MatchTypeDataAnalysis"
]
},
+ "github_com_Tencent_WeKnora_internal_types.MatchedImage": {
+ "type": "object",
+ "properties": {
+ "chunk_id": {
+ "type": "string"
+ },
+ "knowledge_base_id": {
+ "type": "string"
+ },
+ "url": {
+ "type": "string"
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_types.Membership": {
"type": "object",
"properties": {
@@ -20864,7 +22364,7 @@ const docTemplate = `{
}
},
"artifacts": {
- "description": "Skill-generated files produced during this assistant turn (assistant messages only).\nPopulated by ArtifactCollector after the sandbox finishes, referenced by the\nartifact download endpoint. Empty for user messages and turns without skills.",
+ "description": "Skill-generated files produced during this assistant turn (assistant messages only).\nPopulated by ArtifactCollector after the sandbox finishes, referenced by the\nartifact download endpoint. Empty for user messages and turns without skills.\n\nStored in the message_artifacts table, not on the message row: the message\nrepository loads it with every message it returns and writes it whenever it\nis non-nil on create or update. A nil slice leaves the stored rows alone.",
"type": "array",
"items": {
"$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.MessageArtifact"
@@ -20990,6 +22490,10 @@ const docTemplate = `{
"description": "When WeKnora persisted the blob",
"type": "string"
},
+ "deleted_at": {
+ "description": "DeletedAt marks a file the user deleted. The entry stays in the list\nrather than being removed because its position IS the download address\n(msg.Artifacts[index]); dropping it would shift every later file's index\nand hand an old link the wrong blob. Keeping it also keeps the entry in\nArtifactCollector's de-duplication set, so the next collect does not\nre-attach the very file that was deleted — its sandbox mtime has not\nmoved. Clients filter these out; the bytes are already reclaimed.",
+ "type": "string"
+ },
"file_name": {
"description": "Original filename inside the sandbox",
"type": "string"
@@ -21159,6 +22663,14 @@ const docTemplate = `{
"description": "Provider identifier: openai, aliyun, zhipu, generic",
"type": "string"
},
+ "spec": {
+ "description": "Spec overrides the catalog entry for this row: protocol, capability\nflags and the flat protocol compat object. Everything omitted is\ninherited from the vendor catalog (see internal/models/catalog).",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ModelSpecOverride"
+ }
+ ]
+ },
"supports_vision": {
"description": "Whether the model accepts image/multimodal input",
"type": "boolean"
@@ -21252,6 +22764,40 @@ const docTemplate = `{
"ModelSourceAzureOpenAI"
]
},
+ "github_com_Tencent_WeKnora_internal_types.ModelSpecOverride": {
+ "type": "object",
+ "properties": {
+ "api": {
+ "type": "string"
+ },
+ "compat": {
+ "description": "Compat is a plain object so it round-trips through both JSON (API,\ndatabase column) and YAML (builtin_models.yaml).",
+ "type": "object",
+ "additionalProperties": {}
+ },
+ "context_window": {
+ "type": "integer"
+ },
+ "input": {
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ },
+ "max_output_tokens": {
+ "type": "integer"
+ },
+ "reasoning": {
+ "type": "boolean"
+ },
+ "thinking_levels": {
+ "type": "object",
+ "additionalProperties": {
+ "type": "string"
+ }
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_types.ModelType": {
"type": "string",
"enum": [
@@ -21283,6 +22829,17 @@ const docTemplate = `{
"ModelTypeASR"
]
},
+ "github_com_Tencent_WeKnora_internal_types.NamedCount": {
+ "type": "object",
+ "properties": {
+ "count": {
+ "type": "integer"
+ },
+ "name": {
+ "type": "string"
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_types.OBSEngineConfig": {
"type": "object",
"properties": {
@@ -21542,12 +23099,20 @@ const docTemplate = `{
"type": "string"
},
"mineru_model": {
- "description": "MinerU 自建解析参数",
+ "description": "以下 backend、vLLM、公式/表格、语言参数仅对 3.x 及更早版本生效。",
"type": "string"
},
"mineru_parse_method": {
"type": "string"
},
+ "mineru_server_api_key": {
+ "description": "MinerU 自建解析参数。协议按服务端自动识别:4.0+ 走 V1 API,更早版本走 /file_parse。",
+ "type": "string"
+ },
+ "mineru_tier": {
+ "description": "4.0+:flash/basic/standard/advanced,空为服务端默认",
+ "type": "string"
+ },
"mineru_vlm_server_url": {
"description": "vLLM 服务器地址 (vlm-http-client / hybrid-http-client)",
"type": "string"
@@ -21629,6 +23194,15 @@ const docTemplate = `{
"PromptCacheStatusHit"
]
},
+ "github_com_Tencent_WeKnora_internal_types.ProviderMetadata": {
+ "type": "object",
+ "additionalProperties": {
+ "type": "array",
+ "items": {
+ "type": "integer"
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_types.QuestionGenerationConfig": {
"type": "object",
"properties": {
@@ -21784,6 +23358,27 @@ const docTemplate = `{
}
}
},
+ "github_com_Tencent_WeKnora_internal_types.RerankOptions": {
+ "type": "object",
+ "properties": {
+ "enabled": {
+ "description": "Enabled turns the rerank stage off when explicitly false. Nil means on.",
+ "type": "boolean"
+ },
+ "model_id": {
+ "description": "ModelID selects the rerank model. Empty falls back to the tenant\nRetrievalConfig, then to the first rerank model of the tenant.",
+ "type": "string"
+ },
+ "threshold": {
+ "description": "Threshold is the minimum rerank model score. Nil uses the tenant\nRetrievalConfig. A pointer because 0 and negative scores are valid.",
+ "type": "number"
+ },
+ "top_k": {
+ "description": "TopK caps the reranked results. Zero uses the endpoint's result count.",
+ "type": "integer"
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_types.Resource": {
"type": "object",
"properties": {
@@ -22244,6 +23839,14 @@ const docTemplate = `{
"description": "QueryText is required unless query_embedding is provided, keyword matching is disabled,\nand vector matching remains enabled.",
"type": "string"
},
+ "rerank": {
+ "description": "Rerank, when present and enabled, reranks the fused candidates before\ntruncating to MatchCount. Only the hybrid-search API reads it; internal\ncallers rerank on their own.",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.RerankOptions"
+ }
+ ]
+ },
"scope_tag_ids": {
"type": "array",
"items": {
@@ -22344,6 +23947,13 @@ const docTemplate = `{
"description": "MatchedContent is the actual content that was matched in vector search\nFor FAQ: this is the matched question text (standard or similar question)",
"type": "string"
},
+ "matched_images": {
+ "description": "MatchedImages survives merging, deduplication and stored reference replay.\nIt contains storage references, never image bytes.",
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.MatchedImage"
+ }
+ },
"metadata": {
"description": "Metadata",
"type": "object",
@@ -22363,6 +23973,13 @@ const docTemplate = `{
"description": "Seq",
"type": "integer"
},
+ "source_locators": {
+ "description": "SourceLocators point back into the original file for citation\nnavigation. Merged results carry the union of their chunks' locators.",
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.SourceLocator"
+ }
+ },
"start_at": {
"description": "Start at",
"type": "integer"
@@ -22393,6 +24010,10 @@ const docTemplate = `{
"description": "ForkedFromMessageID is the user or assistant message, IN THE PARENT\nSESSION, that the fork branched at. For a user point, messages strictly\nbefore it were copied here. For an assistant point, that answer is\nincluded so the branch continues after it.",
"type": "string"
},
+ "host_workspace_dir": {
+ "description": "HostWorkspaceDir is the user-selected project directory this session\noperates in, for the WeKnora Lite host sandbox only. Empty means the session\ngets an auto-allocated workspace instead.\n\nWritten once at creation and never updated: re-pointing a live session\nat a different directory would leave its transcript describing files\nthat are no longer there. Standard edition never writes it.",
+ "type": "string"
+ },
"id": {
"description": "ID",
"type": "string"
@@ -22481,6 +24102,9 @@ const docTemplate = `{
"model_id": {
"type": "string"
},
+ "reasoning_effort": {
+ "type": "string"
+ },
"skill_names": {
"type": "array",
"items": {
@@ -22538,6 +24162,85 @@ const docTemplate = `{
}
}
},
+ "github_com_Tencent_WeKnora_internal_types.SourceLocator": {
+ "type": "object",
+ "properties": {
+ "bbox": {
+ "description": "BBox is [x0, y0, x1, y1] as fractions of the page width and height,\norigin at the top-left corner of the rendered page.",
+ "type": "array",
+ "items": {
+ "type": "number"
+ }
+ },
+ "block": {
+ "description": "Block is the 1-based index among a Word file's body paragraphs and tables.",
+ "type": "integer"
+ },
+ "end": {
+ "type": "integer"
+ },
+ "end_ms": {
+ "type": "integer"
+ },
+ "mapping": {
+ "description": "Mapping records verified text-to-source alignment. Empty means legacy.",
+ "type": "string"
+ },
+ "page": {
+ "description": "Page is the 1-based PDF page.",
+ "type": "integer"
+ },
+ "partial": {
+ "description": "Partial means some of the chunk has no source mapping.",
+ "type": "boolean"
+ },
+ "quote": {
+ "description": "Quote is the complete cited overlap as parsed. Viewers match it\ninside the rendered original and against the\nanswer sentence that cites the chunk.",
+ "type": "string"
+ },
+ "row_end": {
+ "type": "integer"
+ },
+ "row_start": {
+ "type": "integer"
+ },
+ "section": {
+ "description": "Section is the 1-based EPUB spine index.",
+ "type": "integer"
+ },
+ "sheet": {
+ "description": "Sheet is the spreadsheet sheet name; RowStart/RowEnd are its 1-based\nrow numbers as shown by spreadsheet applications.",
+ "type": "string"
+ },
+ "slide": {
+ "description": "Slide is the 1-based slide number in presentation order.",
+ "type": "integer"
+ },
+ "source_hash": {
+ "description": "SourceHash binds the locator to the knowledge record’s source revision hash.",
+ "type": "string"
+ },
+ "source_id": {
+ "description": "SourceID identifies a block within the parser result, not a rendered item.",
+ "type": "string"
+ },
+ "start": {
+ "description": "Start/End are rune offsets into the original text file.",
+ "type": "integer"
+ },
+ "start_ms": {
+ "description": "StartMs/EndMs bound an audio segment.",
+ "type": "integer"
+ },
+ "title": {
+ "description": "Title names the section or slide when the source has one.",
+ "type": "string"
+ },
+ "type": {
+ "type": "string"
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_types.StarterSuggestionConfig": {
"type": "object",
"properties": {
@@ -23418,6 +25121,14 @@ const docTemplate = `{
"description": "BrowserSearchInstructions customizes browser search for this user. Nil/empty uses the platform default.",
"type": "string"
},
+ "gallery": {
+ "description": "Gallery holds the user's image-gallery UI state: the search\nactivation mode and the per-attribute search-field toggles. Written\nby the gallery's mode / checkbox interactions via the preferences\nPUT; read back by GET /knowledge-bases/:id/gallery-config so the\ngallery renders the same activation state on every device. Nil = the\nuser never touched the gallery controls (mode defaults to \"all\").",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.GalleryUserPrefs"
+ }
+ ]
+ },
"last_active_tenant_id": {
"description": "LastActiveTenantID remembers the last workspace the user actively\nswitched into, so a fresh login (new device, cleared browser, new\nrefresh token) lands them back in that workspace instead of always\nbouncing to their home workspace. Written by the SPA's preferences\nPUT and by service-level SwitchTenant (including when switching\nhome, which stores the home ID). Login / RefreshToken validate that\nthe workspace still exists and the user still has an active membership\n(or CanAccessAllTenants) before honouring this preference; an\ninvalid pointer is best-effort cleared and the user falls back to\nhome. Refresh JWT claims have no tenant_id, so RefreshToken\nre-resolves from this field.\n\nnil = no preference (use user.TenantID, i.e. home)\n*0 = \"clear preference\" sentinel for the partial-update endpoint\n (UpdateUserPreferences turns this into nil). Otherwise treat\n a stored *0 the same as nil.\n*N = preferred workspace id.",
"type": "integer"
@@ -24602,6 +26313,17 @@ const docTemplate = `{
}
}
},
+ "internal_handler.CopyModelRequest": {
+ "type": "object",
+ "required": [
+ "display_name"
+ ],
+ "properties": {
+ "display_name": {
+ "type": "string"
+ }
+ }
+ },
"internal_handler.CreateAgentRequest": {
"type": "object",
"required": [
@@ -25167,6 +26889,26 @@ const docTemplate = `{
}
}
},
+ "internal_handler.ListAllUsersResponse": {
+ "type": "object",
+ "properties": {
+ "has_more": {
+ "type": "boolean"
+ },
+ "limit": {
+ "type": "integer"
+ },
+ "offset": {
+ "type": "integer"
+ },
+ "users": {
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.UserInfo"
+ }
+ }
+ }
+ },
"internal_handler.ListSystemAdminsResponse": {
"type": "object",
"properties": {
@@ -25229,6 +26971,9 @@ const docTemplate = `{
"description": "为空时按需默认为 \"remote\"",
"type": "string"
},
+ "spec": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ModelSpecOverride"
+ },
"supportsDimensionOverride": {
"type": "boolean"
}
@@ -25497,6 +27242,9 @@ const docTemplate = `{
"description": "为空时按需默认为 \"remote\"",
"type": "string"
},
+ "spec": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ModelSpecOverride"
+ },
"supportsDimensionOverride": {
"type": "boolean"
}
@@ -25729,6 +27477,44 @@ const docTemplate = `{
}
}
},
+ "internal_handler.SearchWikiRequest": {
+ "type": "object",
+ "required": [
+ "query"
+ ],
+ "properties": {
+ "knowledge_base_id": {
+ "type": "string"
+ },
+ "knowledge_base_ids": {
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ },
+ "limit": {
+ "type": "integer"
+ },
+ "query": {
+ "type": "string"
+ }
+ }
+ },
+ "internal_handler.SetUserStatusRequest": {
+ "type": "object",
+ "required": [
+ "is_active",
+ "user_id"
+ ],
+ "properties": {
+ "is_active": {
+ "type": "boolean"
+ },
+ "user_id": {
+ "type": "string"
+ }
+ }
+ },
"internal_handler.StorageCheckRequest": {
"type": "object",
"properties": {
@@ -25908,6 +27694,14 @@ const docTemplate = `{
},
"name": {
"type": "string"
+ },
+ "vlm_config": {
+ "description": "VLMConfig updates the knowledge base's multimodal (vision) config.\nOptional: nil means \"no change\", mirroring how CreateKnowledgeBase\naccepts a top-level vlm_config.",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.VLMConfig"
+ }
+ ]
}
}
},
@@ -26146,6 +27940,64 @@ const docTemplate = `{
}
}
},
+ "internal_handler.imageListRequestBody": {
+ "type": "object",
+ "properties": {
+ "attr_filters": {
+ "type": "object",
+ "additionalProperties": {
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ }
+ },
+ "attr_rules": {
+ "type": "object",
+ "additionalProperties": {
+ "type": "object",
+ "additionalProperties": {
+ "type": "string"
+ }
+ }
+ },
+ "is_enabled": {
+ "type": "boolean"
+ },
+ "keyword": {
+ "type": "string"
+ },
+ "knowledge_ids": {
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ },
+ "page": {
+ "description": "Page",
+ "type": "integer",
+ "minimum": 1
+ },
+ "page_size": {
+ "description": "Page size",
+ "type": "integer",
+ "maximum": 1000,
+ "minimum": 1
+ },
+ "search_in": {
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ },
+ "sort_by": {
+ "type": "string"
+ },
+ "sort_order": {
+ "type": "string"
+ }
+ }
+ },
"internal_handler.invitationLookupRequest": {
"type": "object",
"required": [
@@ -26451,6 +28303,10 @@ const docTemplate = `{
}
]
},
+ "reasoning_effort": {
+ "description": "ReasoningEffort overrides thinking for this request; empty inherits the agent configuration.",
+ "type": "string"
+ },
"skill_names": {
"description": "Per-request Skills selected via @mention",
"type": "array",
@@ -26485,6 +28341,10 @@ const docTemplate = `{
"description": "Description for the session (optional)",
"type": "string"
},
+ "project_dir": {
+ "description": "ProjectDir is an optional Lite host-sandbox binding. When set it must\nbe an absolute path already present in the user-approved ProjectDirs\nlist. Empty means the session gets an auto-allocated workspace.",
+ "type": "string"
+ },
"title": {
"description": "Title for the session (optional)",
"type": "string"
@@ -26574,12 +28434,36 @@ const docTemplate = `{
}
}
},
+ "internal_handler_session.RewindSessionRequest": {
+ "type": "object",
+ "required": [
+ "message_id"
+ ],
+ "properties": {
+ "message_id": {
+ "description": "MessageID is the message to rewind to. A user message deletes itself and\neverything after (the client prefills that question). An assistant\nmessage keeps itself and deletes everything after.",
+ "type": "string"
+ }
+ }
+ },
"internal_handler_session.SearchKnowledgeRequest": {
"type": "object",
"required": [
"query"
],
"properties": {
+ "disable_keywords_match": {
+ "description": "Vector recall only",
+ "type": "boolean"
+ },
+ "disable_vector_match": {
+ "description": "Keyword recall only",
+ "type": "boolean"
+ },
+ "keyword_threshold": {
+ "description": "Minimum keyword score",
+ "type": "number"
+ },
"knowledge_base_id": {
"description": "Single knowledge base ID (for backward compatibility)",
"type": "string"
@@ -26598,6 +28482,10 @@ const docTemplate = `{
"type": "string"
}
},
+ "match_count": {
+ "description": "Number of results to return",
+ "type": "integer"
+ },
"mentioned_items": {
"description": "Optional scoped tag mentions",
"type": "array",
@@ -26609,12 +28497,24 @@ const docTemplate = `{
"description": "Query text to search for",
"type": "string"
},
+ "rerank": {
+ "description": "Rerank override",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.RerankOptions"
+ }
+ ]
+ },
"tag_ids": {
"description": "Tag IDs for filtering within a single KB",
"type": "array",
"items": {
"type": "string"
}
+ },
+ "vector_threshold": {
+ "description": "Optional overrides of the tenant retrieval config. Omitted fields keep it.",
+ "type": "number"
}
}
},
diff --git a/docs/release-v0.8.3-audit.md b/docs/release-v0.8.3-audit.md
new file mode 100644
index 00000000..44149b47
--- /dev/null
+++ b/docs/release-v0.8.3-audit.md
@@ -0,0 +1,367 @@
+# v0.8.3 发布整理与文档核对记录
+
+## 基线与范围
+
+- 上个版本:`v0.8.2`(`3e8b0bfc80b845b2d4b2ed683994748741450a97`)。
+- 本次远程 main 截止:`31c6055687d33f9ad7a92cf453a264ce13a1700d`,2026-10-13 拉取,共 282 个提交。
+- 初始整理分支:`chore/release-v0.8.3-docs`;发布 PR 分支:`chore/release-v0.8.3-website`。2026-10-14 提取到最新 `github_public/main`(`3b3b519310721295162ea94b9fd62a92c8cab341`);相较盘点快照增加的 #4278 和 #4283 为基线已有修复,不属于本 PR 的新增代码。
+- 版本选择:延续当前 0.8 系列使用下一个未占用版本 0.8.3;存在需要迁移的行为变化,已单列升级说明。发布日期确定为 2026-10-14;本 PR 更新发布文档与版本元数据,tag 和 GitHub Release 由发布流程创建。
+- 审查对象是版本净变更对应的用户文档、配置/API 契约与发布内容;不是对 282 个提交重新进行完整安全审计。最终生效行为以该 main 快照的源码为准,提交标题只用于完整盘点。
+- 插件系统的 11 个引入提交由 `6c83d7ab4` 撤回,没有作为新功能发布。DeepSeek Harness 等已有外部客户端集成不受此结论影响。
+
+## 已修正与补齐
+
+| 主题 | 原有偏差或遗漏 | 核对源码与处理 |
+| --- | --- | --- |
+| 会话附件 | 未区分解析与读取时机、所有附件 24 小时过期;Agent 说明把阅读绑定到沙箱 | `temporary_document.go`、`temporary_document_prompt.go`、`attachment/reader.go`、repository 的 `expiredQuery`;上传后后台解析、Agent 按需读取并等待、Quick Answer 开流后等待、有效消息引用保留,覆盖会话/Agent/API/配置文档 |
+| 工具选择 | 空数组回退默认、智能推理至少选一项 | `tools/allowed.go`、`custom_agent_tools.go`、迁移 116/35;区分 null 与 [],补工具目录 API、未知名称与开关管理项处理 |
+| 注册策略 | 快速上手、总览、配置与系统设置仍将 invite_only 称作邀请注册 | config/handler 与 `auth.registration_mode`;统一三档语义,单列升级动作 |
+| 用户管理 | 没有用户列表/启停 API,创建用户入口仍在账户与访问 | `SystemSettings.vue`、`UserManagement.vue`、`system.go`、router;补列表分页、停用撤销会话、保护条件与新入口 |
+| 画廊 | 只有底层图片处理参数,没有用户流程和完整 API | `knowledge.go`、`routes_knowledge.go`、`types/image_gallery.go`;补画廊流程、属性契约、GET/POST 过滤与不同 ID 上限 |
+| 图片检索 | 开关要求在长篇底层说明中,缺少检索入口概述 | `knowledgebase_search_image.go`、图片索引与重排/回答链路;区分画廊、图片向量与文本描述,明确默认关闭和重解析 |
+| 图片观察 | 将兜底误写为任何观察失误都不会丢内容 | `image_attr_decision.go`;说明 on_unobserved 可关闭兜底,完整但错误的负值观察也可能跳过 OCR |
+| 模型 | 已加入 Cohere 但仍计数 29,扩展指南遗漏 TEI/Pinecone 协议 | providers 的 Builtins/测试明确 30;更新数字与协议列表,补 WeMM 默认会话批处理说明 |
+| 数据库 | 总清单仍停在 PG 110 / SQLite 30,并声称 SQLite 无 MCP metadata | 迁移文件 111–118 / 31–36;补投影回填、索引、OAuth 时间戳,修正默认镜像和附件状态 |
+| MCP 授权 | 文档声称授权完成关闭整个服务连接 | `CloseClientsAfterAuthorization`;当前主体与旧 client_id 连接清理,补读取注册失败的保守回退,避免夸大为永不影响其他用户 |
+| 本机技能 | 只介绍本机命令执行,漏掉安装及激活版本 | `tenant_skill_host.go`、`host_skills.go`;补本机安装目标、版本切换与失败保留旧版 |
+| Wiki/引用/截图 | 旧图不能展示新分组和原文定位 | 更新说明,接入真实产品截图并核对图注;保留历史版本图片 |
+| 配置 | 遗漏知识库描述预算和出站拨号超时 | `.env.example`、config、描述生成服务;补 2048 token 与 5s 默认,区分拨号和总请求超时 |
+| 发布入口 | 首页、版本导航、徽章仍指 v0.8.2 | v0.8.3 页面、首页画廊、四种 README、共享导航、检查脚本与当前版本元数据统一;历史说明与依赖独立版本保留 |
+
+## 首页长期内容调整
+
+- 中文功能名与前端一致使用「画廊」;画廊与多模态 Embedding 在首页、版本说明、知识库指南及四语 README 分开介绍。
+- 多模态 Embedding 先说明对这一模型类型的支持,以腾讯 WeMM-Embedding、Qwen3-VL-Embedding 等作为已支持的示例;WeMM 的具体预设和部署方式在详情展开,保留启用条件与现有部署文档入口。
+- 本机浏览器使用与版本号无关的独立图文区并复用真实截图;MCP Server 归入生态集成,对话追加、分叉与回滚归入 Agent 概述,不再等权展示。技能与沙箱、Wiki 保留专题,Lite 本机项目目录补入沙箱说明。
+- 语言新增等较小更新不再占据首页版本补充栏,仍在版本说明保留。首页模型厂商数修正为 30。
+
+## 发布表达与历史约定
+
+参考 2026-09-23 Claude 的「v0.8.0 到 v0.8.2 文档同步检查」对话和 v0.8.2 版本页:
+
+- 首页重点说明用途,简短文案配大图;配置前提与协议细节放入功能文档。
+- 当期版本区保留四项重点:画廊、多模态 Embedding、Wiki 图谱、原文引用定位,采用左侧紧凑功能导航与右侧截图,四项说明均保持展开,优先让说明与图片同屏;附件按需读取、平台用户管理、Seafile 与 Docker 终端使用次级链接。
+- 首页 01–07 及使用入口统一背景交替、全宽背景、内容容器和章节间距;浅色背景区统一提供分隔线,深色模式复用主题变量。
+- 本机浏览器保留独立专题,恢复「执行任务/连接扩展」两张图、BrowserSkill 图标与 GitHub 入口,左侧分别展开「执行任务」「连接扩展」的说明,右侧切换对应截图,与技能沙箱之间补齐边界和背景区分。
+- 版本页按用户可以做什么组织,保留升级操作与限制;提交数量、快照、撤回记录在本报告保留,完整技术更新仍见 CHANGELOG。
+
+- Wiki 图谱渲染与交互调整单列版本章节;核对 #4147 的分组布局、主题说明、标签避让、分组切换和返回全图,首页 Wiki 专题首项使用分组图谱实拍截图。
+
+## 第二轮整体复核(2026-10-14)
+
+- 再次按 v0.8.2 至既定 main 快照的 282 个提交核对首页与版本页,未改变发布基线。
+- 首页生态列表补 Seafile,复用前端现有图标;技能专题补 macOS Lite 本机安装。明确 Docker/E2B/Cube 均有终端,图形桌面只适用于 E2B/Cube 桌面模板。
+- 版本页补共享智能体操作调用者本机浏览器、工具列表搜索、检索参数与诊断、渠道会话管理、模型完成事件的实际模型 ID。跨库 Wiki 搜索明确为 API 能力,放回 Wiki 章节。
+- 补扫描 PDF 整页不生成图片向量、原文引用的访问权限与无原文件回退、Seafile 单个未加密资料库及不支持 SeaDoc 的边界。
+- 修正 CHANGELOG 将 #3950 的实体知识图谱方向/来源修复归到 Wiki 的错误;两类图谱分别说明。
+- Swagger 原产物遗漏 10 个现有路径;使用与 go.mod 一致的 swag v1.16.6 重新生成。补模型目录读取/预览/发布的 3 个接口注释(2 个路径),正确声明仅管理员 JWT、版本冲突与 JSON overlay 对象;修复 spans 注释重复 API 基路径及缺失鉴权声明。三种产物均由 `make docs` 生成,不手改产物。
+- 本次新增后端改动仅为 Swagger 注释及 schema 标签,不改接口行为。契约检查覆盖新发布接口、基路径不得重复、overlay 对象类型。
+
+## 发布内容与截图
+
+- 面向用户的版本说明:`website-docs/07-releases/v0.8.3.md`。
+- 英文分类更新清单:根目录 `CHANGELOG.md`;以下附录保留全部提交,可追溯未单独展开的测试/构建/修正文档提交。
+- 截图清单:`website-docs/README.md` 的「v0.8.3 截图清单」,8 个本次场景 + 1 个既有未补齐场景;4 个首页场景复用对应文件。
+- 已接入用户提供的 11 张真实截图并逐张检查,画廊、多模态设置、Wiki 分组图谱、引用定位均已同步到首页。画廊截图保留了原始界面中的部分图片加载失败状态;图注已按实际展示内容调整。未合成产品界面、未执行浏览器外观验收。
+
+## 验证记录
+
+- 使用 Node 24.19.0 执行 `npm run build --prefix website-docs`,成功完成 Next.js 首页与 VitePress 文档构建、内部链接检查、60 个 Mermaid 图检查(0 失败)、首页内容审计和统一站点检查(77 页、376 个本地路由/资源)。
+- `npm run lint --prefix website-docs/homepage` 通过;`git diff --check` 通过。发布分支的 `docs`、`internal/handler`、`internal/application/service` 相对 `github_public/main` 的增量 Go lint 为 0 issues。
+- `go test ./docs` 通过,覆盖 Go 注册文档、swagger.json 和 swagger.yaml 的路由/schema 契约;再次执行 `make docs` 后三份产物逐字节一致。
+- 根 VERSION、文档 VERSION、前端/文档 package.json 与 lockfile 根版本、Helm appVersion 均核对为 0.8.3;`scripts/get_version.sh json` 输出 0.8.3。两个 lockfile 除根版本字段外与 main 完全一致,没有依赖变更。
+- 校验附录包含全部 282 个提交,PostgreSQL 迁移文件为 238 个;历史版本页未修改。
+- 首页 `prebuild` / `predev` 自动同步复用截图,更新文档源图后无需手动复制;同步回归覆盖替换源图、终端文件名映射及保留首页独有图片。预览中的 Wiki 图片已按 HTTP 响应逐字节核对为最新源图,顶部留白收紧,WeMM 名称链接至官方仓库。
+- 静态站点产物位于 `website-docs/static-site/`。11 张文档截图及首页复用的 5 张截图均已补齐;未执行浏览器外观验收或 Docker 部署测试。
+- 本 PR 仅包含文档、官网内容、截图、版本元数据与 Swagger 注释/生成物。附件读取修复已单独通过 #4283 合入,本 PR 不重复携带其代码或测试。
+- 未运行全量 Go/前端业务回归或实际数据库升级。迁移说明依据当前源文件核对,不能替代生产备份与升级演练。
+- 依赖安装提示已有依赖漏洞,未在本次发布文档整理中升级依赖;构建存在依赖包 chunk-size 提示,不影响构建结果。
+
+## 全部提交(时间顺序)
+
+「已撤回」不进入有效功能列表;「构建/测试/文档」仍保留在盘点中。其余按主题归入 CHANGELOG 的新增、改进、修复和升级事项。
+
+| 日期 | 提交 | 内容 | 处理 |
+| --- | --- | --- | --- |
+| 2026-09-24 | [cee697920](https://github.com/Tencent/WeKnora/commit/cee697920e0aebf067b57cf9612aa330863b17c3) | fix(website-docs): serve everything but the homepage from /docs/ | 有效变更,按主题归纳 |
+| 2026-09-24 | [b6b32c24e](https://github.com/Tencent/WeKnora/commit/b6b32c24e152650cf94083c21195fe8bd24328f5) | test(website-docs): update integration test for the v0.8.2 header and release gallery | 构建/测试/文档或体验调整 |
+| 2026-09-24 | [8d4015014](https://github.com/Tencent/WeKnora/commit/8d40150149ce00a952a98a10a80226ad6584abd5) | fix(session): keep auto-generated titles plain text | 有效变更,按主题归纳 |
+| 2026-09-24 | [967ed0970](https://github.com/Tencent/WeKnora/commit/967ed097062c61723045ae103b665982728dc572) | feat(models): add system admin model catalog console (#3662) | 有效变更,按主题归纳 |
+| 2026-09-24 | [eebc1d864](https://github.com/Tencent/WeKnora/commit/eebc1d86436e9cd7b9f6d24c1f973527f318853f) | fix(ui): restore inner padding on form popups (#3673) | 有效变更,按主题归纳 |
+| 2026-09-24 | [e6284f4ea](https://github.com/Tencent/WeKnora/commit/e6284f4eacc89a4b2f1ef2469e9f86241e2841cd) | fix(browser): let shared-space agents drive the caller's local browser | 有效变更,按主题归纳 |
+| 2026-09-24 | [5d00985f7](https://github.com/Tencent/WeKnora/commit/5d00985f720440f88c88ccdf8f65232175d885d5) | docs: restructure the English and Chinese READMEs | 构建/测试/文档或体验调整 |
+| 2026-09-24 | [44a1df604](https://github.com/Tencent/WeKnora/commit/44a1df60460fc06fba9eed9022d57164605a487c) | docs(readme): website-matched design for all four READMEs (#3679) | 构建/测试/文档或体验调整 |
+| 2026-09-24 | [65d8129cb](https://github.com/Tencent/WeKnora/commit/65d8129cb64f4ccc23a7b73896ebe424d31ba228) | fix(retriever/es): keep vectors and enabled state on CopyIndices (#3678) | 有效变更,按主题归纳 |
+| 2026-09-24 | [3cf0be2e9](https://github.com/Tencent/WeKnora/commit/3cf0be2e9745b97dd903feec1c66a936780237d8) | fix(embed): warn at startup when no signing key is configured | 有效变更,按主题归纳 |
+| 2026-09-24 | [e22a8367e](https://github.com/Tencent/WeKnora/commit/e22a8367efc9802353e6981e162226459a998404) | feat(docparser): support MinerU 4.0 V1 API for self-hosted MinerU | 有效变更,按主题归纳 |
+| 2026-09-24 | [9253b2c1d](https://github.com/Tencent/WeKnora/commit/9253b2c1dc8f4fe83fec8e66ba79dbf00144fd43) | fix(frontend): link docs to the public docs site instead of GitHub | 有效变更,按主题归纳 |
+| 2026-09-24 | [9fc08c207](https://github.com/Tencent/WeKnora/commit/9fc08c2079d89173c17c61356dee3838772931c5) | feat(retrieval): unify rerank across search APIs and expose diagnostics (#3675) | 有效变更,按主题归纳 |
+| 2026-09-24 | [65d8a34ea](https://github.com/Tencent/WeKnora/commit/65d8a34ead4c99991d450ecc8c0acdbcf5b0e583) | fix(desktop): pair the Lite browser against the local API listener (#3677) | 有效变更,按主题归纳 |
+| 2026-09-24 | [c029b50ef](https://github.com/Tencent/WeKnora/commit/c029b50ef928d5c21ffb1922f6d541b0d27bb7a4) | fix(rss): 拒绝超限响应以避免正文静默截断 | 有效变更,按主题归纳 |
+| 2026-09-24 | [7e82ee33d](https://github.com/Tencent/WeKnora/commit/7e82ee33dcefc398ca455775700bdf8730054aa5) | docs(misc): add intro video source | 构建/测试/文档或体验调整 |
+| 2026-09-24 | [e2433b129](https://github.com/Tencent/WeKnora/commit/e2433b1293fa350d972902cd3676a1aa9a9c9aac) | feat(lite): install and run skills on the host (#3661) | 有效变更,按主题归纳 |
+| 2026-09-25 | [e896270c4](https://github.com/Tencent/WeKnora/commit/e896270c418c01d230c9c6d11f05d0148677892a) | docs(readme): surface agent tools in the overview and condense feature tables (#3688) | 构建/测试/文档或体验调整 |
+| 2026-09-25 | [04adc1327](https://github.com/Tencent/WeKnora/commit/04adc132726d93c9247898af79ad8f3b542df4f0) | fix(retrieval): fix recall correctness, score scales, cost and agent tools (#3687) | 有效变更,按主题归纳 |
+| 2026-09-25 | [68fe9599c](https://github.com/Tencent/WeKnora/commit/68fe9599cdb853eb1205499d44ed176613f875df) | feat(gallery): add an image gallery tab with schema-driven filters (#3694) | 有效变更,按主题归纳 |
+| 2026-09-25 | [ebbcf96fa](https://github.com/Tencent/WeKnora/commit/ebbcf96fabb651c091c3ee2272d802b80c100c4e) | feat(gallery): rework the gallery toolbar, header tabs and image viewer | 有效变更,按主题归纳 |
+| 2026-09-25 | [1bf717a54](https://github.com/Tencent/WeKnora/commit/1bf717a54d4413e0e5cfe5e2b7ff008489b97c04) | feat(chat): open cited documents at the cited passage (#3700) | 有效变更,按主题归纳 |
+| 2026-09-25 | [ab115c094](https://github.com/Tencent/WeKnora/commit/ab115c0944b6e7f3c5d976970b67cb6070f6db1b) | fix(knowledge): stop parses from getting stuck in processing or finalizing (#3696) | 有效变更,按主题归纳 |
+| 2026-09-25 | [8619bdf3c](https://github.com/Tencent/WeKnora/commit/8619bdf3c5149acc8ca51263e4d485f97da39211) | docs(readme): swap in the new intro video | 构建/测试/文档或体验调整 |
+| 2026-09-25 | [ef6f28c1c](https://github.com/Tencent/WeKnora/commit/ef6f28c1cafec72bd05fcb4fbd3feb8144f93e3d) | fix(agent): recover tool panics instead of crashing the server | 有效变更,按主题归纳 |
+| 2026-09-25 | [8356742dd](https://github.com/Tencent/WeKnora/commit/8356742dd5a75cc84a5f18a527665eac4abf5bea) | fix(llm): treat cut-off streams as broken and tighten LLM retries | 有效变更,按主题归纳 |
+| 2026-09-25 | [ef938f911](https://github.com/Tencent/WeKnora/commit/ef938f9114cc82e799c56b5bde2b3aca9bc3832c) | fix(logger): keep LLM debug records inside their directory | 有效变更,按主题归纳 |
+| 2026-09-25 | [8a75bece6](https://github.com/Tencent/WeKnora/commit/8a75bece6d491a21e8b7b4c6171a8151ad77da9b) | fix(mcp): close MCP connections on shutdown without blocking exit | 有效变更,按主题归纳 |
+| 2026-09-25 | [27b0c84bf](https://github.com/Tencent/WeKnora/commit/27b0c84bfdb2641c66942aea4620f44b328096e1) | fix(docparser): make the MinerU parse timeouts configurable | 有效变更,按主题归纳 |
+| 2026-09-25 | [929dad574](https://github.com/Tencent/WeKnora/commit/929dad5747fb24e3cc85ec86eb1f4b4a3a497441) | fix(weaviate): accept a class another worker created first (#3689) | 有效变更,按主题归纳 |
+| 2026-09-25 | [605d70fc5](https://github.com/Tencent/WeKnora/commit/605d70fc528a948591b647b0b805e7bfd617d4be) | fix(notion): 在资源发现失败时中止增量同步 (#3682) | 有效变更,按主题归纳 |
+| 2026-09-25 | [5614c04d8](https://github.com/Tencent/WeKnora/commit/5614c04d88f2c28620a6bd3e16fc1093b9c54539) | fix(feishu): stop Wiki pagination on repeated page tokens (#3645) | 有效变更,按主题归纳 |
+| 2026-09-25 | [6eb3cce16](https://github.com/Tencent/WeKnora/commit/6eb3cce164f6b1efe9c3a02973502c93f7d45126) | fix(agent): 合并相邻 assistant 消息时保留 tool_calls,避免孤儿 tool 消息导致整轮 400 (#3574) | 有效变更,按主题归纳 |
+| 2026-09-25 | [e5a301ed3](https://github.com/Tencent/WeKnora/commit/e5a301ed3ad156d7139231b43f7664e05eea6806) | fix(files): accept canonical storage:// paths in bare provider services (#3166) | 有效变更,按主题归纳 |
+| 2026-09-25 | [5c717ca04](https://github.com/Tencent/WeKnora/commit/5c717ca045ab6742b5bb57dc1669a54330f6b65b) | fix(im): handle unresolved Feishu resource images (#2761) | 有效变更,按主题归纳 |
+| 2026-09-25 | [591f1c0e4](https://github.com/Tencent/WeKnora/commit/591f1c0e4fb70a9ece1d807b19c6fa0c24e6b102) | docs: drop references to source files that no longer exist (#3631) | 构建/测试/文档或体验调整 |
+| 2026-09-25 | [d9444683d](https://github.com/Tencent/WeKnora/commit/d9444683dae2f382e29655f3efc6cc156d199329) | docs(dev): 补充 Swagger 容器访问与空白页排查步骤 (#3620) | 构建/测试/文档或体验调整 |
+| 2026-09-25 | [0cb1e57fa](https://github.com/Tencent/WeKnora/commit/0cb1e57fae586908ca8d005459eb82acf54ca015) | fix(container): propagate host project lookup errors (#3611) | 有效变更,按主题归纳 |
+| 2026-09-25 | [4668447aa](https://github.com/Tencent/WeKnora/commit/4668447aa60b5190f90a8a53546151df4ec4ed11) | fix(helm): pin application images to chart version (#3669) | 有效变更,按主题归纳 |
+| 2026-09-25 | [22404e92b](https://github.com/Tencent/WeKnora/commit/22404e92b79dfc3ae7d60342ba16b0fd37364b30) | style(files): wrap long test case to satisfy lll | 构建/测试/文档或体验调整 |
+| 2026-09-25 | [508048aee](https://github.com/Tencent/WeKnora/commit/508048aee9739e4d034ab530d6a25603a6d9042d) | fix(frontend): mount the settings modal only once on /platform/settings | 有效变更,按主题归纳 |
+| 2026-09-25 | [745a22d06](https://github.com/Tencent/WeKnora/commit/745a22d066cab0dc555372a7510c017c5ddb9c6a) | refactor(frontend): drop TTL resource caches and remove duplicate startup requests (#3562) | 构建/测试/文档或体验调整 |
+| 2026-09-25 | [c7601b2d2](https://github.com/Tencent/WeKnora/commit/c7601b2d2aaf298939b5ce775970c3005044d74a) | feat(wiki): 增加跨知识库 Wiki 搜索 POST /wiki-search (#3535) | 有效变更,按主题归纳 |
+| 2026-09-25 | [a9d71038b](https://github.com/Tencent/WeKnora/commit/a9d71038b6657fda5e01a85ad085b861c2021758) | fix(models): stop advertising graded thinking levels the wire format drops (#3552) | 有效变更,按主题归纳 |
+| 2026-09-25 | [00e6366fd](https://github.com/Tencent/WeKnora/commit/00e6366fd27a42bca7b08d6fef5b909339bf6be9) | fix(datasource): validate cron schedules and persist manual-only mode (#3653) | 有效变更,按主题归纳 |
+| 2026-09-25 | [0bf105499](https://github.com/Tencent/WeKnora/commit/0bf105499b093578ddbaabf24403ba0aa776657a) | fix(knowledge): reject invalid JSON uploads before async parse (#3360) | 有效变更,按主题归纳 |
+| 2026-09-25 | [a7fc9e13d](https://github.com/Tencent/WeKnora/commit/a7fc9e13dab91cf5bfc3d33c0f09415879957868) | fix(client): accumulate multiline SSE data (#2122) | 有效变更,按主题归纳 |
+| 2026-09-25 | [e7c27ccc9](https://github.com/Tencent/WeKnora/commit/e7c27ccc9e7fc806c583f6bb5de012cba12e5c53) | feat(confluence): add directory sync and private image assets (#3319) | 有效变更,按主题归纳 |
+| 2026-09-25 | [a7ac21ce7](https://github.com/Tencent/WeKnora/commit/a7ac21ce769fa75854c0b7c7cfb04d934464fa73) | fix(docreader): sanitize non-conforming styles.xml fills before openpyxl loads (#3637) (#3638) | 有效变更,按主题归纳 |
+| 2026-09-25 | [1dd3fe867](https://github.com/Tencent/WeKnora/commit/1dd3fe867e0d343df765961eb090ae0c72a3c031) | fix(docreader): drop data validation ranges openpyxl cannot parse instead of failing the sheet (#3599) (#3608) | 有效变更,按主题归纳 |
+| 2026-09-25 | [4364e61af](https://github.com/Tencent/WeKnora/commit/4364e61afa4bf1e086c34b4220bb21449aa15b25) | feat(datasource/yuque): preserve TOC folder hierarchy for synced documents (#3476) | 有效变更,按主题归纳 |
+| 2026-09-26 | [47773b5bd](https://github.com/Tencent/WeKnora/commit/47773b5bd1fc33ecbe96d11824d4aa4b97ae2438) | fix(frontend): initialize chat citation popover after botmsg props | 有效变更,按主题归纳 |
+| 2026-09-26 | [c91a8d11f](https://github.com/Tencent/WeKnora/commit/c91a8d11f8eaa1370970355019a001a36e993117) | feat(plugin): plugin system foundation — builtin plugins, config schemas and workspace switches (#3711) | 已撤回 |
+| 2026-09-26 | [ed63e35cc](https://github.com/Tencent/WeKnora/commit/ed63e35cc7d49856c562b9f519534cba6d6b9ee8) | feat(plugin): install declarative plugins — MCP servers, skills and model vendors (#3716) | 已撤回 |
+| 2026-09-26 | [4328ab26a](https://github.com/Tencent/WeKnora/commit/4328ab26a942a3fb9c397dbea57b41ac1074cc46) | feat(plugin): code plugins — extension protocol, Go SDK, plugin host, web search and data source adapters (#3720) | 已撤回 |
+| 2026-09-26 | [ff49f9491](https://github.com/Tencent/WeKnora/commit/ff49f9491d6b73de5adea18f061a7d04073cbd44) | feat(plugin): Host API with a key-value store, and document parsers from plugins (#3721) | 已撤回 |
+| 2026-09-26 | [728007ada](https://github.com/Tencent/WeKnora/commit/728007ada9df5461d1f70aca64ec358718b891bd) | feat(plugin): remote runtime, Python SDK and python host plugins (P2 batch 3) (#3724) | 已撤回 |
+| 2026-09-26 | [da4ef68d9](https://github.com/Tencent/WeKnora/commit/da4ef68d9d5327dfd236299d882f092943a1a165) | feat(plugin): standalone plugin hosts with Redis discovery (P2 batch 4) (#3725) | 已撤回 |
+| 2026-09-26 | [996696057](https://github.com/Tencent/WeKnora/commit/996696057f7ec229fd1930188955283042f0c852) | feat(plugin): sandboxed plugin pages and the @weknora/plugin-ui bridge (P2 batch 5) (#3727) | 已撤回 |
+| 2026-09-26 | [cf502ba46](https://github.com/Tencent/WeKnora/commit/cf502ba46a629598db393d747be05da3df682c18) | feat(plugin): event delivery and webhooks for plugins (P2 batch 6) (#3728) | 已撤回 |
+| 2026-09-26 | [01222aac8](https://github.com/Tencent/WeKnora/commit/01222aac8e884f9a14db531072c6074f059beb4a) | feat(plugin): dynamic form options, OAuth connections and resource limits (P2 batch 7) (#3729) | 已撤回 |
+| 2026-09-26 | [4f5d475c3](https://github.com/Tencent/WeKnora/commit/4f5d475c3eebe47d7c0e921799bfdb8337f6a024) | feat(examples): Jira Cloud plugin with OAuth, dynamic options and a skill (#3730) | 已撤回 |
+| 2026-09-26 | [a829d7cc9](https://github.com/Tencent/WeKnora/commit/a829d7cc9cc3ac96c3cd1bbd6210203966c97d56) | feat(plugin): agent tools served by plugins, with result views (P2 batch 8) (#3733) | 已撤回 |
+| 2026-09-26 | [6c83d7ab4](https://github.com/Tencent/WeKnora/commit/6c83d7ab479212c136094182a68ac6e238ea370c) | revert: take the plugin system back out of main | 撤回插件系列 |
+| 2026-09-26 | [9fcc2c674](https://github.com/Tencent/WeKnora/commit/9fcc2c674309b42a1d4077e6b4c2a2334d49e57c) | fix(dev): propagate dev.sh exit status and pull MinIO from pgsty/minio | 有效变更,按主题归纳 |
+| 2026-09-26 | [4df7ccbb4](https://github.com/Tencent/WeKnora/commit/4df7ccbb4f650082331452a92b52854f8922da2a) | fix(chat): drop stale minimap-target class binding | 有效变更,按主题归纳 |
+| 2026-09-27 | [9114e4e4f](https://github.com/Tencent/WeKnora/commit/9114e4e4f905be71976a77c684d3f92731e6f9a6) | fix(citations): 修复原文定位、高亮范围与 DocReader 启动失败 (#3794) | 有效变更,按主题归纳 |
+| 2026-09-28 | [a570dffcf](https://github.com/Tencent/WeKnora/commit/a570dffcfa51ad92a5c66a8260ceaaa831567dc0) | test(wiki): 表格夹具改用合成数据 | 构建/测试/文档或体验调整 |
+| 2026-09-28 | [5bf2ae1cc](https://github.com/Tencent/WeKnora/commit/5bf2ae1cc27aa555957dc1c258968c5b36246b26) | fix(faq): actually clean FAQ entry payloads before validation and storage | 有效变更,按主题归纳 |
+| 2026-09-28 | [0b463e459](https://github.com/Tencent/WeKnora/commit/0b463e459aa70ea42036a49664a75e4da95684bf) | test(skills): wait for the host install to release its lock before asserting | 构建/测试/文档或体验调整 |
+| 2026-09-28 | [9cab22685](https://github.com/Tencent/WeKnora/commit/9cab2268531477f6cd879acfd975189973d099cd) | docs: extract active tenant ID in the API quickstart | 构建/测试/文档或体验调整 |
+| 2026-09-28 | [157fb4d3c](https://github.com/Tencent/WeKnora/commit/157fb4d3c1763c47fdda88aa4133b37439bbc54a) | fix(frontend): label host-sandbox skill secret cards like the sandbox cards | 有效变更,按主题归纳 |
+| 2026-09-28 | [bdad5eea2](https://github.com/Tencent/WeKnora/commit/bdad5eea2b46546db4c1481fc83d62bd69351fcc) | fix(chat): clear stale KB scope on session restore | 有效变更,按主题归纳 |
+| 2026-09-28 | [d528d1fc3](https://github.com/Tencent/WeKnora/commit/d528d1fc3ca60f331befe1ba8c0ce28962024bd3) | fix: skip table-summary tasks for missing documents | 有效变更,按主题归纳 |
+| 2026-09-28 | [d270a844a](https://github.com/Tencent/WeKnora/commit/d270a844ac7684ed6eeea7e0469f3f5e37aef8ca) | fix: reject empty table-summary model responses | 有效变更,按主题归纳 |
+| 2026-09-28 | [675b06add](https://github.com/Tencent/WeKnora/commit/675b06addcd76a9f9f0c515bcd6554008efc67d3) | test: consolidate table summary regressions | 构建/测试/文档或体验调整 |
+| 2026-09-28 | [bad7afa6a](https://github.com/Tencent/WeKnora/commit/bad7afa6ae3198b8e574c94614e41267dbe6e005) | fix(sandbox): pass desktop slot timing to the renew goroutine | 有效变更,按主题归纳 |
+| 2026-09-28 | [4ba85e2ed](https://github.com/Tencent/WeKnora/commit/4ba85e2edf71a041fa6846c14eb74ccfa0cfb389) | fix(wiki): fan out follow-up triggers so backlogs drain at ingest_max_inflight | 有效变更,按主题归纳 |
+| 2026-09-28 | [c819abd95](https://github.com/Tencent/WeKnora/commit/c819abd9589c23a3fa504f946fc477499cd0fe70) | fix(wiki): bound outstanding follow-up tasks | 有效变更,按主题归纳 |
+| 2026-09-28 | [ec0ca4633](https://github.com/Tencent/WeKnora/commit/ec0ca46334556e5b49a981e9cf26189e162eca31) | fix(server): run ResourceCleaner on signal shutdown instead of racing Fatalf exit (#3804) | 有效变更,按主题归纳 |
+| 2026-09-29 | [08f7e702c](https://github.com/Tencent/WeKnora/commit/08f7e702cfb3a1344b537641739fa08c523df255) | fix(datasource/dingtalk): 跳过不再静默,展开叶子不再报错 | 有效变更,按主题归纳 |
+| 2026-09-29 | [7d1d71e50](https://github.com/Tencent/WeKnora/commit/7d1d71e50d7123dc14c749d5677b588f728dc7c9) | fix(datasource/dingtalk): 块级渲染不再静默丢内容(列表按线上真实形状渲染 + 未建模块类型可见) (#3822) | 有效变更,按主题归纳 |
+| 2026-09-29 | [72e105f46](https://github.com/Tencent/WeKnora/commit/72e105f46a7747d5acb1a887bd459419d4b9b24d) | fix(gitlab): 按提交快照比较增量同步变更 | 有效变更,按主题归纳 |
+| 2026-09-29 | [a282d1091](https://github.com/Tencent/WeKnora/commit/a282d1091ebf8c5d7d8d64796a25f1e1527bd296) | fix(feishu): 云盘分页补上 #3645 给 wiki 加的重复 page token 防护 (#3770) | 有效变更,按主题归纳 |
+| 2026-09-29 | [b9be7d0d0](https://github.com/Tencent/WeKnora/commit/b9be7d0d0d525435b67caeb49ef0502d6b6e7097) | fix(notion): 修复页面清空后旧正文残留 | 有效变更,按主题归纳 |
+| 2026-09-29 | [15d484552](https://github.com/Tencent/WeKnora/commit/15d484552140eafe491e01d7f7bbfb558515801b) | fix(rss): 修复全文回退后增量同步不重试 | 有效变更,按主题归纳 |
+| 2026-09-29 | [cf079b8dc](https://github.com/Tencent/WeKnora/commit/cf079b8dcd3315b46f1a2c848837f851ec61e04c) | fix(models): keep Retry-After on HTTPError and honour it in agent LLM retries | 有效变更,按主题归纳 |
+| 2026-09-29 | [436d02fdf](https://github.com/Tencent/WeKnora/commit/436d02fdf3865279df3ff3bc14a91057b56d0464) | fix(compaction): 流因取消而关闭时不得当成空摘要成功返回 | 有效变更,按主题归纳 |
+| 2026-09-29 | [e39043f34](https://github.com/Tencent/WeKnora/commit/e39043f34b22bec26ed6ddb4d08ff66e9555f3ea) | fix(repository): 关键词 LIKE/ILIKE 检索补显式 ESCAPE,修复 Lite 上静默漏结果 | 有效变更,按主题归纳 |
+| 2026-09-29 | [1912ba26c](https://github.com/Tencent/WeKnora/commit/1912ba26c544025adb869459a74c2cbac2a95cd9) | test(weaviate): 让并发建表的用例真的每个 worker 都去创建,消掉随机失败 | 构建/测试/文档或体验调整 |
+| 2026-09-29 | [f46c99056](https://github.com/Tencent/WeKnora/commit/f46c9905677b36a454aa6669a2619f9eecf9fd23) | fix(embed): 开启 Host context 后点击追问建议不再返回 400 | 有效变更,按主题归纳 |
+| 2026-09-29 | [75d2c1b58](https://github.com/Tencent/WeKnora/commit/75d2c1b58343b2872051a24b78185298fb8b88f3) | fix(frontend): normalize style guard paths | 有效变更,按主题归纳 |
+| 2026-09-29 | [0c6b30cfa](https://github.com/Tencent/WeKnora/commit/0c6b30cfa705c1d2f2ccd084cd76174738a7fb4b) | fix(docs): spawn nested npm through Node on Windows | 有效变更,按主题归纳 |
+| 2026-09-29 | [9842e51d4](https://github.com/Tencent/WeKnora/commit/9842e51d40a54ca2dbb61153f3fb587da1265b7a) | fix(frontend): 修复复制模型时模型名称被改写的问题 (#3830) | 有效变更,按主题归纳 |
+| 2026-09-30 | [5225edc0f](https://github.com/Tencent/WeKnora/commit/5225edc0f746ae68e15dc4bd6fde9343efde1142) | feat(auth): 新增仅限邀请注册档位 (#3900) | 有效变更,按主题归纳 |
+| 2026-09-30 | [176aa0d02](https://github.com/Tencent/WeKnora/commit/176aa0d02b30b64634d55d000b6b44e00ec0310a) | fix(sandbox): build real arm64 sandbox images and ship arm64 Cube variants | 有效变更,按主题归纳 |
+| 2026-09-30 | [dba912755](https://github.com/Tencent/WeKnora/commit/dba9127558c3df160c8ff181e2a0a99172ae7c75) | docs(api): 补充模型复制接口说明 | 构建/测试/文档或体验调整 |
+| 2026-09-30 | [2ec177462](https://github.com/Tencent/WeKnora/commit/2ec1774623a77608902287b2ba42dcde3e4daaab) | fix(auth): 令牌写入失败不再当成登录成功,查询失败不再谎称已撤销 | 有效变更,按主题归纳 |
+| 2026-09-30 | [b71b31c89](https://github.com/Tencent/WeKnora/commit/b71b31c897efae7b6b8320ab0e5bdf796eb1e006) | fix(ima): 笔记正文被清空后旧正文仍留在知识库 | 有效变更,按主题归纳 |
+| 2026-09-30 | [1dbe9d0c1](https://github.com/Tencent/WeKnora/commit/1dbe9d0c1a1b9dd8d7c345c5c1672f76269b86c5) | fix(gitlab): 仓库文件下载补体积上限,超限报错而不是无界读入 (#3887) | 有效变更,按主题归纳 |
+| 2026-09-30 | [75fc0ad15](https://github.com/Tencent/WeKnora/commit/75fc0ad1590786202af57546a6879bb074671ab9) | feat(docparser): upgrade AnyDoc to 0.2.4 and parse assets once | 有效变更,按主题归纳 |
+| 2026-09-30 | [bccb4b151](https://github.com/Tencent/WeKnora/commit/bccb4b151bae403508da77fbb174efc79dc47c1a) | fix(embed): 内嵌网页只传图片不输入文字时不再返回 400 | 有效变更,按主题归纳 |
+| 2026-10-07 | [402f69832](https://github.com/Tencent/WeKnora/commit/402f69832e395d8d193d89fef7a248a1c48ad24c) | fix(notion): 查询撞厂商上限被截断时不再静默,截断轮次不下发删除 (#3845) | 有效变更,按主题归纳 |
+| 2026-10-07 | [0ec945fe5](https://github.com/Tencent/WeKnora/commit/0ec945fe5df0def761351aa640bdb588bd8e1bb2) | fix(dataset): use supported default QA model (#3965) | 有效变更,按主题归纳 |
+| 2026-10-07 | [72b8a106c](https://github.com/Tencent/WeKnora/commit/72b8a106c77a3337bd5d785c595c8575686b6025) | fix(sqlite): 检索后端批量更新索引副本失败不再吞掉错误 (#3768) | 有效变更,按主题归纳 |
+| 2026-10-07 | [a8a2c17b8](https://github.com/Tencent/WeKnora/commit/a8a2c17b892c378686e4d0f3916b064d2b38da80) | fix(sqlite): 索引行写入失败不再被丢弃,BatchSave 回滚而不是留下「主行在、索引行不在」 (#3843) | 有效变更,按主题归纳 |
+| 2026-10-07 | [f9eeb8dc8](https://github.com/Tencent/WeKnora/commit/f9eeb8dc8535d678e206b00c6e24540db542b868) | fix(retriever/es): _bulk 的 per-item 失败不再被吞,全部被拒时 BatchSave 返回错误 (#3842) | 有效变更,按主题归纳 |
+| 2026-10-07 | [5f5093b9f](https://github.com/Tencent/WeKnora/commit/5f5093b9f74e2c8621c11e1c53886758784d68b2) | fix(retriever): Qdrant/Milvus 关键词检索整批失败返回 error,不再当成「没有匹配」 (#3847) | 有效变更,按主题归纳 |
+| 2026-10-07 | [5648d24a9](https://github.com/Tencent/WeKnora/commit/5648d24a93d353dca6d719178752e26d72a2478a) | fix(retriever/milvus): CopyIndices 按映射分块读源行,去掉会静默漏向量的 offset 窗口 (#3883) | 有效变更,按主题归纳 |
+| 2026-10-07 | [9490035f1](https://github.com/Tencent/WeKnora/commit/9490035f1517bd68545df024e05b2244e4d1ac9a) | fix(agent): 图谱查询的三处截断必须可见,模型不再把子图当全集 (#3885) | 有效变更,按主题归纳 |
+| 2026-10-07 | [9fa9e7c29](https://github.com/Tencent/WeKnora/commit/9fa9e7c29d8eb727825b14f0b0ca23907969b91a) | fix: use consistent qdrant env settings (#3960) | 有效变更,按主题归纳 |
+| 2026-10-08 | [afc6cbd50](https://github.com/Tencent/WeKnora/commit/afc6cbd50264a54018faa9552346d45f964834fe) | fix(milvus): validate connections and preserve gRPC resolver defaults (#3654) | 有效变更,按主题归纳 |
+| 2026-10-08 | [e29301115](https://github.com/Tencent/WeKnora/commit/e29301115a276afc4e7be7f0ffe568066c0c3386) | fix(graph): preserve relationship direction and document provenance (#3950) | 有效变更,按主题归纳 |
+| 2026-10-08 | [7c284e7fc](https://github.com/Tencent/WeKnora/commit/7c284e7fc9beb065c4da2587b82bed52b5073bef) | fix(faq): 保留克隆状态同步时的正文和索引状态 (#3993) | 有效变更,按主题归纳 |
+| 2026-10-08 | [a8c291c6f](https://github.com/Tencent/WeKnora/commit/a8c291c6ff52985e6729a8b3e0d404add08ecfd3) | fix: keep failed data-analysis table loads retryable (#3745) | 有效变更,按主题归纳 |
+| 2026-10-08 | [27b9ee74a](https://github.com/Tencent/WeKnora/commit/27b9ee74a02f2f33689d0adb2de091c5bd8d5342) | fix(storage): use virtual-hosted addressing for Tencent COS through the S3 driver (#3722) | 有效变更,按主题归纳 |
+| 2026-10-08 | [a15dfdd94](https://github.com/Tencent/WeKnora/commit/a15dfdd941497b32198f8abaec71510b304ee070) | fix(summary): 坏 JSON 不再冒充文档摘要(截断/畸形 JSON 走失败路径,纯文本兜底保留) (#3769) | 有效变更,按主题归纳 |
+| 2026-10-08 | [e915adb14](https://github.com/Tencent/WeKnora/commit/e915adb147d721fc265b22db24a56a535b5bcf6c) | fix(chunker): preserve source offsets for table child chunks (#3969) | 有效变更,按主题归纳 |
+| 2026-10-08 | [98feef38d](https://github.com/Tencent/WeKnora/commit/98feef38de5c39e57ef1c817e43ef6a4879f4029) | fix(chunker): count an escaped pipe as cell text when matching table widths (#3979) | 有效变更,按主题归纳 |
+| 2026-10-08 | [0d3a053eb](https://github.com/Tencent/WeKnora/commit/0d3a053eb62c5150c8ad01442776de0225d06ea2) | fix(yuque): don't acknowledge failed detail fetches in the incremental cursor (#3702) | 有效变更,按主题归纳 |
+| 2026-10-08 | [b5c74b2dd](https://github.com/Tencent/WeKnora/commit/b5c74b2dd6f183f1e31ef98a166696664b716570) | fix(feishu): 多维表格字段分页补重复 page_token 防护与条数上限 (#3840) | 有效变更,按主题归纳 |
+| 2026-10-08 | [a2aeb97a0](https://github.com/Tencent/WeKnora/commit/a2aeb97a01203b23b6093eba2369583b460aaa66) | fix(gitlab): 分页循环补重复页防护与跳数上限 (#3841) | 有效变更,按主题归纳 |
+| 2026-10-08 | [a6c6c35ad](https://github.com/Tencent/WeKnora/commit/a6c6c35ad292716570c44a42221f9e5ec68d2543) | fix(ima): 三条游标分页循环补重复游标防护与跳数上限 (#3889) | 有效变更,按主题归纳 |
+| 2026-10-08 | [c9723b119](https://github.com/Tencent/WeKnora/commit/c9723b1198bfa42d31278b4e44f1199593ee5a08) | fix(docreader): read EPUB chapters in spine order (#3977) | 有效变更,按主题归纳 |
+| 2026-10-08 | [9de0946e5](https://github.com/Tencent/WeKnora/commit/9de0946e53348d063180b2af87008a51c4fd9dcc) | fix(docparser): anydoc 内嵌 EMF/WMF 图片的扩展名与 Rust 侧对齐 (#3938) | 有效变更,按主题归纳 |
+| 2026-10-08 | [188d9026d](https://github.com/Tencent/WeKnora/commit/188d9026d8d344464676c7919c09730a5181adb6) | fix(models): honor OpenRouter native thinking for legacy defaults (#3807) | 有效变更,按主题归纳 |
+| 2026-10-08 | [f6119ae5f](https://github.com/Tencent/WeKnora/commit/f6119ae5fd4032e00a621a2cc280972390d0cfa4) | fix(agent): 保留数据分析 SQL 字符串字面量 (#3812) | 有效变更,按主题归纳 |
+| 2026-10-08 | [52893006a](https://github.com/Tencent/WeKnora/commit/52893006ad26558164ddceabce89360a1b17b984) | fix(agent): validate tool string lengths by Unicode code points (#3937) | 有效变更,按主题归纳 |
+| 2026-10-08 | [fa4f85d3d](https://github.com/Tencent/WeKnora/commit/fa4f85d3dd1edba7689a3e235abd3f159b2af381) | fix(agent): 拒绝将断流摘要保存为记忆检查点 (#3994) | 有效变更,按主题归纳 |
+| 2026-10-08 | [e394c7ade](https://github.com/Tencent/WeKnora/commit/e394c7adecbe38019f66b8490ee06d254efaa05a) | fix(agent): 工具自起 goroutine 的 panic 自行兜底,MCP 预热不再带走进程 (#3846) | 有效变更,按主题归纳 |
+| 2026-10-08 | [412ce63f0](https://github.com/Tencent/WeKnora/commit/412ce63f0b00ab45811f81b0ca254259e205acca) | fix(chat): clear stale file and tool selections on session restore (#3802) | 有效变更,按主题归纳 |
+| 2026-10-08 | [f55947349](https://github.com/Tencent/WeKnora/commit/f559473498719ff0f5c94fc9c6918199986aaf3e) | fix(mcp): keep the client secret from dynamic client registration (#3763) | 有效变更,按主题归纳 |
+| 2026-10-08 | [28734725a](https://github.com/Tencent/WeKnora/commit/28734725a3567d5d12a0b190b0e5bccbb91b252c) | fix(mcp): reconnect after the server forgets the session (#3765) | 有效变更,按主题归纳 |
+| 2026-10-08 | [172d0a85a](https://github.com/Tencent/WeKnora/commit/172d0a85ab17e74abe8f987927ad1b9fd3eeb29a) | fix(mcp): pass embedded resources in tool results to the agent (#3766) | 有效变更,按主题归纳 |
+| 2026-10-08 | [e8cc9c763](https://github.com/Tencent/WeKnora/commit/e8cc9c763e639749bab486a0fc291d91c6b09597) | fix(mcp): include shared knowledge bases in endpoint scope (#3848) | 有效变更,按主题归纳 |
+| 2026-10-08 | [69a7fd993](https://github.com/Tencent/WeKnora/commit/69a7fd9931a0867dd7e1003262b315a330e5d606) | fix(mcp): close legacy chat responses on HTTP errors (#3930) | 有效变更,按主题归纳 |
+| 2026-10-08 | [c0c194f62](https://github.com/Tencent/WeKnora/commit/c0c194f6253ba38ddfc868767bda21fd6f4ea900) | fix(im/wecom): preserve image in mixed messages with text captions (#3901) | 有效变更,按主题归纳 |
+| 2026-10-08 | [95691daab](https://github.com/Tencent/WeKnora/commit/95691daab2e292a8c45e0ae200737b1c9e5de222) | fix(knowledge): retry document processing after transient indexing failures (#3915) | 有效变更,按主题归纳 |
+| 2026-10-08 | [fac9ebabf](https://github.com/Tencent/WeKnora/commit/fac9ebabf2d4578f6192528b1e68f69af48d9a96) | fix(knowledge): close spans left open on terminal rows and superseded attempts (#3855) | 有效变更,按主题归纳 |
+| 2026-10-08 | [4d7e66312](https://github.com/Tencent/WeKnora/commit/4d7e663128e70f4f4ba496f062fc5a7c23633358) | fix(wiki): preserve pending work when source lookups fail (#3715) | 有效变更,按主题归纳 |
+| 2026-10-08 | [626185a12](https://github.com/Tencent/WeKnora/commit/626185a120ef55a741c5add46147c0af146c18aa) | fix(wiki): 重建链接逐页写失败不再静默,半成品不再回 200 (#3884) | 有效变更,按主题归纳 |
+| 2026-10-08 | [11bfacad9](https://github.com/Tencent/WeKnora/commit/11bfacad97c7748542a988717bd3f4ffc53244b4) | chore(deps): bump the cli-deps group in /cli with 2 updates (#3920) | 构建/测试/文档或体验调整 |
+| 2026-10-08 | [3206e15e8](https://github.com/Tencent/WeKnora/commit/3206e15e831672f0ae0d837d822150ca19415594) | fix(docs): make `make docs` reproduce the committed Swagger again (#3751) | 有效变更,按主题归纳 |
+| 2026-10-08 | [9d4cc233c](https://github.com/Tencent/WeKnora/commit/9d4cc233c68a3f3e7ee7fd81926708dd5c898b36) | feat(ui): add search to Skills and MCP toolbox lists (#3731) | 有效变更,按主题归纳 |
+| 2026-10-08 | [4ea94afcb](https://github.com/Tencent/WeKnora/commit/4ea94afcb91b7f2388d9440495f30cf555e88a5b) | fix(cli): propagate paginated response read errors (#3983) | 有效变更,按主题归纳 |
+| 2026-10-08 | [4a0c7fa91](https://github.com/Tencent/WeKnora/commit/4a0c7fa91f2f646e6f56181ba723a9eebf368ed9) | fix(cli): render paginated API results in text mode (#3984) | 有效变更,按主题归纳 |
+| 2026-10-08 | [7d75ce3a6](https://github.com/Tencent/WeKnora/commit/7d75ce3a6f871ae59668b440439ea27a9959c351) | fix(client): make multipart uploads replayable (#3989) | 有效变更,按主题归纳 |
+| 2026-10-08 | [cefd5920f](https://github.com/Tencent/WeKnora/commit/cefd5920f885b604cc760478e9dd87b5778a272e) | fix(cli): apply list limits after filtering and sorting (#3990) | 有效变更,按主题归纳 |
+| 2026-10-08 | [ff3318443](https://github.com/Tencent/WeKnora/commit/ff33184433c438eea0ec40293289b819fa817c97) | fix(cli): preserve integer precision in jq output (#3991) | 有效变更,按主题归纳 |
+| 2026-10-08 | [b413023d2](https://github.com/Tencent/WeKnora/commit/b413023d2885742048103d4d617922465788c0f2) | fix(cli): classify wrapped SDK HTTP errors correctly (#3992) | 有效变更,按主题归纳 |
+| 2026-10-08 | [562127c65](https://github.com/Tencent/WeKnora/commit/562127c6542d9f3f073f390658cb67ae65b97484) | fix(client): accept optional spaces in SSE event fields (#3975) | 有效变更,按主题归纳 |
+| 2026-10-08 | [b8d83eab1](https://github.com/Tencent/WeKnora/commit/b8d83eab1cf426d19c8e0183b9d896b77a62872b) | fix(skills): skip vendored dep trees in install verification script enumeration (#3943) | 有效变更,按主题归纳 |
+| 2026-10-08 | [dad81cb8c](https://github.com/Tencent/WeKnora/commit/dad81cb8c10c52efd5d70dfe1abcc9f0c5672c99) | fix(sandbox): reclassify exec races with the idle sweep as NotFound so rebinding heals (#3980) | 有效变更,按主题归纳 |
+| 2026-10-08 | [20038cbda](https://github.com/Tencent/WeKnora/commit/20038cbda6e3c810b4ab3951d56bbcf8f81af19c) | fix(asynqdl): truncate dead-letter errors on a rune boundary (#3964) | 有效变更,按主题归纳 |
+| 2026-10-08 | [6057a8101](https://github.com/Tencent/WeKnora/commit/6057a81017375a8d680cd32afca1c2ff03b0b17c) | fix(middleware): stop buffering whole request bodies in the access log (#3927) | 有效变更,按主题归纳 |
+| 2026-10-08 | [5077f3580](https://github.com/Tencent/WeKnora/commit/5077f3580003bcf55eeb89694d43479ca216adae) | fix(notion): 超过 block 上限截断时补 Warn,避免内容与子页面静默缺失 (#3771) | 有效变更,按主题归纳 |
+| 2026-10-08 | [dabbdaf15](https://github.com/Tencent/WeKnora/commit/dabbdaf156b8dfcb58dd68e9c0be3842d27d72b5) | fix(notion): 分页循环补重复游标防护与跳数上限 (#3890) | 有效变更,按主题归纳 |
+| 2026-10-08 | [e7aaf8ae0](https://github.com/Tencent/WeKnora/commit/e7aaf8ae02608d1d9e0f84aa864106a5f1ae78a3) | fix(docreader): keep the space held by a whitespace-only inline element (#3978) | 有效变更,按主题归纳 |
+| 2026-10-08 | [336f84c3b](https://github.com/Tencent/WeKnora/commit/336f84c3be4a4831c49b2708f98431dae2ab3312) | fix(data-analysis): skip style-only worksheets before DuckDB loading (#3660) | 有效变更,按主题归纳 |
+| 2026-10-08 | [4dd19b535](https://github.com/Tencent/WeKnora/commit/4dd19b535603d4a9dc8b64033192592572db5180) | fix(frontend): render single tilde literally instead of GFM strikethrough (#3963) | 有效变更,按主题归纳 |
+| 2026-10-08 | [4eb23a856](https://github.com/Tencent/WeKnora/commit/4eb23a85607c0cabfa050f5e87eb71fb0ce2554c) | fix(qdrant): CopyIndices 用服务端返回的 next_page_offset 翻页,不再按页尾 id 回填 (#3882) | 有效变更,按主题归纳 |
+| 2026-10-08 | [e6ecb9adc](https://github.com/Tencent/WeKnora/commit/e6ecb9adcc8ddbb500957f77490ca5cb4c2fa8d4) | fix(dataset): allow selecting the QA answer model (#3944) | 有效变更,按主题归纳 |
+| 2026-10-08 | [49e262147](https://github.com/Tencent/WeKnora/commit/49e262147ca504a28680930930ed0851a622d0c5) | fix(cli): preserve directory structure in doc upload --recursive (#3753) | 有效变更,按主题归纳 |
+| 2026-10-08 | [de2dfd303](https://github.com/Tencent/WeKnora/commit/de2dfd3034a7b506fa269910401bbdcbd4efc71d) | fix(multimodal): bound remote image downloads by the file size limit (#3906) | 有效变更,按主题归纳 |
+| 2026-10-08 | [cbfd66fd9](https://github.com/Tencent/WeKnora/commit/cbfd66fd922df3632448e0bfd5925a6c10196fc8) | fix(knowledge): delete written chunks when the storage quota check fails (#3656) | 有效变更,按主题归纳 |
+| 2026-10-08 | [11aa00f85](https://github.com/Tencent/WeKnora/commit/11aa00f85f22ee54a6c37963010ac3f4d20d694b) | fix(agent): preserve untouched numeric tool arguments during casting (#3936) | 有效变更,按主题归纳 |
+| 2026-10-08 | [9a2cd2006](https://github.com/Tencent/WeKnora/commit/9a2cd200637359081d87fb4c51b6bc911a0ff6a2) | fix(frontend): keep target="_blank" on sanitized markdown links (#3818) | 有效变更,按主题归纳 |
+| 2026-10-08 | [a75e8e9ac](https://github.com/Tencent/WeKnora/commit/a75e8e9ac0ac10ea23c3108e789a8d71f6f071e9) | fix(datasource): 语雀/Notion/IMA/飞书连接器 API 响应体补体积上限 (#3985) | 有效变更,按主题归纳 |
+| 2026-10-08 | [b5fd79edf](https://github.com/Tencent/WeKnora/commit/b5fd79edf283f4ac6a272403f231c3c07db11500) | fix(model): treat 400 as reachable in chat connection test again (#3929) | 有效变更,按主题归纳 |
+| 2026-10-08 | [1826f46b1](https://github.com/Tencent/WeKnora/commit/1826f46b1217867fe15f188706a9334cacec4165) | fix(docparser): 云解析超时改为跟随既有配置,不再硬编码 (#3891) | 有效变更,按主题归纳 |
+| 2026-10-08 | [99b4c3a9b](https://github.com/Tencent/WeKnora/commit/99b4c3a9b0d3c2d3440036c6e3f4235c6f276390) | feat(rerank): support Hugging Face TEI protocol (#3956) | 有效变更,按主题归纳 |
+| 2026-10-08 | [d84cdabac](https://github.com/Tencent/WeKnora/commit/d84cdabac49046dc546ba04716afcc3785240a88) | fix(rss): refresh titles and source links when article text is unchanged (#3824) | 有效变更,按主题归纳 |
+| 2026-10-08 | [bea0d7b68](https://github.com/Tencent/WeKnora/commit/bea0d7b6891ccbf94d3e1c6f649a3a6092978a12) | fix(docreader): preserve PPTX fallback image associations (#3973) | 有效变更,按主题归纳 |
+| 2026-10-08 | [3fb6b9d63](https://github.com/Tencent/WeKnora/commit/3fb6b9d631d399104778cf8d4a79facd23ed8853) | fix(kb): let the update API set vlm_config (#3704) | 有效变更,按主题归纳 |
+| 2026-10-08 | [b5c8d90cc](https://github.com/Tencent/WeKnora/commit/b5c8d90cc20747be3f64356150ca69224d063be0) | fix(chat_pipeline): treat non-JSON model refusals as skipped graph extraction (#3606) | 有效变更,按主题归纳 |
+| 2026-10-08 | [ae9054626](https://github.com/Tencent/WeKnora/commit/ae90546262a84dcfd90b5da370b56bd0774cc8f2) | fix(frontend): avoid false unsaved changes warnings | 有效变更,按主题归纳 |
+| 2026-10-08 | [bcbd65a45](https://github.com/Tencent/WeKnora/commit/bcbd65a459e912f70ad4a9b0f39f8848db4a2d9f) | fix(datasource/dingtalk): Validate 不再因为某个工作区读不到就整库判失败 (#3782) | 有效变更,按主题归纳 |
+| 2026-10-08 | [19542f2a1](https://github.com/Tencent/WeKnora/commit/19542f2a1d958d26383bbb586adecae6be07a7d5) | feat(datasource): add Seafile data source connector (#3616) | 有效变更,按主题归纳 |
+| 2026-10-08 | [d2bc0df6a](https://github.com/Tencent/WeKnora/commit/d2bc0df6a2af4f953485e69c63972cf7d1cc2646) | feat(i18n): add Traditional Chinese (zh-TW) locale support (#3916) | 有效变更,按主题归纳 |
+| 2026-10-08 | [1c528d83d](https://github.com/Tencent/WeKnora/commit/1c528d83de39d79e870fe8b46d21addf20136a60) | fix(i18n): add the zh-TW Seafile strings missing after #3616 and #3916 | 有效变更,按主题归纳 |
+| 2026-10-08 | [0de62b2f8](https://github.com/Tencent/WeKnora/commit/0de62b2f801513982291311b59fad381f7959560) | fix(chat): restore citation source navigation and soften highlights | 有效变更,按主题归纳 |
+| 2026-10-08 | [a729d1fc8](https://github.com/Tencent/WeKnora/commit/a729d1fc8c1c7e05119b69696ef9f61bc2627cfb) | docs(models): state which rerank limits split requests and which fail the call | 构建/测试/文档或体验调整 |
+| 2026-10-08 | [ea7cf8d5d](https://github.com/Tencent/WeKnora/commit/ea7cf8d5d77cf0deb46215836262405ad91d4b44) | fix(models): emit explicit empty content for assistant tool-call turns (#3605) | 有效变更,按主题归纳 |
+| 2026-10-08 | [5a25e2614](https://github.com/Tencent/WeKnora/commit/5a25e261421db4c76b720e9d7392bd8f88f36585) | feat(rerank): support Pinecone Inference protocol (#3957) | 有效变更,按主题归纳 |
+| 2026-10-08 | [b7f318106](https://github.com/Tencent/WeKnora/commit/b7f31810664e22e44cb6da2dae3b2115c2060f49) | docs: count 29 built-in model vendors and list Hugging Face TEI and Pinecone (#4048) | 构建/测试/文档或体验调整 |
+| 2026-10-08 | [0c5d28abb](https://github.com/Tencent/WeKnora/commit/0c5d28abb7917b49d36f79bafb78fb59796a068b) | feat: index and recall images with multimodal embedding models (#3683) | 有效变更,按主题归纳 |
+| 2026-10-08 | [676a59307](https://github.com/Tencent/WeKnora/commit/676a59307e5ca7efabfe8896354b9425703aef67) | fix(settings): improve modal sizing and integration page layout | 有效变更,按主题归纳 |
+| 2026-10-08 | [2a02807a6](https://github.com/Tencent/WeKnora/commit/2a02807a618a18196ebc1ddbe9888fb742b24117) | fix(chunker): respect Markdown fences during heading detection | 有效变更,按主题归纳 |
+| 2026-10-08 | [005627b18](https://github.com/Tencent/WeKnora/commit/005627b180c6c7509660b0a14401e463f48d53f8) | fix(agent): compare tool enum values without string coercion | 有效变更,按主题归纳 |
+| 2026-10-09 | [b14bcbcd5](https://github.com/Tencent/WeKnora/commit/b14bcbcd541fc0fdd67eef518828a28302358e18) | fix(notion): 保留正文抓取失败时的同步游标 (#3693) | 有效变更,按主题归纳 |
+| 2026-10-09 | [04da26024](https://github.com/Tencent/WeKnora/commit/04da26024c2412b34b3b30e7b283c98197ef95f1) | fix(weaviate): CopyIndices 补游标防护,并修掉 GraphQL 报错时的 panic (#3881) | 有效变更,按主题归纳 |
+| 2026-10-09 | [5d40f81ee](https://github.com/Tencent/WeKnora/commit/5d40f81ee4020d8f3173ac64df4dc4ae140351c1) | fix(im): retry Yunzhijia attachment downloads (#3809) | 有效变更,按主题归纳 |
+| 2026-10-09 | [419f17fc6](https://github.com/Tencent/WeKnora/commit/419f17fc651de93a34eb3586cb894a9e7f3db4a4) | fix(wiki): resolve titles for unloaded backlinks (#3931) | 有效变更,按主题归纳 |
+| 2026-10-09 | [cbfc0fed4](https://github.com/Tencent/WeKnora/commit/cbfc0fed4a1cb4a3ff0ffabe943d6f1edfa3b21e) | fix(chat): 在模型上下文中标注 FILTER_TOP_K 的子集截断,避免把前 N 条当作全集作答 (#3758) | 有效变更,按主题归纳 |
+| 2026-10-09 | [abc365f91](https://github.com/Tencent/WeKnora/commit/abc365f919191317020d395e9af19e4129357ff3) | fix(wiki): 唯一写出口 UpdatePage 拦住丢表格行的机器重写(agent 整页写入不再静默改短页面) (#3793) | 有效变更,按主题归纳 |
+| 2026-10-09 | [8319d59dd](https://github.com/Tencent/WeKnora/commit/8319d59dd014fccb7daf04eed9d50d8a3ffdecf4) | feat(datasource/dingtalk): 摄取钉钉"上传文件"(docx/pptx/xlsx/pdf) (#3786) | 有效变更,按主题归纳 |
+| 2026-10-09 | [09411106f](https://github.com/Tencent/WeKnora/commit/09411106f0ec19e840e9c24cabf391b8c2852a8e) | fix(session): allow tenant admins to manage channel sessions (#3939) | 有效变更,按主题归纳 |
+| 2026-10-09 | [3d01bebce](https://github.com/Tencent/WeKnora/commit/3d01bebce3d7aa90aaa52914534f22d9cadf683b) | fix(docreader): resolve EPUB images relative to their chapters (#3924) | 有效变更,按主题归纳 |
+| 2026-10-09 | [4243018ff](https://github.com/Tencent/WeKnora/commit/4243018ff1f55ccd92baebf4da0a1057317cc644) | fix(mcp): report advertised directory failures in connection tests (#3905) | 有效变更,按主题归纳 |
+| 2026-10-09 | [a1450d382](https://github.com/Tencent/WeKnora/commit/a1450d382e39a06977cc395aaf822eebab87c69e) | fix(knowledge): claim cloned and released document images in the resource catalog (#3752) | 有效变更,按主题归纳 |
+| 2026-10-09 | [65f88f412](https://github.com/Tencent/WeKnora/commit/65f88f4121433702ea053ae9127647329bf98b05) | feat: judge image hits by their image in rerank and show them to vision models (#4013) | 有效变更,按主题归纳 |
+| 2026-10-09 | [52617f641](https://github.com/Tencent/WeKnora/commit/52617f64106e2ea9af770ff783a1818e1ce7b8fb) | fix(sandbox): omit env overrides when creating Aliyun E2B sandboxes from snapshots (#4054) | 有效变更,按主题归纳 |
+| 2026-10-09 | [602d0c388](https://github.com/Tencent/WeKnora/commit/602d0c388bf17d894f467fe0cd65d1d62c73d303) | feat: support app and docreader build download mirrors (#4073) | 有效变更,按主题归纳 |
+| 2026-10-10 | [d074ec7b0](https://github.com/Tencent/WeKnora/commit/d074ec7b0b63903f117543e09deb31f5f3d19ecb) | fix(websearch): return results for non-webpage Metaso scopes (#4010) | 有效变更,按主题归纳 |
+| 2026-10-10 | [40eb34bc8](https://github.com/Tencent/WeKnora/commit/40eb34bc849cb29dc53ccddef971c9a7c4410f7b) | fix(agent): align web_search count limit with the configured maximum (#4011) | 有效变更,按主题归纳 |
+| 2026-10-10 | [c3f077ed1](https://github.com/Tencent/WeKnora/commit/c3f077ed18f6e00aaf705d405023ed88c7c15b27) | fix(models): align configuration and multimodal embedding workflows (#4080) | 有效变更,按主题归纳 |
+| 2026-10-10 | [0c71fff2a](https://github.com/Tencent/WeKnora/commit/0c71fff2a387b539c449cd6b66a3922ed6839f76) | fix(sandbox): treat E2B in_progress template builds as building (#4108) | 有效变更,按主题归纳 |
+| 2026-10-10 | [6e2e5d8bb](https://github.com/Tencent/WeKnora/commit/6e2e5d8bb07cd9c8db360c03b80ca47ea2e81fd5) | fix(frontend): follow configured upload limit in hints (#4110) | 有效变更,按主题归纳 |
+| 2026-10-10 | [097bd475f](https://github.com/Tencent/WeKnora/commit/097bd475f601d48a82ed7aaa2f9676dd82f15d97) | fix(docker): install CA certificates before the browserskill mirror swap (#4104) | 有效变更,按主题归纳 |
+| 2026-10-10 | [8125ed250](https://github.com/Tencent/WeKnora/commit/8125ed250f5ea404445e7952359a6ac686d90afc) | fix(chat): keep custom metadata on knowledge-chat references (#4087) | 有效变更,按主题归纳 |
+| 2026-10-10 | [d6269994b](https://github.com/Tencent/WeKnora/commit/d6269994b33688c2b85488c24014d95db1948d52) | refactor: 出站拨号超时默认 5s 并改为 env 可配 (WEKNORA_OUTBOUND_DIAL_TIMEOUT) (#4122) | 有效变更,按主题归纳 |
+| 2026-10-10 | [03e2c47ae](https://github.com/Tencent/WeKnora/commit/03e2c47aecbf60120820216baaa8abb2e98844f7) | feat(mcp): support 2026-07-28 alongside legacy clients (#4095) | 有效变更,按主题归纳 |
+| 2026-10-10 | [e372b8e42](https://github.com/Tencent/WeKnora/commit/e372b8e42ef71f8ba9e1c4ada6546d79ec056f91) | perf(embedding): batch Qwen3-VL message inputs (#4126) | 有效变更,按主题归纳 |
+| 2026-10-10 | [8cb53fdbc](https://github.com/Tencent/WeKnora/commit/8cb53fdbc58e4c15aafdba973d65742551e4c9cd) | fix(observability): distinguish retrieval diagnostic spans (#4128) | 有效变更,按主题归纳 |
+| 2026-10-10 | [bb878af50](https://github.com/Tencent/WeKnora/commit/bb878af50873b919a68134dc007ba9ae859c6c39) | fix(gallery): improve image detail panel readability (#4130) | 有效变更,按主题归纳 |
+| 2026-10-10 | [d27ecade6](https://github.com/Tencent/WeKnora/commit/d27ecade656a1f2335baa2520d2622a297036ea5) | fix(ocr): reject invalid output and record extraction failures (#4132) | 有效变更,按主题归纳 |
+| 2026-10-10 | [5f18308b0](https://github.com/Tencent/WeKnora/commit/5f18308b0a39bd44f3b44203c167663d0430a40b) | fix(ocr): reject split repetition and preserve no-text responses (#4135) | 有效变更,按主题归纳 |
+| 2026-10-10 | [a58fc1320](https://github.com/Tencent/WeKnora/commit/a58fc13206c7d1e89c784d61189ab67295727744) | fix(chat): smooth streamed citation reveal cadence | 有效变更,按主题归纳 |
+| 2026-10-10 | [aa535c22f](https://github.com/Tencent/WeKnora/commit/aa535c22fed1f4ba5c965b7f208361c0b1c2d4e2) | fix(chat): keep queued follow-up answers attached to their own question | 有效变更,按主题归纳 |
+| 2026-10-10 | [50a296316](https://github.com/Tencent/WeKnora/commit/50a29631608b9b95c88b4ea98e09867bdbcad991) | fix(preview): refine source citation highlights (#4139) | 有效变更,按主题归纳 |
+| 2026-10-10 | [d8fa5b49a](https://github.com/Tencent/WeKnora/commit/d8fa5b49a7f4aa93ded9978c40d1b57346375211) | feat(embed): support pasting screenshots into chat | 有效变更,按主题归纳 |
+| 2026-10-10 | [98e9dc078](https://github.com/Tencent/WeKnora/commit/98e9dc078625265ba1ea85ee3538a547f85a5be0) | perf(reranking): accelerate MMR selection with sparse postings | 有效变更,按主题归纳 |
+| 2026-10-10 | [54e8d7ffd](https://github.com/Tencent/WeKnora/commit/54e8d7ffde31ccc00ea59cec7b2eaec18b01f137) | fix(mcp): 隔离慢连接清理对其他服务的阻塞 | 有效变更,按主题归纳 |
+| 2026-10-10 | [dccb1d6d2](https://github.com/Tencent/WeKnora/commit/dccb1d6d2c974062976e8b8841c46bbddb06d226) | fix(mcp): stop advertising read-only endpoint tools as destructive | 有效变更,按主题归纳 |
+| 2026-10-10 | [de616ed4b](https://github.com/Tencent/WeKnora/commit/de616ed4b7d2a6e55c9d380bd6068f792fe9f3a1) | test(service): assert RRF ranks ignore list arrival order | 构建/测试/文档或体验调整 |
+| 2026-10-10 | [1a21af1b8](https://github.com/Tencent/WeKnora/commit/1a21af1b827072589eddcc9def0b91433399480c) | fix(mcp): preserve structured tool results for agents | 有效变更,按主题归纳 |
+| 2026-10-10 | [d9237166c](https://github.com/Tencent/WeKnora/commit/d9237166cc4cb618818d2e65e62f820f1cc03f24) | fix(im/wecom): accept custom wss long-connection endpoints behind SSRF host checks | 有效变更,按主题归纳 |
+| 2026-10-10 | [def582b98](https://github.com/Tencent/WeKnora/commit/def582b98a2960ceb53d7051f27b90014c08858f) | test(im/wecom): satisfy errcheck and lll in endpoint tests | 构建/测试/文档或体验调整 |
+| 2026-10-10 | [769a5aeca](https://github.com/Tencent/WeKnora/commit/769a5aeca84331b1f1f5ad5f3b93934ec33d4fbf) | fix(frontend): preserve the caret when outdenting Markdown lists | 有效变更,按主题归纳 |
+| 2026-10-10 | [57f8a65fe](https://github.com/Tencent/WeKnora/commit/57f8a65fec7b86fb20aa2c19a0e21b67a3864c73) | fix(datasource/dingtalk): tighten Validate sampling edges (#4079) | 有效变更,按主题归纳 |
+| 2026-10-10 | [932b750b2](https://github.com/Tencent/WeKnora/commit/932b750b2f5ef1ce2537ac57a117dec7423f46ab) | fix(agent): preserve steering order across persistence failures (#4078) | 有效变更,按主题归纳 |
+| 2026-10-10 | [0466c294e](https://github.com/Tencent/WeKnora/commit/0466c294ef1ad728e889058d331a5cfc37b2bd23) | fix(agent): 将空闲超时覆盖到模型流建立阶段 | 有效变更,按主题归纳 |
+| 2026-10-10 | [3f9af2afa](https://github.com/Tencent/WeKnora/commit/3f9af2afa912e16ebaec1ae172963a99dfc2e419) | fix(sqlite): add MCP metadata migration | 有效变更,按主题归纳 |
+| 2026-10-10 | [45a022aac](https://github.com/Tencent/WeKnora/commit/45a022aac23448e270f23046c9018cf8b2e37761) | fix: wait for qdrant writes to finish | 有效变更,按主题归纳 |
+| 2026-10-10 | [89e3273da](https://github.com/Tencent/WeKnora/commit/89e3273daf3c11f1a4498183b75c8257efd8db6c) | fix(mcp): preserve disabled state on service creation | 有效变更,按主题归纳 |
+| 2026-10-10 | [0d39af121](https://github.com/Tencent/WeKnora/commit/0d39af1214ff9d02382a4e851d320f5be8957985) | fix(retrieval): skip KB retrieval for summarize intent | 有效变更,按主题归纳 |
+| 2026-10-10 | [a6327f8e3](https://github.com/Tencent/WeKnora/commit/a6327f8e39fe62b299348e16a9bc81763c0fcf00) | fix(i18n): correct Korean username rule message to match the pattern | 有效变更,按主题归纳 |
+| 2026-10-10 | [b1d92048c](https://github.com/Tencent/WeKnora/commit/b1d92048cf61d732a6cd9171237c8d05770183e1) | fix(agent): account for file paths in read pagination budgets | 有效变更,按主题归纳 |
+| 2026-10-10 | [54cb6e9f9](https://github.com/Tencent/WeKnora/commit/54cb6e9f9be8cf486bb88b62916f13e16351f9a5) | test(docreader): add unit tests for utils.split text-splitting helpers | 构建/测试/文档或体验调整 |
+| 2026-10-10 | [150d8cc60](https://github.com/Tencent/WeKnora/commit/150d8cc605ed9305a5fc4ad1b8789dae238decce) | feat(sandbox): Docker 后端支持交互式终端 (#4023) | 有效变更,按主题归纳 |
+| 2026-10-11 | [231a5c0b1](https://github.com/Tencent/WeKnora/commit/231a5c0b15b817bb13ee3a876c3fbaf337cb4d5e) | feat(wiki): improve graph layout, theme captions and group focus (#4147) | 有效变更,按主题归纳 |
+| 2026-10-11 | [d2dbf7209](https://github.com/Tencent/WeKnora/commit/d2dbf7209963f0a9f6fafb70358ef9baec999aa3) | style: polish sidebar, list pages, toolbox and citation UI (#4154) | 构建/测试/文档或体验调整 |
+| 2026-10-11 | [17dc7da51](https://github.com/Tencent/WeKnora/commit/17dc7da514e1fd86f6c2fe77ab05c799cf34c27a) | feat(gallery): replace pagination with infinite masonry scrolling (#4156) | 有效变更,按主题归纳 |
+| 2026-10-11 | [643f64180](https://github.com/Tencent/WeKnora/commit/643f64180eb1898a7a2e32d517ff73f978d5448f) | feat(system): add platform user management and streamline settings (#4153) | 有效变更,按主题归纳 |
+| 2026-10-11 | [5343ac47b](https://github.com/Tencent/WeKnora/commit/5343ac47b730d451ad267004530084e08a5f599b) | fix(models): stop sending temperature to Claude models that reject it (#4155) | 有效变更,按主题归纳 |
+| 2026-10-11 | [00474f734](https://github.com/Tencent/WeKnora/commit/00474f734d6c7c0911d7485465d7c7286d0047cb) | fix: align settings layouts and tolerate existing migration objects (#4164) | 有效变更,按主题归纳 |
+| 2026-10-12 | [9efe94cc6](https://github.com/Tencent/WeKnora/commit/9efe94cc6f6d623590098f1fb355dd1f2e63b8f5) | feat(models): add Claude Opus 5.5, Sonnet 5.5 and Haiku 5.5 to the catalog (#4158) | 有效变更,按主题归纳 |
+| 2026-10-12 | [d272a83cf](https://github.com/Tencent/WeKnora/commit/d272a83cfc5d4a92aa3655f6a6c806aa40a34c70) | fix(deploy): pass WEKNORA_OUTBOUND_DIAL_TIMEOUT through to the app container (#4157) | 有效变更,按主题归纳 |
+| 2026-10-12 | [be581c41e](https://github.com/Tencent/WeKnora/commit/be581c41e0f17db25fd50493c345a5c43a8db138) | fix(mcp): close only the authorizing user's connection after OAuth (#4138) | 有效变更,按主题归纳 |
+| 2026-10-12 | [21ef506fc](https://github.com/Tencent/WeKnora/commit/21ef506fc995259e48680477ce8eb06556bbd957) | fix(agent): scope database queries to authorized shared knowledge bases (#4136) | 有效变更,按主题归纳 |
+| 2026-10-12 | [3c36cb19a](https://github.com/Tencent/WeKnora/commit/3c36cb19a2bb4214835cf8e5d520d4aa64ac8f23) | fix(deploy): bump paradedb image to v0.25.11-pg17 (#4120) | 有效变更,按主题归纳 |
+| 2026-10-12 | [c6bfe8189](https://github.com/Tencent/WeKnora/commit/c6bfe81894be411c910ad573bd1c80d88452414b) | feat(helm): allow Service name overrides with matching consumer references (#4109) | 有效变更,按主题归纳 |
+| 2026-10-12 | [e5f9f9dea](https://github.com/Tencent/WeKnora/commit/e5f9f9deab36898676fe85a9fda6b3e6f00288fa) | feat(rerank): add Cohere provider for the v2 rerank API | 有效变更,按主题归纳 |
+| 2026-10-12 | [fe444c851](https://github.com/Tencent/WeKnora/commit/fe444c851ef4430bcdcbef4cb10342839c57807f) | fix(feishu): add shared request pacing and improve HTTP 429 / error 99991400 retries in Feishu Wiki synchronization (#4039) | 有效变更,按主题归纳 |
+| 2026-10-12 | [24b5ca3f2](https://github.com/Tencent/WeKnora/commit/24b5ca3f2325e28acb4122384459ca66acc85ef4) | docs: clarify frontend development proxy configuration (#4171) | 构建/测试/文档或体验调整 |
+| 2026-10-12 | [7925f774d](https://github.com/Tencent/WeKnora/commit/7925f774d1baa72e8244b5fc65324542049677a8) | fix(datasource/gitlab): emptied files no longer leave their old body searchable (#4008) | 有效变更,按主题归纳 |
+| 2026-10-12 | [9b031a657](https://github.com/Tencent/WeKnora/commit/9b031a657656cddaddafe3cb2e124ccfc2f0693e) | fix: accept qdrant collection created concurrently (#3999) | 有效变更,按主题归纳 |
+| 2026-10-12 | [7982c6a3b](https://github.com/Tencent/WeKnora/commit/7982c6a3bda4b91da402faf39ca381cabd4832ca) | fix(agent): accept adjacent prefixed Python string literals (#4006) | 有效变更,按主题归纳 |
+| 2026-10-12 | [6a7094cea](https://github.com/Tencent/WeKnora/commit/6a7094ceacd3d9844ab9f2ad9083075e1f7cd088) | fix(mcp): bound resource listing pagination (#4007) | 有效变更,按主题归纳 |
+| 2026-10-12 | [ecdcd14e1](https://github.com/Tencent/WeKnora/commit/ecdcd14e1db5452e929ea846b0a1677502f81464) | docs: refresh LICENSE third-party notices (#4178) | 构建/测试/文档或体验调整 |
+| 2026-10-12 | [c2edec192](https://github.com/Tencent/WeKnora/commit/c2edec192b5b4abbe77c0e210b3e5abfb51792cc) | docs: explain knowledge upload folder paths (#4177) | 构建/测试/文档或体验调整 |
+| 2026-10-12 | [089711f92](https://github.com/Tencent/WeKnora/commit/089711f92c50ecad782ab95dbce8b04278a9cb35) | fix(mcp): ignore discover probe in resource list tests (#4179) | 有效变更,按主题归纳 |
+| 2026-10-12 | [98362dd29](https://github.com/Tencent/WeKnora/commit/98362dd29698f4b78df5d8ecf3f5020517e50ddc) | feat(gallery): filter images by source document (#4159) | 有效变更,按主题归纳 |
+| 2026-10-12 | [1c54beb55](https://github.com/Tencent/WeKnora/commit/1c54beb551673f15c6054ffb7f224a4bfdb495e3) | refactor(gallery): turn source document filter into a toolbar popover (#4190) | 有效变更,按主题归纳 |
+| 2026-10-12 | [ef54e835c](https://github.com/Tencent/WeKnora/commit/ef54e835c907718cedf7316bfa93daf69a28b2ba) | docs(api): clarify knowledge-search scope union of knowledge_base_ids and knowledge_ids (#4191) | 构建/测试/文档或体验调整 |
+| 2026-10-12 | [8fae51c64](https://github.com/Tencent/WeKnora/commit/8fae51c64084a5cc938eadb32d27c61eea4f4d87) | style(frontend): polish knowledge base settings hints and sub-sections (#4193) | 构建/测试/文档或体验调整 |
+| 2026-10-12 | [fd4cc4584](https://github.com/Tencent/WeKnora/commit/fd4cc4584884f7d21d8949d40de5404aa6b14a40) | fix(frontend): respect Markdown indentation selection boundaries (#4195) | 有效变更,按主题归纳 |
+| 2026-10-12 | [97796c637](https://github.com/Tencent/WeKnora/commit/97796c637aa53713edc846bd0a39470d687f0fdf) | fix(helm): configure app rollouts and document shared storage (#4180) | 有效变更,按主题归纳 |
+| 2026-10-13 | [809bda7b2](https://github.com/Tencent/WeKnora/commit/809bda7b2f2d44952eaa8765d91e3eb666a42c1c) | fix(agent): make agent tool selection match the tools the runtime registers (#4211) | 有效变更,按主题归纳 |
+| 2026-10-13 | [194bffbb0](https://github.com/Tencent/WeKnora/commit/194bffbb0a2b07a9a4bcf35355d18ef56dc6535b) | feat(agent): read chat attachments on demand (#4203) | 有效变更,按主题归纳 |
+| 2026-10-13 | [e0712df3e](https://github.com/Tencent/WeKnora/commit/e0712df3e020c3b9744a333cfa102b5a6e6371f7) | feat(mcp): let callers without a browser obtain MCP OAuth tokens (#4229) | 有效变更,按主题归纳 |
+| 2026-10-13 | [ec382d791](https://github.com/Tencent/WeKnora/commit/ec382d79114cef327da3f857004077f725da110c) | fix(frontend): preserve inline images in EPUB previews (#4226) | 有效变更,按主题归纳 |
+| 2026-10-13 | [a7a5f4762](https://github.com/Tencent/WeKnora/commit/a7a5f47625f31698897a62c152b1e186c95b43c7) | fix(frontend): preserve nonempty lists when pressing Enter (#4224) | 有效变更,按主题归纳 |
+| 2026-10-13 | [7972cdd25](https://github.com/Tencent/WeKnora/commit/7972cdd251acdb599eb41dcb66ad312982f82360) | fix(knowledge): make the knowledge base description token budget configurable (#4230) | 有效变更,按主题归纳 |
+| 2026-10-13 | [621bc9b98](https://github.com/Tencent/WeKnora/commit/621bc9b98ad19b32f553da83f6cbb14f7f74a4d6) | fix(website-docs): prebundle fastdom when mermaid imports it (#4208) | 有效变更,按主题归纳 |
+| 2026-10-13 | [8af7938ae](https://github.com/Tencent/WeKnora/commit/8af7938aeddfaaf55fb0578817482a3468f2a045) | fix(mcp): keep OAuth token expiry independent of the process time zone (#4207) | 有效变更,按主题归纳 |
+| 2026-10-13 | [94fae6de4](https://github.com/Tencent/WeKnora/commit/94fae6de43a36e44d64aecbbdf19c0b72b352aa8) | feat(models): add WeMM embedding presets and Transformers serving (#4238) | 有效变更,按主题归纳 |
+| 2026-10-13 | [4356f8c16](https://github.com/Tencent/WeKnora/commit/4356f8c16dcd5cae85652e9afebf63dc280334fc) | fix(skill): do not fail installs on non-blocking observations (#4246) | 有效变更,按主题归纳 |
+| 2026-10-13 | [e51dee829](https://github.com/Tencent/WeKnora/commit/e51dee8297e10db8b0037a8e0cf5d684010cea44) | feat(chat): polish rewind confirm and mark sessions with unread replies (#4247) | 有效变更,按主题归纳 |
+| 2026-10-13 | [2511a1f93](https://github.com/Tencent/WeKnora/commit/2511a1f932a7a0dc0e9a04edb62986df0b147cca) | fix(mcp): offer OAuth prompt after tool approval and tidy tools panel (#4248) | 有效变更,按主题归纳 |
+| 2026-10-13 | [923822086](https://github.com/Tencent/WeKnora/commit/92382208674ae1b9a2dd62ac0fa98d67ce348bfe) | feat(chat): add global search to the @ mention menu and align its styling (#4249) | 有效变更,按主题归纳 |
+| 2026-10-13 | [31c605568](https://github.com/Tencent/WeKnora/commit/31c6055687d33f9ad7a92cf453a264ce13a1700d) | feat(agent): name the turn's chat model on the completion event (#4252) | 有效变更,按主题归纳 |
diff --git a/docs/swagger.json b/docs/swagger.json
index 458c7edd..44208bad 100644
--- a/docs/swagger.json
+++ b/docs/swagger.json
@@ -775,31 +775,47 @@
}
}
},
- "/api/v1/knowledge/{id}/spans": {
+ "/artifacts": {
"get": {
- "description": "返回该知识在解析流水线的 trace tree(root → stage → subspan):每段状态、耗时、input/output、错误码、langfuse_trace_id。支持 ?attempt=N 查看历史尝试;不传则返回最新尝试。前端用于渲染时间线 + 多模态/embedding 子节点 + 一键跳转 Langfuse。",
- "consumes": [
- "application/json"
+ "security": [
+ {
+ "Bearer": []
+ },
+ {
+ "ApiKeyAuth": []
+ }
],
+ "description": "跨会话列出当前用户可见会话中技能生成的文件,每个文件只返回最新版本(不含存储 URL)",
"produces": [
"application/json"
],
"tags": [
- "知识管理"
+ "会话"
],
- "summary": "获取知识文档解析的 Span 树(含历史尝试)",
+ "summary": "列出我的产物",
"parameters": [
{
"type": "string",
- "description": "知识ID",
- "name": "id",
- "in": "path",
- "required": true
+ "description": "按文件名过滤",
+ "name": "keyword",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "逗号分隔的扩展名,如 .pdf,.pptx",
+ "name": "file_types",
+ "in": "query"
},
{
"type": "integer",
- "description": "指定尝试号;省略=最新",
- "name": "attempt",
+ "description": "页码",
+ "name": "page",
+ "in": "query"
+ },
+ {
+ "type": "integer",
+ "description": "每页数量(最大 100)",
+ "name": "page_size",
"in": "query"
}
],
@@ -812,6 +828,65 @@
}
}
}
+ },
+ "delete": {
+ "security": [
+ {
+ "Bearer": []
+ }
+ ],
+ "description": "删除产物库中的一个文件。产物库一行代表一个文件的最新版本,默认连同该文件在本会话中的所有历史版本一并删除",
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "会话"
+ ],
+ "summary": "删除产物库中的文件",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "会话ID",
+ "name": "session_id",
+ "in": "query",
+ "required": true
+ },
+ {
+ "type": "string",
+ "description": "消息ID",
+ "name": "message_id",
+ "in": "query",
+ "required": true
+ },
+ {
+ "type": "integer",
+ "description": "产物在消息中的位置",
+ "name": "index",
+ "in": "query",
+ "required": true
+ },
+ {
+ "type": "boolean",
+ "description": "是否删除所有版本,默认 true",
+ "name": "all_versions",
+ "in": "query"
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "OK",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "404": {
+ "description": "Not Found",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ }
+ }
}
},
"/auth/auto-setup": {
@@ -1338,7 +1413,7 @@
},
"/auth/register-by-invite": {
"post": {
- "description": "通过 Owner 生成的共享邀请链接 token 完成注册,绕过 invite_only 模式拦截。\n注册者自填邮箱(与 token 不绑定);注册成功后自动加入对应空间。",
+ "description": "通过共享邀请链接 token 完成注册,仅 self_serve / invite_register 模式允许。\n注册者自填邮箱(与 token 不绑定);注册成功后自动加入对应空间。",
"consumes": [
"application/json"
],
@@ -1373,6 +1448,12 @@
"$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
}
},
+ "403": {
+ "description": "注册功能已禁用",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ },
"409": {
"description": "邮箱已注册",
"schema": {
@@ -2659,7 +2740,7 @@
"ApiKeyAuth": []
}
],
- "description": "更新指定 IM 渠道的名称、模式、知识库、凭证或启用状态",
+ "description": "更新指定 IM 渠道的名称、模式、回复语言、知识库、凭证或启用状态",
"consumes": [
"application/json"
],
@@ -2679,7 +2760,7 @@
"required": true
},
{
- "description": "更新字段(name/mode/output_mode/knowledge_base_id/credentials/enabled)",
+ "description": "更新字段(含 locale)",
"name": "request",
"in": "body",
"required": true,
@@ -4932,7 +5013,7 @@
"ApiKeyAuth": []
}
],
- "description": "在知识库中执行向量和关键词混合搜索。推荐使用 POST;GET 携带 JSON 请求体仍受支持(兼容旧客户端)。",
+ "description": "底层召回:向量+关键词混合检索,默认不 rerank(可用 rerank 字段开启);一般检索请用 /knowledge-search。推荐 POST,GET 带 JSON 体仅兼容旧客户端。",
"consumes": [
"application/json"
],
@@ -4997,7 +5078,7 @@
"ApiKeyAuth": []
}
],
- "description": "在知识库中执行向量和关键词混合搜索。推荐使用 POST;GET 携带 JSON 请求体仍受支持(兼容旧客户端)。",
+ "description": "底层召回:向量+关键词混合检索,默认不 rerank(可用 rerank 字段开启);一般检索请用 /knowledge-search。推荐 POST,GET 带 JSON 体仅兼容旧客户端。",
"consumes": [
"application/json"
],
@@ -5054,6 +5135,166 @@
}
}
},
+ "/knowledge-bases/{id}/images": {
+ "get": {
+ "security": [
+ {
+ "Bearer": []
+ },
+ {
+ "ApiKeyAuth": []
+ }
+ ],
+ "description": "浏览某知识库内的全部图片资产,支持关键字搜索(caption/OCR 文本)、按图片属性筛选(attr_filters / attr_rules)与排序分页。",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "knowledge"
+ ],
+ "summary": "列出知识库图片资产",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "知识库 ID",
+ "name": "id",
+ "in": "path",
+ "required": true
+ },
+ {
+ "type": "integer",
+ "description": "页码,默认 1",
+ "name": "page",
+ "in": "query"
+ },
+ {
+ "type": "integer",
+ "description": "每页数量,默认 20,最大 1000",
+ "name": "page_size",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "关键字,对 search_in 指定字段做不区分大小写的子串匹配",
+ "name": "keyword",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "参与搜索的属性 ID(逗号分隔);须为 in_searchfield=true",
+ "name": "search_in",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "排序属性 ID;须为 in_sortfield=true,默认 created_at",
+ "name": "sort_by",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "排序方向:asc / desc,默认 desc",
+ "name": "sort_order",
+ "in": "query"
+ },
+ {
+ "type": "boolean",
+ "description": "仅包含启用状态的图片",
+ "name": "is_enabled",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "来源文档 ID 列表(逗号分隔,最多 50 个),仅返回这些文档的图片",
+ "name": "knowledge_ids",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "属性筛选 JSON,同一属性多值 OR、属性间 AND",
+ "name": "attr_filters",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "属性逐值裁决 JSON(off 隐藏 / on 强制显示,on 优先)",
+ "name": "attr_rules",
+ "in": "query"
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "图片资产分页列表",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "400": {
+ "description": "请求参数错误",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ }
+ }
+ },
+ "post": {
+ "security": [
+ {
+ "Bearer": []
+ },
+ {
+ "ApiKeyAuth": []
+ }
+ ],
+ "description": "与 GET /images 同义,筛选参数改经 JSON body 传递:knowledge_ids 不再受 URL 长度限制(上限 1000),供来源文档全选使用。",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "knowledge"
+ ],
+ "summary": "列出知识库图片资产(body 版)",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "知识库 ID",
+ "name": "id",
+ "in": "path",
+ "required": true
+ },
+ {
+ "description": "筛选参数(与 GET /images 同名参数同义)",
+ "name": "body",
+ "in": "body",
+ "schema": {
+ "$ref": "#/definitions/internal_handler.imageListRequestBody"
+ }
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "图片资产分页列表",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "400": {
+ "description": "请求参数错误",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ }
+ }
+ }
+ },
"/knowledge-bases/{id}/knowledge": {
"get": {
"security": [
@@ -5148,6 +5389,18 @@
"description": "为 true 时同时返回子文件夹内的文档",
"name": "folder_recursive",
"in": "query"
+ },
+ {
+ "type": "string",
+ "description": "排序字段: updated_at/created_at/file_name,默认 created_at",
+ "name": "sort_by",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "排序方向: asc/desc,默认 desc",
+ "name": "sort_order",
+ "in": "query"
}
],
"responses": {
@@ -5740,6 +5993,50 @@
}
}
},
+ "/knowledge-bases/{id}/profile/generate": {
+ "post": {
+ "security": [
+ {
+ "Bearer": []
+ },
+ {
+ "ApiKeyAuth": []
+ }
+ ],
+ "description": "基于文档画像聚合,立即重新生成知识库的 AI 描述(不覆盖手写描述)",
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "知识库"
+ ],
+ "summary": "生成知识库描述",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "知识库ID",
+ "name": "id",
+ "in": "path",
+ "required": true
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "生成的知识库画像",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "400": {
+ "description": "知识库类型不支持或未配置模型",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ }
+ }
+ }
+ },
"/knowledge-bases/{id}/shares": {
"get": {
"security": [
@@ -6270,7 +6567,7 @@
"ApiKeyAuth": []
}
],
- "description": "在知识库中搜索(不使用LLM总结)",
+ "description": "在知识库中搜索(不使用LLM总结)。与产品内问答使用同一检索流程(召回、rerank、合并),外部检索首选;可覆盖召回参数与 rerank",
"consumes": [
"application/json"
],
@@ -7258,6 +7555,53 @@
}
}
},
+ "/knowledge/{id}/spans": {
+ "get": {
+ "security": [
+ {
+ "Bearer": []
+ },
+ {
+ "ApiKeyAuth": []
+ }
+ ],
+ "description": "返回该知识在解析流水线的 trace tree(root → stage → subspan):每段状态、耗时、input/output、错误码、langfuse_trace_id。支持 ?attempt=N 查看历史尝试;不传则返回最新尝试。前端用于渲染时间线 + 多模态/embedding 子节点 + 一键跳转 Langfuse。",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "知识管理"
+ ],
+ "summary": "获取知识文档解析的 Span 树(含历史尝试)",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "知识ID",
+ "name": "id",
+ "in": "path",
+ "required": true
+ },
+ {
+ "type": "integer",
+ "description": "指定尝试号;省略=最新",
+ "name": "attempt",
+ "in": "query"
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "OK",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ }
+ }
+ }
+ },
"/knowledgebase/{kb_id}/wiki/auto-fix": {
"post": {
"security": [
@@ -8676,6 +9020,20 @@
"MCP服务"
],
"summary": "获取MCP服务列表",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "Agent ID; needs agent_source_tenant_id",
+ "name": "agent_id",
+ "in": "query"
+ },
+ {
+ "type": "integer",
+ "description": "Shared agent source workspace",
+ "name": "agent_source_tenant_id",
+ "in": "query"
+ }
+ ],
"responses": {
"200": {
"description": "MCP服务列表",
@@ -8689,6 +9047,12 @@
"schema": {
"$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
}
+ },
+ "403": {
+ "description": "无权使用该共享智能体",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
}
}
},
@@ -10817,6 +11181,65 @@
}
}
},
+ "/models/catalog/resolve": {
+ "get": {
+ "security": [
+ {
+ "Bearer": []
+ },
+ {
+ "ApiKeyAuth": []
+ }
+ ],
+ "description": "根据厂商、模型名、Base URL 与 extra_config 返回目录解析结果(协议、思考等级、上下文等),供模型编辑器实时展示",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "模型管理"
+ ],
+ "summary": "解析模型的有效接入配置",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "厂商标识",
+ "name": "provider",
+ "in": "query",
+ "required": true
+ },
+ {
+ "type": "string",
+ "description": "模型名",
+ "name": "model",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "Base URL",
+ "name": "base_url",
+ "in": "query"
+ },
+ {
+ "type": "string",
+ "description": "模型类型",
+ "name": "model_type",
+ "in": "query"
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "解析结果",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ }
+ }
+ }
+ },
"/models/providers": {
"get": {
"security": [
@@ -10827,7 +11250,7 @@
"ApiKeyAuth": []
}
],
- "description": "根据模型类型获取支持的厂商列表及配置信息",
+ "description": "根据模型类型获取支持的厂商定义(含图标、默认地址、额外字段、内置模型目录与思考能力)",
"consumes": [
"application/json"
],
@@ -10841,7 +11264,7 @@
"parameters": [
{
"type": "string",
- "description": "模型类型 (chat, embedding, rerank, vllm)",
+ "description": "模型类型 (chat, embedding, rerank, vllm, asr)",
"name": "model_type",
"in": "query"
}
@@ -11045,6 +11468,68 @@
}
}
},
+ "/models/{id}/copy": {
+ "post": {
+ "security": [
+ {
+ "Bearer": []
+ },
+ {
+ "ApiKeyAuth": []
+ }
+ ],
+ "description": "复制模型配置。name 与源模型一致,凭证从已存记录复制,不接受客户端重传。",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "模型管理"
+ ],
+ "summary": "复制模型",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "源模型ID",
+ "name": "id",
+ "in": "path",
+ "required": true
+ },
+ {
+ "description": "副本展示名",
+ "name": "request",
+ "in": "body",
+ "required": true,
+ "schema": {
+ "$ref": "#/definitions/internal_handler.CopyModelRequest"
+ }
+ }
+ ],
+ "responses": {
+ "201": {
+ "description": "复制的模型",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "400": {
+ "description": "请求参数错误",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ },
+ "404": {
+ "description": "模型不存在",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ }
+ }
+ }
+ },
"/organizations": {
"get": {
"security": [
@@ -13793,6 +14278,67 @@
"responses": {}
}
},
+ "/sessions/{session_id}/messages/{message_id}/artifacts/{index}": {
+ "delete": {
+ "security": [
+ {
+ "Bearer": []
+ }
+ ],
+ "description": "删除某条 assistant 消息生成的单个文件;all_versions=true 时连同本会话中该文件的所有历史版本一并删除",
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "会话"
+ ],
+ "summary": "删除会话中的产物文件",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "会话ID",
+ "name": "session_id",
+ "in": "path",
+ "required": true
+ },
+ {
+ "type": "string",
+ "description": "消息ID",
+ "name": "message_id",
+ "in": "path",
+ "required": true
+ },
+ {
+ "type": "integer",
+ "description": "产物在消息中的位置",
+ "name": "index",
+ "in": "path",
+ "required": true
+ },
+ {
+ "type": "boolean",
+ "description": "是否删除该文件的所有版本",
+ "name": "all_versions",
+ "in": "query"
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "OK",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "404": {
+ "description": "Not Found",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ }
+ }
+ }
+ },
"/sessions/{session_id}/messages/{message_id}/artifacts/{index}/download": {
"get": {
"responses": {}
@@ -13947,6 +14493,80 @@
}
}
},
+ "/sessions/{session_id}/rewind": {
+ "post": {
+ "security": [
+ {
+ "Bearer": []
+ },
+ {
+ "ApiKeyAuth": []
+ }
+ ],
+ "description": "回滚当前会话到指定用户或助手消息:删除其后消息,可到达时 git reset 工作区。",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "会话"
+ ],
+ "summary": "回滚会话",
+ "parameters": [
+ {
+ "type": "string",
+ "description": "会话 ID",
+ "name": "session_id",
+ "in": "path",
+ "required": true
+ },
+ {
+ "description": "回滚请求",
+ "name": "request",
+ "in": "body",
+ "required": true,
+ "schema": {
+ "$ref": "#/definitions/internal_handler_session.RewindSessionRequest"
+ }
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "回滚结果",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "400": {
+ "description": "请求参数错误 / 回滚点角色不支持",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ },
+ "404": {
+ "description": "会话或消息不存在",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ },
+ "409": {
+ "description": "会话正在生成中 / 无检查点 / 沙箱已更换",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ },
+ "500": {
+ "description": "工作区回滚失败",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ }
+ }
+ }
+ },
"/sessions/{session_id}/steer": {
"post": {
"security": [
@@ -14318,9 +14938,21 @@
"parameters": [
{
"type": "string",
- "description": "Sandbox config ID",
+ "description": "Sandbox config ID; ignored for a shared agent",
"name": "sandbox_config_id",
"in": "query"
+ },
+ {
+ "type": "string",
+ "description": "Agent ID; needs agent_source_tenant_id",
+ "name": "agent_id",
+ "in": "query"
+ },
+ {
+ "type": "integer",
+ "description": "Shared agent source workspace",
+ "name": "agent_source_tenant_id",
+ "in": "query"
}
],
"responses": {
@@ -14330,6 +14962,12 @@
"type": "object",
"additionalProperties": true
}
+ },
+ "403": {
+ "description": "无权使用该共享智能体",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
}
}
}
@@ -15182,6 +15820,198 @@
}
}
},
+ "/system/admin/model-catalog": {
+ "get": {
+ "security": [
+ {
+ "Bearer": []
+ }
+ ],
+ "description": "System administrator JWT only; API keys are not accepted. Returns built-in, deployment and effective\ncatalogs, the current overlay, version/baseline tokens, and the latest 20 published revisions.",
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "系统管理"
+ ],
+ "summary": "Get the system model catalog",
+ "responses": {
+ "200": {
+ "description": "OK",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogState"
+ }
+ },
+ "401": {
+ "description": "Unauthorized",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "403": {
+ "description": "Forbidden",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "503": {
+ "description": "Service Unavailable",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ }
+ }
+ },
+ "put": {
+ "security": [
+ {
+ "Bearer": []
+ }
+ ],
+ "description": "System administrator JWT only; API keys are not accepted. Validates and publishes a new overlay\nrevision using the current version and baseline. Stale values return 409. To restore a historical\nconfiguration, submit its overlay with the current version and baseline.",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "系统管理"
+ ],
+ "summary": "Publish a model catalog overlay",
+ "parameters": [
+ {
+ "description": "Version, baseline and overlay object (max 1 MiB)",
+ "name": "request",
+ "in": "body",
+ "required": true,
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogUpdate"
+ }
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "OK",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogState"
+ }
+ },
+ "400": {
+ "description": "Bad Request",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "401": {
+ "description": "Unauthorized",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "403": {
+ "description": "Forbidden",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "409": {
+ "description": "Conflict",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "500": {
+ "description": "Internal Server Error",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ }
+ }
+ }
+ },
+ "/system/admin/model-catalog/preview": {
+ "post": {
+ "security": [
+ {
+ "Bearer": []
+ }
+ ],
+ "description": "System administrator JWT only; API keys are not accepted. Validates the overlay without publishing\nit. Supply the current version and baseline; stale values return 409. The response contains the\ncandidate effective catalog and normalized overlay; history, builtin and deployment are null.",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "系统管理"
+ ],
+ "summary": "Preview a model catalog overlay",
+ "parameters": [
+ {
+ "description": "Version, baseline and overlay object (max 1 MiB)",
+ "name": "request",
+ "in": "body",
+ "required": true,
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogUpdate"
+ }
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "OK",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogState"
+ }
+ },
+ "400": {
+ "description": "Bad Request",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "401": {
+ "description": "Unauthorized",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "403": {
+ "description": "Forbidden",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "409": {
+ "description": "Conflict",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "500": {
+ "description": "Internal Server Error",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ }
+ }
+ }
+ },
"/system/admin/promote": {
"post": {
"description": "Grant system administrator privileges to a user (SystemAdmin only).\nIdempotent: re-promoting an existing system admin returns 200 with no DB write.\nIdentify the user by email (preferred for human operators) or user_id (UUID, for API clients).",
@@ -15610,6 +16440,49 @@
}
}
},
+ "/system/admin/users": {
+ "get": {
+ "description": "Retrieve a paginated list of every user across tenants (SystemAdmin only).\nSupports `offset` (default 0) and `limit` (default 50, max 200).\nPassword/hashes are never serialised — only the redacted UserInfo shape.",
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "System Admin"
+ ],
+ "summary": "List all platform users",
+ "parameters": [
+ {
+ "type": "integer",
+ "default": 0,
+ "description": "Page offset",
+ "name": "offset",
+ "in": "query"
+ },
+ {
+ "type": "integer",
+ "default": 50,
+ "description": "Page size (max 200)",
+ "name": "limit",
+ "in": "query"
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "Users retrieved successfully",
+ "schema": {
+ "$ref": "#/definitions/internal_handler.ListAllUsersResponse"
+ }
+ },
+ "403": {
+ "description": "Forbidden: not a system admin",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ }
+ }
+ }
+ },
"/system/admin/users/create": {
"post": {
"description": "Provision a new local user account (SystemAdmin only).\nWhen `password` is omitted or null, a cryptographically random\npassword is generated (OIDC-style crypto/rand + base64url)\nand returned once in the response body. Any provided value,\nincluding empty string, is policy-checked. Tenant provisioning\nfollows the shared auth.default_tenant_mode policy.",
@@ -15734,6 +16607,55 @@
}
}
},
+ "/system/admin/users/status": {
+ "post": {
+ "description": "Disable revokes the target's active sessions immediately.\nSelf-disable and disabling the last active system administrator are rejected.",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "System Admin"
+ ],
+ "summary": "Enable or disable a user account",
+ "parameters": [
+ {
+ "description": "Account status request",
+ "name": "request",
+ "in": "body",
+ "required": true,
+ "schema": {
+ "$ref": "#/definitions/internal_handler.SetUserStatusRequest"
+ }
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "Status updated",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "400": {
+ "description": "Self-disable / last-admin guard",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "404": {
+ "description": "User not found",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ }
+ }
+ }
+ },
"/system/capabilities": {
"get": {
"description": "返回当前部署版本及实际注册的后端路由所对应的功能能力;仅 supported=false 表示入口应隐藏",
@@ -18039,9 +18961,183 @@
}
}
}
+ },
+ "/wiki-search": {
+ "post": {
+ "security": [
+ {
+ "Bearer": []
+ },
+ {
+ "ApiKeyAuth": []
+ }
+ ],
+ "description": "Cross-KB wiki search using the same POSIX regex ranking as single-KB wiki search",
+ "consumes": [
+ "application/json"
+ ],
+ "produces": [
+ "application/json"
+ ],
+ "tags": [
+ "Wiki"
+ ],
+ "summary": "Cross-KB wiki search",
+ "parameters": [
+ {
+ "description": "Wiki search request",
+ "name": "request",
+ "in": "body",
+ "required": true,
+ "schema": {
+ "$ref": "#/definitions/internal_handler.SearchWikiRequest"
+ }
+ }
+ ],
+ "responses": {
+ "200": {
+ "description": "OK",
+ "schema": {
+ "type": "object",
+ "additionalProperties": true
+ }
+ },
+ "400": {
+ "description": "Bad Request",
+ "schema": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError"
+ }
+ }
+ }
+ }
}
},
"definitions": {
+ "github_com_Tencent_WeKnora_internal_application_service.CatalogProviderView": {
+ "type": "object",
+ "properties": {
+ "api": {
+ "type": "string"
+ },
+ "id": {
+ "type": "string"
+ },
+ "model_thinking_levels": {
+ "description": "ModelThinkingLevels lists, per chat model id, the levels offered once\nthinking is enabled, resolved against the vendor map and protocol so\nthe console does not re-implement level inheritance.",
+ "type": "object",
+ "additionalProperties": {
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_models_api.ReasoningEffort"
+ }
+ }
+ },
+ "model_types": {
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ModelType"
+ }
+ },
+ "models": {
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_models.ModelSpec"
+ }
+ },
+ "name": {
+ "type": "string"
+ },
+ "settings": {
+ "type": "object",
+ "additionalProperties": {}
+ },
+ "vendor_thinking_levels": {
+ "description": "VendorThinkingLevels is what a chat model without its own map offers.",
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_models_api.ReasoningEffort"
+ }
+ }
+ }
+ },
+ "github_com_Tencent_WeKnora_internal_application_service.CatalogRevision": {
+ "type": "object",
+ "properties": {
+ "overlay": {
+ "type": "object"
+ },
+ "updated_at": {
+ "type": "string"
+ },
+ "updated_by": {
+ "type": "string"
+ },
+ "version": {
+ "type": "integer"
+ }
+ }
+ },
+ "github_com_Tencent_WeKnora_internal_application_service.CatalogState": {
+ "type": "object",
+ "properties": {
+ "applied_version": {
+ "type": "integer"
+ },
+ "baseline": {
+ "type": "string"
+ },
+ "builtin": {
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogProviderView"
+ }
+ },
+ "deployment": {
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogProviderView"
+ }
+ },
+ "deployment_error": {
+ "type": "string"
+ },
+ "effective": {
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogProviderView"
+ }
+ },
+ "history": {
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogRevision"
+ }
+ },
+ "overlay": {
+ "type": "object"
+ },
+ "sync_error": {
+ "type": "string"
+ },
+ "version": {
+ "type": "integer"
+ }
+ }
+ },
+ "github_com_Tencent_WeKnora_internal_application_service.CatalogUpdate": {
+ "type": "object",
+ "properties": {
+ "baseline": {
+ "type": "string"
+ },
+ "overlay": {
+ "type": "object"
+ },
+ "version": {
+ "type": "integer"
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_application_service.WikiLintIssue": {
"type": "object",
"properties": {
@@ -18317,6 +19413,152 @@
}
}
},
+ "github_com_Tencent_WeKnora_internal_models.ModelCost": {
+ "type": "object",
+ "properties": {
+ "cache_read": {
+ "type": "number"
+ },
+ "cache_write": {
+ "type": "number"
+ },
+ "currency": {
+ "type": "string"
+ },
+ "input": {
+ "type": "number"
+ },
+ "output": {
+ "type": "number"
+ }
+ }
+ },
+ "github_com_Tencent_WeKnora_internal_models.ModelSpec": {
+ "type": "object",
+ "properties": {
+ "aliases": {
+ "description": "Aliases are additional ids that resolve to this entry (dated\nsnapshots, vendor-prefixed names on gateways).",
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ },
+ "api": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_models_api.API"
+ },
+ "compat": {
+ "description": "Compat is the protocol-specific overlay for this model (flat object,\nfields depend on API). Decoded lazily by Resolve.",
+ "type": "array",
+ "items": {
+ "type": "integer"
+ }
+ },
+ "context_window": {
+ "type": "integer"
+ },
+ "cost": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_models.ModelCost"
+ },
+ "deprecated": {
+ "description": "Deprecated hides the entry from pickers but keeps resolution working.",
+ "type": "boolean"
+ },
+ "dimension": {
+ "description": "Dimension is the embedding vector width (embedding entries only).",
+ "type": "integer"
+ },
+ "id": {
+ "type": "string"
+ },
+ "input": {
+ "description": "Input lists accepted modalities: text, image, audio, video.",
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ },
+ "match": {
+ "description": "Match is a case-insensitive glob (\"gpt-5*\", \"*deepseek-r1*\") that lets\none entry describe a family. Exact ids and aliases win over patterns;\namong patterns the longest literal prefix wins.",
+ "type": "string"
+ },
+ "max_output_tokens": {
+ "type": "integer"
+ },
+ "name": {
+ "type": "string"
+ },
+ "reasoning": {
+ "description": "Reasoning reports whether the model can emit thinking at all.",
+ "type": "boolean"
+ },
+ "source": {
+ "description": "Source is the documentation URL the facts were taken from.",
+ "type": "string"
+ },
+ "thinking_levels": {
+ "description": "ThinkingLevels maps protocol-neutral levels to vendor values; missing\nkeys inherit the vendor map. See api.ThinkingLevelMap.",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_models_api.ThinkingLevelMap"
+ }
+ ]
+ },
+ "type": {
+ "description": "Type is the WeKnora model type this entry describes. Chat models with\nimage input are still KnowledgeQA; the UI derives VLM eligibility from\nInput. Embedding / Rerank / ASR entries exist for the model picker.",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ModelType"
+ }
+ ]
+ }
+ }
+ },
+ "github_com_Tencent_WeKnora_internal_models_api.API": {
+ "type": "string",
+ "enum": [
+ "openai-completions",
+ "openai-responses",
+ "anthropic-messages",
+ "google-generative-ai",
+ "ollama"
+ ],
+ "x-enum-varnames": [
+ "APIOpenAICompletions",
+ "APIOpenAIResponses",
+ "APIAnthropicMessages",
+ "APIGoogleGenerativeAI",
+ "APIOllama"
+ ]
+ },
+ "github_com_Tencent_WeKnora_internal_models_api.ReasoningEffort": {
+ "type": "string",
+ "enum": [
+ "off",
+ "auto",
+ "minimal",
+ "low",
+ "medium",
+ "high",
+ "xhigh",
+ "max"
+ ],
+ "x-enum-varnames": [
+ "ReasoningOff",
+ "ReasoningAuto",
+ "ReasoningMinimal",
+ "ReasoningLow",
+ "ReasoningMedium",
+ "ReasoningHigh",
+ "ReasoningXHigh",
+ "ReasoningMax"
+ ]
+ },
+ "github_com_Tencent_WeKnora_internal_models_api.ThinkingLevelMap": {
+ "type": "object",
+ "additionalProperties": {
+ "type": "string"
+ }
+ },
"github_com_Tencent_WeKnora_internal_models_limiter.RuntimeStat": {
"type": "object",
"properties": {
@@ -18400,6 +19642,13 @@
"description": "ReasoningContent stores the OpenAI-protocol reasoning_content emitted by the\nmodel in this round. Persisted on AgentStep so cross-turn replay can put it\nback on the assistant message — required by MiMo / DeepSeek V3.2+ thinking\nmode, ignored by providers that don't recognize the field.",
"type": "string"
},
+ "reasoning_metadata": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ProviderMetadata"
+ },
+ "reasoning_signature": {
+ "description": "ReasoningSignature / ReasoningMetadata are the provider artifacts that\nmust accompany ReasoningContent on replay (Anthropic signatures,\nOpenAI Responses encrypted reasoning items).",
+ "type": "string"
+ },
"thought": {
"description": "LLM's reasoning/thinking (Think phase)",
"type": "string"
@@ -18415,6 +19664,10 @@
"$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ToolCall"
}
},
+ "truncated": {
+ "description": "Truncated marks the round the completion-token cap cut off. It rides in\nthe agent_steps JSON so a reloaded, shared or re-opened conversation can\nstill show that the answer stops mid-sentence by design, rather than\nlooking finished. Live streaming carries the same fact on the answer\nevent; this is what survives the round trip.",
+ "type": "boolean"
+ },
"user_messages_before": {
"description": "UserMessagesBefore records consumed steer rows in delivery order, before\nthis model response. Unlike timestamps, this remains unambiguous on replay.",
"type": "array",
@@ -19163,6 +20416,10 @@
}
]
},
+ "reasoning_effort": {
+ "description": "ReasoningEffort selects the thinking intensity (off | auto | minimal |\nlow | medium | high | xhigh | max). Empty keeps the boolean Thinking\nsemantics: true means \"auto\". See internal/models/api.ReasoningEffort.",
+ "type": "string"
+ },
"rerank_model_id": {
"description": "ReRank model ID for retrieval",
"type": "string"
@@ -19729,6 +20986,22 @@
}
}
},
+ "github_com_Tencent_WeKnora_internal_types.GalleryUserPrefs": {
+ "type": "object",
+ "properties": {
+ "mode": {
+ "description": "Mode is \"all\" (every search-eligible field active) or \"custom\"\n(per-field Status governs). Empty = \"all\".",
+ "type": "string"
+ },
+ "status": {
+ "description": "Status maps a namespaced attribute id (\"builtin:caption\") to\n\"on\"/\"off\". Only consulted in custom mode; an unrecorded field is\noff. Source-declared on/off does not exist — activation is always\nthe user's own record.",
+ "type": "object",
+ "additionalProperties": {
+ "type": "string"
+ }
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_types.GraphNode": {
"type": "object",
"properties": {
@@ -19763,9 +21036,60 @@
}
}
},
+ "github_com_Tencent_WeKnora_internal_types.ImageActionsConfig": {
+ "type": "object",
+ "properties": {
+ "ocr": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ImageOCRAction"
+ }
+ }
+ },
+ "github_com_Tencent_WeKnora_internal_types.ImageAttrCondition": {
+ "type": "object",
+ "properties": {
+ "is": {
+ "type": "string"
+ },
+ "prop": {
+ "type": "string"
+ }
+ }
+ },
+ "github_com_Tencent_WeKnora_internal_types.ImageOCRAction": {
+ "type": "object",
+ "properties": {
+ "on": {
+ "description": "On is the list of attribute conditions that enable OCR. Matching any one\nturns OCR on.",
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ImageAttrCondition"
+ }
+ },
+ "on_unobserved": {
+ "description": "OnUnobserved is the OCR decision used when the observation does not\nanswer for one of the attributes the policy reads — the attribute is\nabsent because the model skipped the line, or because the value it wrote\nwas not one of the allowed ones. The default is true: an incomplete\nobservation keeps extracting, so a missed block of text costs one extra\ncall instead of being lost. Turn it off only with a model reliable enough\nto answer every attribute line.",
+ "type": "boolean"
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_types.ImageProcessingConfig": {
"type": "object",
"properties": {
+ "image_actions": {
+ "description": "ImageActions overrides the built-in attribute→work table (see\nDefaultImageActions). A nil value keeps the conservative default: run OCR\nfor block text, data visuals, and unobserved images.",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ImageActionsConfig"
+ }
+ ]
+ },
+ "image_attrs_enabled": {
+ "description": "ImageAttrsEnabled turns the attribute-observed image pipeline on for this\nknowledge base. It is off by default on purpose, so upgrading an\nexisting deployment never changes what happens to documents already\nbeing ingested: with it off every image is described and OCR'd exactly\nas before. When on, the describe round also observes image attributes and\nthe attribute policy decides whether OCR runs for it. A single upload can\noverride it per document through\nKnowledgeProcessOverrides.ImageAttrsEnabled.",
+ "type": "boolean"
+ },
+ "image_vector_enabled": {
+ "description": "ImageVectorEnabled lets a knowledge base whose embedding model takes\nimages index each image by its own vector and recall it with one. It\nis off by default on purpose: a model that takes images is also used\nfor plain text, and turning this on costs an embedding call per image\nand widens vector recall. Existing knowledge bases stay off after an\nupgrade because the stored JSON has no such key.",
+ "type": "boolean"
+ },
"model_id": {
"description": "Model ID",
"type": "string"
@@ -19869,7 +21193,7 @@
],
"properties": {
"representative_user_id": {
- "description": "RepresentativeUserID identifies the user attached to the OTM row for\ndisplay/audit. Optional: when unset, the handler picks a stable default\n(the user from the legacy UserID field, or the workspace's owner).",
+ "description": "RepresentativeUserID is accepted for compatibility and ignored: a\ndirect add attaches no user of the enrolled workspace, since the\ninviter must not choose whose details the roster shows.",
"type": "string"
},
"role": {
@@ -19885,7 +21209,7 @@
"type": "integer"
},
"user_id": {
- "description": "UserID is retained for backward compatibility. When set without\nTenantID, the handler resolves the user's TenantID and uses this\nuser as the representative.",
+ "description": "UserID is retained for backward compatibility. When set without\nTenantID, the handler resolves the user's TenantID.",
"type": "string"
}
}
@@ -20040,6 +21364,14 @@
}
]
},
+ "generated_profile": {
+ "description": "GeneratedProfile is the machine-generated description: a gist, merged\ntopics, typical questions and the aggregate snapshot they came from. It\nnever overwrites the user-authored Description; both are shown to agents.",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfile"
+ }
+ ]
+ },
"id": {
"description": "Unique identifier of the knowledge base",
"type": "string"
@@ -20088,6 +21420,14 @@
"description": "ProcessingCount indicates the number of knowledge items being processed (for document type knowledge bases)",
"type": "integer"
},
+ "profile_config": {
+ "description": "ProfileConfig controls automatic generation of the knowledge-base\ndescription from per-document profiles (document knowledge bases only).",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfileConfig"
+ }
+ ]
+ },
"question_generation_config": {
"description": "QuestionGenerationConfig stores question generation configuration for document knowledge bases",
"allOf": [
@@ -20186,7 +21526,7 @@
]
},
"image_processing_config": {
- "description": "Image processing configuration",
+ "description": "Image processing configuration.\n\nnil means \"no change\" when updating, the same contract IndexingStrategy\nuses. A request that does not mention this field must leave the image\nsettings alone: the attribute-observation switch and the attribute policy\nare things a knowledge base accumulates, and a client that predates them\nwould otherwise reset the lot on any save. Sending an object — empty\nincluded — replaces the whole configuration.",
"allOf": [
{
"$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ImageProcessingConfig"
@@ -20201,6 +21541,14 @@
}
]
},
+ "profile_config": {
+ "description": "ProfileConfig controls optional automatic knowledge-base description\ngeneration. nil means \"no change\" when updating.",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfileConfig"
+ }
+ ]
+ },
"wiki_config": {
"description": "Wiki configuration (only for wiki-enabled knowledge bases)",
"allOf": [
@@ -20211,6 +21559,128 @@
}
}
},
+ "github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfile": {
+ "type": "object",
+ "properties": {
+ "aggregate_hash": {
+ "description": "AggregateHash identifies the aggregate the gist was generated from. A\nmismatch with the live aggregate means the text is stale.",
+ "type": "string"
+ },
+ "error": {
+ "description": "Error holds the last generation error when Status is failed.",
+ "type": "string"
+ },
+ "generated_at": {
+ "description": "GeneratedAt records when the text was produced.",
+ "type": "string"
+ },
+ "gist": {
+ "description": "Gist is a short statement of what the knowledge base covers.",
+ "type": "string"
+ },
+ "model_id": {
+ "description": "ModelID records which chat model produced the text.",
+ "type": "string"
+ },
+ "stats": {
+ "description": "Stats is the aggregate snapshot the text was generated from.",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfileStats"
+ }
+ ]
+ },
+ "status": {
+ "description": "Status is one of the KnowledgeBaseProfileStatus* values.",
+ "type": "string"
+ },
+ "topics": {
+ "description": "Topics is the merged topic list (at most KnowledgeBaseProfileMaxTopics).",
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ },
+ "typical_questions": {
+ "description": "TypicalQuestions lists questions this knowledge base can answer.",
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ }
+ }
+ },
+ "github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfileConfig": {
+ "type": "object",
+ "properties": {
+ "custom_instructions": {
+ "description": "CustomInstructions is appended to the stable system prompt, e.g. the\naudience the description should address or terms to keep.",
+ "type": "string"
+ },
+ "enabled": {
+ "type": "boolean"
+ },
+ "model_id": {
+ "description": "ModelID overrides the chat model; empty falls back to the KB summary model.",
+ "type": "string"
+ }
+ }
+ },
+ "github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfileStats": {
+ "type": "object",
+ "properties": {
+ "doc_types": {
+ "description": "DocTypes counts documents per model-assigned document type.",
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.NamedCount"
+ }
+ },
+ "document_count": {
+ "description": "DocumentCount is the number of eligible documents at generation time.",
+ "type": "integer"
+ },
+ "earliest_at": {
+ "description": "EarliestAt and LatestAt bound document creation times.",
+ "type": "string"
+ },
+ "file_types": {
+ "description": "FileTypes counts documents per file extension.",
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.NamedCount"
+ }
+ },
+ "folders": {
+ "description": "Folders lists the top-level folder names present.",
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ },
+ "latest_at": {
+ "type": "string"
+ },
+ "profiled_count": {
+ "description": "ProfiledCount is how many of them carried a document profile.",
+ "type": "integer"
+ },
+ "raw_topics": {
+ "description": "RawTopics counts documents per normalized topic keyword before merging.",
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.NamedCount"
+ }
+ },
+ "tags": {
+ "description": "Tags counts documents per knowledge-base tag.",
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.NamedCount"
+ }
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_types.KnowledgeBaseShareResponse": {
"type": "object",
"properties": {
@@ -20334,6 +21804,22 @@
"graph_enabled": {
"type": "boolean"
},
+ "image_actions": {
+ "description": "ImageActions overrides rows of the attribute-\u003ework table for this document\n(merged per action key on top of the knowledge base's table).\nnil keeps the knowledge base's setting.",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ImageActionsConfig"
+ }
+ ]
+ },
+ "image_attrs_enabled": {
+ "description": "ImageAttrsEnabled overrides the knowledge base's attribute-observed image\npipeline switch for this one document. nil keeps the knowledge base's\nsetting.",
+ "type": "boolean"
+ },
+ "image_vector_enabled": {
+ "description": "ImageVectorEnabled overrides image vector indexing for this document.\nIt does not change the KB-wide retrieval setting or embedding model.",
+ "type": "boolean"
+ },
"parser_engine_overrides": {
"description": "ParserEngineOverrides passes key-value configuration to docreader parsers\n(e.g. pdf_force_scanned=true). Merged with workspace-level overrides in the\nparse pipeline; per-upload values take priority on conflict.",
"type": "object",
@@ -20740,6 +22226,20 @@
"MatchTypeDataAnalysis"
]
},
+ "github_com_Tencent_WeKnora_internal_types.MatchedImage": {
+ "type": "object",
+ "properties": {
+ "chunk_id": {
+ "type": "string"
+ },
+ "knowledge_base_id": {
+ "type": "string"
+ },
+ "url": {
+ "type": "string"
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_types.Membership": {
"type": "object",
"properties": {
@@ -20857,7 +22357,7 @@
}
},
"artifacts": {
- "description": "Skill-generated files produced during this assistant turn (assistant messages only).\nPopulated by ArtifactCollector after the sandbox finishes, referenced by the\nartifact download endpoint. Empty for user messages and turns without skills.",
+ "description": "Skill-generated files produced during this assistant turn (assistant messages only).\nPopulated by ArtifactCollector after the sandbox finishes, referenced by the\nartifact download endpoint. Empty for user messages and turns without skills.\n\nStored in the message_artifacts table, not on the message row: the message\nrepository loads it with every message it returns and writes it whenever it\nis non-nil on create or update. A nil slice leaves the stored rows alone.",
"type": "array",
"items": {
"$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.MessageArtifact"
@@ -20983,6 +22483,10 @@
"description": "When WeKnora persisted the blob",
"type": "string"
},
+ "deleted_at": {
+ "description": "DeletedAt marks a file the user deleted. The entry stays in the list\nrather than being removed because its position IS the download address\n(msg.Artifacts[index]); dropping it would shift every later file's index\nand hand an old link the wrong blob. Keeping it also keeps the entry in\nArtifactCollector's de-duplication set, so the next collect does not\nre-attach the very file that was deleted — its sandbox mtime has not\nmoved. Clients filter these out; the bytes are already reclaimed.",
+ "type": "string"
+ },
"file_name": {
"description": "Original filename inside the sandbox",
"type": "string"
@@ -21152,6 +22656,14 @@
"description": "Provider identifier: openai, aliyun, zhipu, generic",
"type": "string"
},
+ "spec": {
+ "description": "Spec overrides the catalog entry for this row: protocol, capability\nflags and the flat protocol compat object. Everything omitted is\ninherited from the vendor catalog (see internal/models/catalog).",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ModelSpecOverride"
+ }
+ ]
+ },
"supports_vision": {
"description": "Whether the model accepts image/multimodal input",
"type": "boolean"
@@ -21245,6 +22757,40 @@
"ModelSourceAzureOpenAI"
]
},
+ "github_com_Tencent_WeKnora_internal_types.ModelSpecOverride": {
+ "type": "object",
+ "properties": {
+ "api": {
+ "type": "string"
+ },
+ "compat": {
+ "description": "Compat is a plain object so it round-trips through both JSON (API,\ndatabase column) and YAML (builtin_models.yaml).",
+ "type": "object",
+ "additionalProperties": {}
+ },
+ "context_window": {
+ "type": "integer"
+ },
+ "input": {
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ },
+ "max_output_tokens": {
+ "type": "integer"
+ },
+ "reasoning": {
+ "type": "boolean"
+ },
+ "thinking_levels": {
+ "type": "object",
+ "additionalProperties": {
+ "type": "string"
+ }
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_types.ModelType": {
"type": "string",
"enum": [
@@ -21276,6 +22822,17 @@
"ModelTypeASR"
]
},
+ "github_com_Tencent_WeKnora_internal_types.NamedCount": {
+ "type": "object",
+ "properties": {
+ "count": {
+ "type": "integer"
+ },
+ "name": {
+ "type": "string"
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_types.OBSEngineConfig": {
"type": "object",
"properties": {
@@ -21535,12 +23092,20 @@
"type": "string"
},
"mineru_model": {
- "description": "MinerU 自建解析参数",
+ "description": "以下 backend、vLLM、公式/表格、语言参数仅对 3.x 及更早版本生效。",
"type": "string"
},
"mineru_parse_method": {
"type": "string"
},
+ "mineru_server_api_key": {
+ "description": "MinerU 自建解析参数。协议按服务端自动识别:4.0+ 走 V1 API,更早版本走 /file_parse。",
+ "type": "string"
+ },
+ "mineru_tier": {
+ "description": "4.0+:flash/basic/standard/advanced,空为服务端默认",
+ "type": "string"
+ },
"mineru_vlm_server_url": {
"description": "vLLM 服务器地址 (vlm-http-client / hybrid-http-client)",
"type": "string"
@@ -21622,6 +23187,15 @@
"PromptCacheStatusHit"
]
},
+ "github_com_Tencent_WeKnora_internal_types.ProviderMetadata": {
+ "type": "object",
+ "additionalProperties": {
+ "type": "array",
+ "items": {
+ "type": "integer"
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_types.QuestionGenerationConfig": {
"type": "object",
"properties": {
@@ -21777,6 +23351,27 @@
}
}
},
+ "github_com_Tencent_WeKnora_internal_types.RerankOptions": {
+ "type": "object",
+ "properties": {
+ "enabled": {
+ "description": "Enabled turns the rerank stage off when explicitly false. Nil means on.",
+ "type": "boolean"
+ },
+ "model_id": {
+ "description": "ModelID selects the rerank model. Empty falls back to the tenant\nRetrievalConfig, then to the first rerank model of the tenant.",
+ "type": "string"
+ },
+ "threshold": {
+ "description": "Threshold is the minimum rerank model score. Nil uses the tenant\nRetrievalConfig. A pointer because 0 and negative scores are valid.",
+ "type": "number"
+ },
+ "top_k": {
+ "description": "TopK caps the reranked results. Zero uses the endpoint's result count.",
+ "type": "integer"
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_types.Resource": {
"type": "object",
"properties": {
@@ -22237,6 +23832,14 @@
"description": "QueryText is required unless query_embedding is provided, keyword matching is disabled,\nand vector matching remains enabled.",
"type": "string"
},
+ "rerank": {
+ "description": "Rerank, when present and enabled, reranks the fused candidates before\ntruncating to MatchCount. Only the hybrid-search API reads it; internal\ncallers rerank on their own.",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.RerankOptions"
+ }
+ ]
+ },
"scope_tag_ids": {
"type": "array",
"items": {
@@ -22337,6 +23940,13 @@
"description": "MatchedContent is the actual content that was matched in vector search\nFor FAQ: this is the matched question text (standard or similar question)",
"type": "string"
},
+ "matched_images": {
+ "description": "MatchedImages survives merging, deduplication and stored reference replay.\nIt contains storage references, never image bytes.",
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.MatchedImage"
+ }
+ },
"metadata": {
"description": "Metadata",
"type": "object",
@@ -22356,6 +23966,13 @@
"description": "Seq",
"type": "integer"
},
+ "source_locators": {
+ "description": "SourceLocators point back into the original file for citation\nnavigation. Merged results carry the union of their chunks' locators.",
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.SourceLocator"
+ }
+ },
"start_at": {
"description": "Start at",
"type": "integer"
@@ -22386,6 +24003,10 @@
"description": "ForkedFromMessageID is the user or assistant message, IN THE PARENT\nSESSION, that the fork branched at. For a user point, messages strictly\nbefore it were copied here. For an assistant point, that answer is\nincluded so the branch continues after it.",
"type": "string"
},
+ "host_workspace_dir": {
+ "description": "HostWorkspaceDir is the user-selected project directory this session\noperates in, for the WeKnora Lite host sandbox only. Empty means the session\ngets an auto-allocated workspace instead.\n\nWritten once at creation and never updated: re-pointing a live session\nat a different directory would leave its transcript describing files\nthat are no longer there. Standard edition never writes it.",
+ "type": "string"
+ },
"id": {
"description": "ID",
"type": "string"
@@ -22474,6 +24095,9 @@
"model_id": {
"type": "string"
},
+ "reasoning_effort": {
+ "type": "string"
+ },
"skill_names": {
"type": "array",
"items": {
@@ -22531,6 +24155,85 @@
}
}
},
+ "github_com_Tencent_WeKnora_internal_types.SourceLocator": {
+ "type": "object",
+ "properties": {
+ "bbox": {
+ "description": "BBox is [x0, y0, x1, y1] as fractions of the page width and height,\norigin at the top-left corner of the rendered page.",
+ "type": "array",
+ "items": {
+ "type": "number"
+ }
+ },
+ "block": {
+ "description": "Block is the 1-based index among a Word file's body paragraphs and tables.",
+ "type": "integer"
+ },
+ "end": {
+ "type": "integer"
+ },
+ "end_ms": {
+ "type": "integer"
+ },
+ "mapping": {
+ "description": "Mapping records verified text-to-source alignment. Empty means legacy.",
+ "type": "string"
+ },
+ "page": {
+ "description": "Page is the 1-based PDF page.",
+ "type": "integer"
+ },
+ "partial": {
+ "description": "Partial means some of the chunk has no source mapping.",
+ "type": "boolean"
+ },
+ "quote": {
+ "description": "Quote is the complete cited overlap as parsed. Viewers match it\ninside the rendered original and against the\nanswer sentence that cites the chunk.",
+ "type": "string"
+ },
+ "row_end": {
+ "type": "integer"
+ },
+ "row_start": {
+ "type": "integer"
+ },
+ "section": {
+ "description": "Section is the 1-based EPUB spine index.",
+ "type": "integer"
+ },
+ "sheet": {
+ "description": "Sheet is the spreadsheet sheet name; RowStart/RowEnd are its 1-based\nrow numbers as shown by spreadsheet applications.",
+ "type": "string"
+ },
+ "slide": {
+ "description": "Slide is the 1-based slide number in presentation order.",
+ "type": "integer"
+ },
+ "source_hash": {
+ "description": "SourceHash binds the locator to the knowledge record’s source revision hash.",
+ "type": "string"
+ },
+ "source_id": {
+ "description": "SourceID identifies a block within the parser result, not a rendered item.",
+ "type": "string"
+ },
+ "start": {
+ "description": "Start/End are rune offsets into the original text file.",
+ "type": "integer"
+ },
+ "start_ms": {
+ "description": "StartMs/EndMs bound an audio segment.",
+ "type": "integer"
+ },
+ "title": {
+ "description": "Title names the section or slide when the source has one.",
+ "type": "string"
+ },
+ "type": {
+ "type": "string"
+ }
+ }
+ },
"github_com_Tencent_WeKnora_internal_types.StarterSuggestionConfig": {
"type": "object",
"properties": {
@@ -23411,6 +25114,14 @@
"description": "BrowserSearchInstructions customizes browser search for this user. Nil/empty uses the platform default.",
"type": "string"
},
+ "gallery": {
+ "description": "Gallery holds the user's image-gallery UI state: the search\nactivation mode and the per-attribute search-field toggles. Written\nby the gallery's mode / checkbox interactions via the preferences\nPUT; read back by GET /knowledge-bases/:id/gallery-config so the\ngallery renders the same activation state on every device. Nil = the\nuser never touched the gallery controls (mode defaults to \"all\").",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.GalleryUserPrefs"
+ }
+ ]
+ },
"last_active_tenant_id": {
"description": "LastActiveTenantID remembers the last workspace the user actively\nswitched into, so a fresh login (new device, cleared browser, new\nrefresh token) lands them back in that workspace instead of always\nbouncing to their home workspace. Written by the SPA's preferences\nPUT and by service-level SwitchTenant (including when switching\nhome, which stores the home ID). Login / RefreshToken validate that\nthe workspace still exists and the user still has an active membership\n(or CanAccessAllTenants) before honouring this preference; an\ninvalid pointer is best-effort cleared and the user falls back to\nhome. Refresh JWT claims have no tenant_id, so RefreshToken\nre-resolves from this field.\n\nnil = no preference (use user.TenantID, i.e. home)\n*0 = \"clear preference\" sentinel for the partial-update endpoint\n (UpdateUserPreferences turns this into nil). Otherwise treat\n a stored *0 the same as nil.\n*N = preferred workspace id.",
"type": "integer"
@@ -24595,6 +26306,17 @@
}
}
},
+ "internal_handler.CopyModelRequest": {
+ "type": "object",
+ "required": [
+ "display_name"
+ ],
+ "properties": {
+ "display_name": {
+ "type": "string"
+ }
+ }
+ },
"internal_handler.CreateAgentRequest": {
"type": "object",
"required": [
@@ -25160,6 +26882,26 @@
}
}
},
+ "internal_handler.ListAllUsersResponse": {
+ "type": "object",
+ "properties": {
+ "has_more": {
+ "type": "boolean"
+ },
+ "limit": {
+ "type": "integer"
+ },
+ "offset": {
+ "type": "integer"
+ },
+ "users": {
+ "type": "array",
+ "items": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.UserInfo"
+ }
+ }
+ }
+ },
"internal_handler.ListSystemAdminsResponse": {
"type": "object",
"properties": {
@@ -25222,6 +26964,9 @@
"description": "为空时按需默认为 \"remote\"",
"type": "string"
},
+ "spec": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ModelSpecOverride"
+ },
"supportsDimensionOverride": {
"type": "boolean"
}
@@ -25490,6 +27235,9 @@
"description": "为空时按需默认为 \"remote\"",
"type": "string"
},
+ "spec": {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.ModelSpecOverride"
+ },
"supportsDimensionOverride": {
"type": "boolean"
}
@@ -25722,6 +27470,44 @@
}
}
},
+ "internal_handler.SearchWikiRequest": {
+ "type": "object",
+ "required": [
+ "query"
+ ],
+ "properties": {
+ "knowledge_base_id": {
+ "type": "string"
+ },
+ "knowledge_base_ids": {
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ },
+ "limit": {
+ "type": "integer"
+ },
+ "query": {
+ "type": "string"
+ }
+ }
+ },
+ "internal_handler.SetUserStatusRequest": {
+ "type": "object",
+ "required": [
+ "is_active",
+ "user_id"
+ ],
+ "properties": {
+ "is_active": {
+ "type": "boolean"
+ },
+ "user_id": {
+ "type": "string"
+ }
+ }
+ },
"internal_handler.StorageCheckRequest": {
"type": "object",
"properties": {
@@ -25901,6 +27687,14 @@
},
"name": {
"type": "string"
+ },
+ "vlm_config": {
+ "description": "VLMConfig updates the knowledge base's multimodal (vision) config.\nOptional: nil means \"no change\", mirroring how CreateKnowledgeBase\naccepts a top-level vlm_config.",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.VLMConfig"
+ }
+ ]
}
}
},
@@ -26139,6 +27933,64 @@
}
}
},
+ "internal_handler.imageListRequestBody": {
+ "type": "object",
+ "properties": {
+ "attr_filters": {
+ "type": "object",
+ "additionalProperties": {
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ }
+ },
+ "attr_rules": {
+ "type": "object",
+ "additionalProperties": {
+ "type": "object",
+ "additionalProperties": {
+ "type": "string"
+ }
+ }
+ },
+ "is_enabled": {
+ "type": "boolean"
+ },
+ "keyword": {
+ "type": "string"
+ },
+ "knowledge_ids": {
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ },
+ "page": {
+ "description": "Page",
+ "type": "integer",
+ "minimum": 1
+ },
+ "page_size": {
+ "description": "Page size",
+ "type": "integer",
+ "maximum": 1000,
+ "minimum": 1
+ },
+ "search_in": {
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ },
+ "sort_by": {
+ "type": "string"
+ },
+ "sort_order": {
+ "type": "string"
+ }
+ }
+ },
"internal_handler.invitationLookupRequest": {
"type": "object",
"required": [
@@ -26444,6 +28296,10 @@
}
]
},
+ "reasoning_effort": {
+ "description": "ReasoningEffort overrides thinking for this request; empty inherits the agent configuration.",
+ "type": "string"
+ },
"skill_names": {
"description": "Per-request Skills selected via @mention",
"type": "array",
@@ -26478,6 +28334,10 @@
"description": "Description for the session (optional)",
"type": "string"
},
+ "project_dir": {
+ "description": "ProjectDir is an optional Lite host-sandbox binding. When set it must\nbe an absolute path already present in the user-approved ProjectDirs\nlist. Empty means the session gets an auto-allocated workspace.",
+ "type": "string"
+ },
"title": {
"description": "Title for the session (optional)",
"type": "string"
@@ -26567,12 +28427,36 @@
}
}
},
+ "internal_handler_session.RewindSessionRequest": {
+ "type": "object",
+ "required": [
+ "message_id"
+ ],
+ "properties": {
+ "message_id": {
+ "description": "MessageID is the message to rewind to. A user message deletes itself and\neverything after (the client prefills that question). An assistant\nmessage keeps itself and deletes everything after.",
+ "type": "string"
+ }
+ }
+ },
"internal_handler_session.SearchKnowledgeRequest": {
"type": "object",
"required": [
"query"
],
"properties": {
+ "disable_keywords_match": {
+ "description": "Vector recall only",
+ "type": "boolean"
+ },
+ "disable_vector_match": {
+ "description": "Keyword recall only",
+ "type": "boolean"
+ },
+ "keyword_threshold": {
+ "description": "Minimum keyword score",
+ "type": "number"
+ },
"knowledge_base_id": {
"description": "Single knowledge base ID (for backward compatibility)",
"type": "string"
@@ -26591,6 +28475,10 @@
"type": "string"
}
},
+ "match_count": {
+ "description": "Number of results to return",
+ "type": "integer"
+ },
"mentioned_items": {
"description": "Optional scoped tag mentions",
"type": "array",
@@ -26602,12 +28490,24 @@
"description": "Query text to search for",
"type": "string"
},
+ "rerank": {
+ "description": "Rerank override",
+ "allOf": [
+ {
+ "$ref": "#/definitions/github_com_Tencent_WeKnora_internal_types.RerankOptions"
+ }
+ ]
+ },
"tag_ids": {
"description": "Tag IDs for filtering within a single KB",
"type": "array",
"items": {
"type": "string"
}
+ },
+ "vector_threshold": {
+ "description": "Optional overrides of the tenant retrieval config. Omitted fields keep it.",
+ "type": "number"
}
}
},
@@ -26682,4 +28582,4 @@
"in": "header"
}
}
-}
+}
\ No newline at end of file
diff --git a/docs/swagger.yaml b/docs/swagger.yaml
index 8b9136b0..23f40490 100644
--- a/docs/swagger.yaml
+++ b/docs/swagger.yaml
@@ -1,5 +1,92 @@
basePath: /api/v1
definitions:
+ github_com_Tencent_WeKnora_internal_application_service.CatalogProviderView:
+ properties:
+ api:
+ type: string
+ id:
+ type: string
+ model_thinking_levels:
+ additionalProperties:
+ items:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_models_api.ReasoningEffort'
+ type: array
+ description: |-
+ ModelThinkingLevels lists, per chat model id, the levels offered once
+ thinking is enabled, resolved against the vendor map and protocol so
+ the console does not re-implement level inheritance.
+ type: object
+ model_types:
+ items:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.ModelType'
+ type: array
+ models:
+ items:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_models.ModelSpec'
+ type: array
+ name:
+ type: string
+ settings:
+ additionalProperties: {}
+ type: object
+ vendor_thinking_levels:
+ description: VendorThinkingLevels is what a chat model without its own map
+ offers.
+ items:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_models_api.ReasoningEffort'
+ type: array
+ type: object
+ github_com_Tencent_WeKnora_internal_application_service.CatalogRevision:
+ properties:
+ overlay:
+ type: object
+ updated_at:
+ type: string
+ updated_by:
+ type: string
+ version:
+ type: integer
+ type: object
+ github_com_Tencent_WeKnora_internal_application_service.CatalogState:
+ properties:
+ applied_version:
+ type: integer
+ baseline:
+ type: string
+ builtin:
+ items:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogProviderView'
+ type: array
+ deployment:
+ items:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogProviderView'
+ type: array
+ deployment_error:
+ type: string
+ effective:
+ items:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogProviderView'
+ type: array
+ history:
+ items:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogRevision'
+ type: array
+ overlay:
+ type: object
+ sync_error:
+ type: string
+ version:
+ type: integer
+ type: object
+ github_com_Tencent_WeKnora_internal_application_service.CatalogUpdate:
+ properties:
+ baseline:
+ type: string
+ overlay:
+ type: object
+ version:
+ type: integer
+ type: object
github_com_Tencent_WeKnora_internal_application_service.WikiLintIssue:
properties:
auto_fixable:
@@ -209,6 +296,123 @@ definitions:
tier:
$ref: '#/definitions/github_com_Tencent_WeKnora_internal_infrastructure_chunker.StrategyTier'
type: object
+ github_com_Tencent_WeKnora_internal_models.ModelCost:
+ properties:
+ cache_read:
+ type: number
+ cache_write:
+ type: number
+ currency:
+ type: string
+ input:
+ type: number
+ output:
+ type: number
+ type: object
+ github_com_Tencent_WeKnora_internal_models.ModelSpec:
+ properties:
+ aliases:
+ description: |-
+ Aliases are additional ids that resolve to this entry (dated
+ snapshots, vendor-prefixed names on gateways).
+ items:
+ type: string
+ type: array
+ api:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_models_api.API'
+ compat:
+ description: |-
+ Compat is the protocol-specific overlay for this model (flat object,
+ fields depend on API). Decoded lazily by Resolve.
+ items:
+ type: integer
+ type: array
+ context_window:
+ type: integer
+ cost:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_models.ModelCost'
+ deprecated:
+ description: Deprecated hides the entry from pickers but keeps resolution
+ working.
+ type: boolean
+ dimension:
+ description: Dimension is the embedding vector width (embedding entries only).
+ type: integer
+ id:
+ type: string
+ input:
+ description: 'Input lists accepted modalities: text, image, audio, video.'
+ items:
+ type: string
+ type: array
+ match:
+ description: |-
+ Match is a case-insensitive glob ("gpt-5*", "*deepseek-r1*") that lets
+ one entry describe a family. Exact ids and aliases win over patterns;
+ among patterns the longest literal prefix wins.
+ type: string
+ max_output_tokens:
+ type: integer
+ name:
+ type: string
+ reasoning:
+ description: Reasoning reports whether the model can emit thinking at all.
+ type: boolean
+ source:
+ description: Source is the documentation URL the facts were taken from.
+ type: string
+ thinking_levels:
+ allOf:
+ - $ref: '#/definitions/github_com_Tencent_WeKnora_internal_models_api.ThinkingLevelMap'
+ description: |-
+ ThinkingLevels maps protocol-neutral levels to vendor values; missing
+ keys inherit the vendor map. See api.ThinkingLevelMap.
+ type:
+ allOf:
+ - $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.ModelType'
+ description: |-
+ Type is the WeKnora model type this entry describes. Chat models with
+ image input are still KnowledgeQA; the UI derives VLM eligibility from
+ Input. Embedding / Rerank / ASR entries exist for the model picker.
+ type: object
+ github_com_Tencent_WeKnora_internal_models_api.API:
+ enum:
+ - openai-completions
+ - openai-responses
+ - anthropic-messages
+ - google-generative-ai
+ - ollama
+ type: string
+ x-enum-varnames:
+ - APIOpenAICompletions
+ - APIOpenAIResponses
+ - APIAnthropicMessages
+ - APIGoogleGenerativeAI
+ - APIOllama
+ github_com_Tencent_WeKnora_internal_models_api.ReasoningEffort:
+ enum:
+ - "off"
+ - auto
+ - minimal
+ - low
+ - medium
+ - high
+ - xhigh
+ - max
+ type: string
+ x-enum-varnames:
+ - ReasoningOff
+ - ReasoningAuto
+ - ReasoningMinimal
+ - ReasoningLow
+ - ReasoningMedium
+ - ReasoningHigh
+ - ReasoningXHigh
+ - ReasoningMax
+ github_com_Tencent_WeKnora_internal_models_api.ThinkingLevelMap:
+ additionalProperties:
+ type: string
+ type: object
github_com_Tencent_WeKnora_internal_models_limiter.RuntimeStat:
properties:
active:
@@ -273,6 +477,14 @@ definitions:
back on the assistant message — required by MiMo / DeepSeek V3.2+ thinking
mode, ignored by providers that don't recognize the field.
type: string
+ reasoning_metadata:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.ProviderMetadata'
+ reasoning_signature:
+ description: |-
+ ReasoningSignature / ReasoningMetadata are the provider artifacts that
+ must accompany ReasoningContent on replay (Anthropic signatures,
+ OpenAI Responses encrypted reasoning items).
+ type: string
thought:
description: LLM's reasoning/thinking (Think phase)
type: string
@@ -284,6 +496,14 @@ definitions:
items:
$ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.ToolCall'
type: array
+ truncated:
+ description: |-
+ Truncated marks the round the completion-token cap cut off. It rides in
+ the agent_steps JSON so a reloaded, shared or re-opened conversation can
+ still show that the answer stops mid-sentence by design, rather than
+ looking finished. Live streaming carries the same fact on the answer
+ event; this is what survives the round trip.
+ type: boolean
user_messages_before:
description: |-
UserMessagesBefore records consumed steer rows in delivery order, before
@@ -888,7 +1108,8 @@ definitions:
this factor
type: number
history_turns:
- description: Number of history turns to keep in context. Quick-answer only; smart-reasoning sizes history by context window
+ description: Number of history turns to keep in context. Quick-answer only;
+ smart-reasoning sizes history by context window
type: integer
image_storage_provider:
description: |-
@@ -982,6 +1203,12 @@ definitions:
QuestionSuggestions owns both the static/knowledge-backed prompts shown
before the first user turn and the contextual follow-up questions shown
after a completed assistant answer.
+ reasoning_effort:
+ description: |-
+ ReasoningEffort selects the thinking intensity (off | auto | minimal |
+ low | medium | high | xhigh | max). Empty keeps the boolean Thinking
+ semantics: true means "auto". See internal/models/api.ReasoningEffort.
+ type: string
rerank_model_id:
description: ReRank model ID for retrieval
type: string
@@ -1438,6 +1665,23 @@ definitions:
suppress_when_answer_asks_question:
type: boolean
type: object
+ github_com_Tencent_WeKnora_internal_types.GalleryUserPrefs:
+ properties:
+ mode:
+ description: |-
+ Mode is "all" (every search-eligible field active) or "custom"
+ (per-field Status governs). Empty = "all".
+ type: string
+ status:
+ additionalProperties:
+ type: string
+ description: |-
+ Status maps a namespaced attribute id ("builtin:caption") to
+ "on"/"off". Only consulted in custom mode; an unrecorded field is
+ off. Source-declared on/off does not exist — activation is always
+ the user's own record.
+ type: object
+ type: object
github_com_Tencent_WeKnora_internal_types.GraphNode:
properties:
attributes:
@@ -1460,8 +1704,67 @@ definitions:
type:
type: string
type: object
+ github_com_Tencent_WeKnora_internal_types.ImageActionsConfig:
+ properties:
+ ocr:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.ImageOCRAction'
+ type: object
+ github_com_Tencent_WeKnora_internal_types.ImageAttrCondition:
+ properties:
+ is:
+ type: string
+ prop:
+ type: string
+ type: object
+ github_com_Tencent_WeKnora_internal_types.ImageOCRAction:
+ properties:
+ "on":
+ description: |-
+ On is the list of attribute conditions that enable OCR. Matching any one
+ turns OCR on.
+ items:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.ImageAttrCondition'
+ type: array
+ on_unobserved:
+ description: |-
+ OnUnobserved is the OCR decision used when the observation does not
+ answer for one of the attributes the policy reads — the attribute is
+ absent because the model skipped the line, or because the value it wrote
+ was not one of the allowed ones. The default is true: an incomplete
+ observation keeps extracting, so a missed block of text costs one extra
+ call instead of being lost. Turn it off only with a model reliable enough
+ to answer every attribute line.
+ type: boolean
+ type: object
github_com_Tencent_WeKnora_internal_types.ImageProcessingConfig:
properties:
+ image_actions:
+ allOf:
+ - $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.ImageActionsConfig'
+ description: |-
+ ImageActions overrides the built-in attribute→work table (see
+ DefaultImageActions). A nil value keeps the conservative default: run OCR
+ for block text, data visuals, and unobserved images.
+ image_attrs_enabled:
+ description: |-
+ ImageAttrsEnabled turns the attribute-observed image pipeline on for this
+ knowledge base. It is off by default on purpose, so upgrading an
+ existing deployment never changes what happens to documents already
+ being ingested: with it off every image is described and OCR'd exactly
+ as before. When on, the describe round also observes image attributes and
+ the attribute policy decides whether OCR runs for it. A single upload can
+ override it per document through
+ KnowledgeProcessOverrides.ImageAttrsEnabled.
+ type: boolean
+ image_vector_enabled:
+ description: |-
+ ImageVectorEnabled lets a knowledge base whose embedding model takes
+ images index each image by its own vector and recall it with one. It
+ is off by default on purpose: a model that takes images is also used
+ for plain text, and turning this on costs an embedding call per image
+ and widens vector recall. Existing knowledge bases stay off after an
+ upgrade because the stored JSON has no such key.
+ type: boolean
model_id:
description: Model ID
type: string
@@ -1541,9 +1844,9 @@ definitions:
properties:
representative_user_id:
description: |-
- RepresentativeUserID identifies the user attached to the OTM row for
- display/audit. Optional: when unset, the handler picks a stable default
- (the user from the legacy UserID field, or the workspace's owner).
+ RepresentativeUserID is accepted for compatibility and ignored: a
+ direct add attaches no user of the enrolled workspace, since the
+ inviter must not choose whose details the roster shows.
type: string
role:
allOf:
@@ -1556,8 +1859,7 @@ definitions:
user_id:
description: |-
UserID is retained for backward compatibility. When set without
- TenantID, the handler resolves the user's TenantID and uses this
- user as the representative.
+ TenantID, the handler resolves the user's TenantID.
type: string
required:
- role
@@ -1667,6 +1969,13 @@ definitions:
- $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.FAQConfig'
description: FAQConfig stores FAQ specific configuration such as indexing
strategy
+ generated_profile:
+ allOf:
+ - $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfile'
+ description: |-
+ GeneratedProfile is the machine-generated description: a gist, merged
+ topics, typical questions and the aggregate snapshot they came from. It
+ never overwrites the user-authored Description; both are shown to agents.
id:
description: Unique identifier of the knowledge base
type: string
@@ -1714,6 +2023,12 @@ definitions:
description: ProcessingCount indicates the number of knowledge items being
processed (for document type knowledge bases)
type: integer
+ profile_config:
+ allOf:
+ - $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfileConfig'
+ description: |-
+ ProfileConfig controls automatic generation of the knowledge-base
+ description from per-document profiles (document knowledge bases only).
question_generation_config:
allOf:
- $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.QuestionGenerationConfig'
@@ -1787,18 +2102,124 @@ definitions:
image_processing_config:
allOf:
- $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.ImageProcessingConfig'
- description: Image processing configuration
+ description: |-
+ Image processing configuration.
+
+ nil means "no change" when updating, the same contract IndexingStrategy
+ uses. A request that does not mention this field must leave the image
+ settings alone: the attribute-observation switch and the attribute policy
+ are things a knowledge base accumulates, and a client that predates them
+ would otherwise reset the lot on any save. Sending an object — empty
+ included — replaces the whole configuration.
indexing_strategy:
allOf:
- $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.IndexingStrategy'
description: |-
IndexingStrategy controls which indexing pipelines are active.
nil means "no change" when updating (preserves existing strategy).
+ profile_config:
+ allOf:
+ - $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfileConfig'
+ description: |-
+ ProfileConfig controls optional automatic knowledge-base description
+ generation. nil means "no change" when updating.
wiki_config:
allOf:
- $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.WikiConfig'
description: Wiki configuration (only for wiki-enabled knowledge bases)
type: object
+ github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfile:
+ properties:
+ aggregate_hash:
+ description: |-
+ AggregateHash identifies the aggregate the gist was generated from. A
+ mismatch with the live aggregate means the text is stale.
+ type: string
+ error:
+ description: Error holds the last generation error when Status is failed.
+ type: string
+ generated_at:
+ description: GeneratedAt records when the text was produced.
+ type: string
+ gist:
+ description: Gist is a short statement of what the knowledge base covers.
+ type: string
+ model_id:
+ description: ModelID records which chat model produced the text.
+ type: string
+ stats:
+ allOf:
+ - $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfileStats'
+ description: Stats is the aggregate snapshot the text was generated from.
+ status:
+ description: Status is one of the KnowledgeBaseProfileStatus* values.
+ type: string
+ topics:
+ description: Topics is the merged topic list (at most KnowledgeBaseProfileMaxTopics).
+ items:
+ type: string
+ type: array
+ typical_questions:
+ description: TypicalQuestions lists questions this knowledge base can answer.
+ items:
+ type: string
+ type: array
+ type: object
+ github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfileConfig:
+ properties:
+ custom_instructions:
+ description: |-
+ CustomInstructions is appended to the stable system prompt, e.g. the
+ audience the description should address or terms to keep.
+ type: string
+ enabled:
+ type: boolean
+ model_id:
+ description: ModelID overrides the chat model; empty falls back to the KB
+ summary model.
+ type: string
+ type: object
+ github_com_Tencent_WeKnora_internal_types.KnowledgeBaseProfileStats:
+ properties:
+ doc_types:
+ description: DocTypes counts documents per model-assigned document type.
+ items:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.NamedCount'
+ type: array
+ document_count:
+ description: DocumentCount is the number of eligible documents at generation
+ time.
+ type: integer
+ earliest_at:
+ description: EarliestAt and LatestAt bound document creation times.
+ type: string
+ file_types:
+ description: FileTypes counts documents per file extension.
+ items:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.NamedCount'
+ type: array
+ folders:
+ description: Folders lists the top-level folder names present.
+ items:
+ type: string
+ type: array
+ latest_at:
+ type: string
+ profiled_count:
+ description: ProfiledCount is how many of them carried a document profile.
+ type: integer
+ raw_topics:
+ description: RawTopics counts documents per normalized topic keyword before
+ merging.
+ items:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.NamedCount'
+ type: array
+ tags:
+ description: Tags counts documents per knowledge-base tag.
+ items:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.NamedCount'
+ type: array
+ type: object
github_com_Tencent_WeKnora_internal_types.KnowledgeBaseShareResponse:
properties:
chunk_count:
@@ -1884,6 +2305,24 @@ definitions:
$ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.ExtractConfig'
graph_enabled:
type: boolean
+ image_actions:
+ allOf:
+ - $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.ImageActionsConfig'
+ description: |-
+ ImageActions overrides rows of the attribute->work table for this document
+ (merged per action key on top of the knowledge base's table).
+ nil keeps the knowledge base's setting.
+ image_attrs_enabled:
+ description: |-
+ ImageAttrsEnabled overrides the knowledge base's attribute-observed image
+ pipeline switch for this one document. nil keeps the knowledge base's
+ setting.
+ type: boolean
+ image_vector_enabled:
+ description: |-
+ ImageVectorEnabled overrides image vector indexing for this document.
+ It does not change the KB-wide retrieval setting or embedding model.
+ type: boolean
parser_engine_overrides:
additionalProperties:
type: string
@@ -2197,6 +2636,15 @@ definitions:
- MatchTypeWebSearch
- MatchTypeDirectLoad
- MatchTypeDataAnalysis
+ github_com_Tencent_WeKnora_internal_types.MatchedImage:
+ properties:
+ chunk_id:
+ type: string
+ knowledge_base_id:
+ type: string
+ url:
+ type: string
+ type: object
github_com_Tencent_WeKnora_internal_types.Membership:
properties:
role:
@@ -2325,6 +2773,10 @@ definitions:
Skill-generated files produced during this assistant turn (assistant messages only).
Populated by ArtifactCollector after the sandbox finishes, referenced by the
artifact download endpoint. Empty for user messages and turns without skills.
+
+ Stored in the message_artifacts table, not on the message row: the message
+ repository loads it with every message it returns and writes it whenever it
+ is non-nil on create or update. A nil slice leaves the stored rows alone.
items:
$ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.MessageArtifact'
type: array
@@ -2430,6 +2882,16 @@ definitions:
created_at:
description: When WeKnora persisted the blob
type: string
+ deleted_at:
+ description: |-
+ DeletedAt marks a file the user deleted. The entry stays in the list
+ rather than being removed because its position IS the download address
+ (msg.Artifacts[index]); dropping it would shift every later file's index
+ and hand an old link the wrong blob. Keeping it also keeps the entry in
+ ArtifactCollector's de-duplication set, so the next collect does not
+ re-attach the very file that was deleted — its sandbox mtime has not
+ moved. Clients filter these out; the bytes are already reclaimed.
+ type: string
file_name:
description: Original filename inside the sandbox
type: string
@@ -2566,6 +3028,13 @@ definitions:
provider:
description: 'Provider identifier: openai, aliyun, zhipu, generic'
type: string
+ spec:
+ allOf:
+ - $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.ModelSpecOverride'
+ description: |-
+ Spec overrides the catalog entry for this row: protocol, capability
+ flags and the flat protocol compat object. Everything omitted is
+ inherited from the vendor catalog (see internal/models/catalog).
supports_vision:
description: Whether the model accepts image/multimodal input
type: boolean
@@ -2652,6 +3121,31 @@ definitions:
- ModelSourceNvidia
- ModelSourceNovita
- ModelSourceAzureOpenAI
+ github_com_Tencent_WeKnora_internal_types.ModelSpecOverride:
+ properties:
+ api:
+ type: string
+ compat:
+ additionalProperties: {}
+ description: |-
+ Compat is a plain object so it round-trips through both JSON (API,
+ database column) and YAML (builtin_models.yaml).
+ type: object
+ context_window:
+ type: integer
+ input:
+ items:
+ type: string
+ type: array
+ max_output_tokens:
+ type: integer
+ reasoning:
+ type: boolean
+ thinking_levels:
+ additionalProperties:
+ type: string
+ type: object
+ type: object
github_com_Tencent_WeKnora_internal_types.ModelType:
enum:
- Embedding
@@ -2678,6 +3172,13 @@ definitions:
- ModelTypeKnowledgeQA
- ModelTypeVLLM
- ModelTypeASR
+ github_com_Tencent_WeKnora_internal_types.NamedCount:
+ properties:
+ count:
+ type: integer
+ name:
+ type: string
+ type: object
github_com_Tencent_WeKnora_internal_types.OBSEngineConfig:
properties:
access_key:
@@ -2863,10 +3364,16 @@ definitions:
mineru_language:
type: string
mineru_model:
- description: MinerU 自建解析参数
+ description: 以下 backend、vLLM、公式/表格、语言参数仅对 3.x 及更早版本生效。
type: string
mineru_parse_method:
type: string
+ mineru_server_api_key:
+ description: MinerU 自建解析参数。协议按服务端自动识别:4.0+ 走 V1 API,更早版本走 /file_parse。
+ type: string
+ mineru_tier:
+ description: 4.0+:flash/basic/standard/advanced,空为服务端默认
+ type: string
mineru_vlm_server_url:
description: vLLM 服务器地址 (vlm-http-client / hybrid-http-client)
type: string
@@ -2929,6 +3436,12 @@ definitions:
- PromptCacheStatusUnreported
- PromptCacheStatusMiss
- PromptCacheStatusHit
+ github_com_Tencent_WeKnora_internal_types.ProviderMetadata:
+ additionalProperties:
+ items:
+ type: integer
+ type: array
+ type: object
github_com_Tencent_WeKnora_internal_types.QuestionGenerationConfig:
properties:
custom_instructions:
@@ -3039,6 +3552,27 @@ definitions:
required:
- requested_role
type: object
+ github_com_Tencent_WeKnora_internal_types.RerankOptions:
+ properties:
+ enabled:
+ description: Enabled turns the rerank stage off when explicitly false. Nil
+ means on.
+ type: boolean
+ model_id:
+ description: |-
+ ModelID selects the rerank model. Empty falls back to the tenant
+ RetrievalConfig, then to the first rerank model of the tenant.
+ type: string
+ threshold:
+ description: |-
+ Threshold is the minimum rerank model score. Nil uses the tenant
+ RetrievalConfig. A pointer because 0 and negative scores are valid.
+ type: number
+ top_k:
+ description: TopK caps the reranked results. Zero uses the endpoint's result
+ count.
+ type: integer
+ type: object
github_com_Tencent_WeKnora_internal_types.Resource:
properties:
description:
@@ -3394,6 +3928,13 @@ definitions:
QueryText is required unless query_embedding is provided, keyword matching is disabled,
and vector matching remains enabled.
type: string
+ rerank:
+ allOf:
+ - $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.RerankOptions'
+ description: |-
+ Rerank, when present and enabled, reranks the fused candidates before
+ truncating to MatchCount. Only the hybrid-search API reads it; internal
+ callers rerank on their own.
scope_tag_ids:
items:
type: string
@@ -3477,6 +4018,13 @@ definitions:
MatchedContent is the actual content that was matched in vector search
For FAQ: this is the matched question text (standard or similar question)
type: string
+ matched_images:
+ description: |-
+ MatchedImages survives merging, deduplication and stored reference replay.
+ It contains storage references, never image bytes.
+ items:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.MatchedImage'
+ type: array
metadata:
additionalProperties:
type: string
@@ -3491,6 +4039,13 @@ definitions:
seq:
description: Seq
type: integer
+ source_locators:
+ description: |-
+ SourceLocators point back into the original file for citation
+ navigation. Merged results carry the union of their chunks' locators.
+ items:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.SourceLocator'
+ type: array
start_at:
description: Start at
type: integer
@@ -3516,6 +4071,16 @@ definitions:
before it were copied here. For an assistant point, that answer is
included so the branch continues after it.
type: string
+ host_workspace_dir:
+ description: |-
+ HostWorkspaceDir is the user-selected project directory this session
+ operates in, for the WeKnora Lite host sandbox only. Empty means the session
+ gets an auto-allocated workspace instead.
+
+ Written once at creation and never updated: re-pointing a live session
+ at a different directory would leave its transcript describing files
+ that are no longer there. Standard edition never writes it.
+ type: string
id:
description: ID
type: string
@@ -3599,6 +4164,8 @@ definitions:
type: array
model_id:
type: string
+ reasoning_effort:
+ type: string
skill_names:
items:
type: string
@@ -3646,6 +4213,74 @@ definitions:
template.
type: string
type: object
+ github_com_Tencent_WeKnora_internal_types.SourceLocator:
+ properties:
+ bbox:
+ description: |-
+ BBox is [x0, y0, x1, y1] as fractions of the page width and height,
+ origin at the top-left corner of the rendered page.
+ items:
+ type: number
+ type: array
+ block:
+ description: Block is the 1-based index among a Word file's body paragraphs
+ and tables.
+ type: integer
+ end:
+ type: integer
+ end_ms:
+ type: integer
+ mapping:
+ description: Mapping records verified text-to-source alignment. Empty means
+ legacy.
+ type: string
+ page:
+ description: Page is the 1-based PDF page.
+ type: integer
+ partial:
+ description: Partial means some of the chunk has no source mapping.
+ type: boolean
+ quote:
+ description: |-
+ Quote is the complete cited overlap as parsed. Viewers match it
+ inside the rendered original and against the
+ answer sentence that cites the chunk.
+ type: string
+ row_end:
+ type: integer
+ row_start:
+ type: integer
+ section:
+ description: Section is the 1-based EPUB spine index.
+ type: integer
+ sheet:
+ description: |-
+ Sheet is the spreadsheet sheet name; RowStart/RowEnd are its 1-based
+ row numbers as shown by spreadsheet applications.
+ type: string
+ slide:
+ description: Slide is the 1-based slide number in presentation order.
+ type: integer
+ source_hash:
+ description: SourceHash binds the locator to the knowledge record’s source
+ revision hash.
+ type: string
+ source_id:
+ description: SourceID identifies a block within the parser result, not a rendered
+ item.
+ type: string
+ start:
+ description: Start/End are rune offsets into the original text file.
+ type: integer
+ start_ms:
+ description: StartMs/EndMs bound an audio segment.
+ type: integer
+ title:
+ description: Title names the section or slide when the source has one.
+ type: string
+ type:
+ type: string
+ type: object
github_com_Tencent_WeKnora_internal_types.StarterSuggestionConfig:
properties:
count:
@@ -4310,6 +4945,16 @@ definitions:
description: BrowserSearchInstructions customizes browser search for this
user. Nil/empty uses the platform default.
type: string
+ gallery:
+ allOf:
+ - $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.GalleryUserPrefs'
+ description: |-
+ Gallery holds the user's image-gallery UI state: the search
+ activation mode and the per-attribute search-field toggles. Written
+ by the gallery's mode / checkbox interactions via the preferences
+ PUT; read back by GET /knowledge-bases/:id/gallery-config so the
+ gallery renders the same activation state on every device. Nil = the
+ user never touched the gallery controls (mode defaults to "all").
last_active_tenant_id:
description: |-
LastActiveTenantID remembers the last workspace the user actively
@@ -5288,6 +5933,13 @@ definitions:
required:
- source_id
type: object
+ internal_handler.CopyModelRequest:
+ properties:
+ display_name:
+ type: string
+ required:
+ - display_name
+ type: object
internal_handler.CreateAgentRequest:
properties:
avatar:
@@ -5677,6 +6329,19 @@ definitions:
required:
- llmModelId
type: object
+ internal_handler.ListAllUsersResponse:
+ properties:
+ has_more:
+ type: boolean
+ limit:
+ type: integer
+ offset:
+ type: integer
+ users:
+ items:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.UserInfo'
+ type: array
+ type: object
internal_handler.ListSystemAdminsResponse:
properties:
admins:
@@ -5725,6 +6390,8 @@ definitions:
source:
description: 为空时按需默认为 "remote"
type: string
+ spec:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.ModelSpecOverride'
supportsDimensionOverride:
type: boolean
required:
@@ -5917,6 +6584,8 @@ definitions:
source:
description: 为空时按需默认为 "remote"
type: string
+ spec:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.ModelSpecOverride'
supportsDimensionOverride:
type: boolean
required:
@@ -6086,6 +6755,31 @@ definitions:
required:
- query
type: object
+ internal_handler.SearchWikiRequest:
+ properties:
+ knowledge_base_id:
+ type: string
+ knowledge_base_ids:
+ items:
+ type: string
+ type: array
+ limit:
+ type: integer
+ query:
+ type: string
+ required:
+ - query
+ type: object
+ internal_handler.SetUserStatusRequest:
+ properties:
+ is_active:
+ type: boolean
+ user_id:
+ type: string
+ required:
+ - is_active
+ - user_id
+ type: object
internal_handler.StorageCheckRequest:
properties:
cos:
@@ -6208,6 +6902,13 @@ definitions:
type: string
name:
type: string
+ vlm_config:
+ allOf:
+ - $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.VLMConfig'
+ description: |-
+ VLMConfig updates the knowledge base's multimodal (vision) config.
+ Optional: nil means "no change", mirroring how CreateKnowledgeBase
+ accepts a top-level vlm_config.
required:
- name
type: object
@@ -6368,6 +7069,46 @@ definitions:
required:
- name
type: object
+ internal_handler.imageListRequestBody:
+ properties:
+ attr_filters:
+ additionalProperties:
+ items:
+ type: string
+ type: array
+ type: object
+ attr_rules:
+ additionalProperties:
+ additionalProperties:
+ type: string
+ type: object
+ type: object
+ is_enabled:
+ type: boolean
+ keyword:
+ type: string
+ knowledge_ids:
+ items:
+ type: string
+ type: array
+ page:
+ description: Page
+ minimum: 1
+ type: integer
+ page_size:
+ description: Page size
+ maximum: 1000
+ minimum: 1
+ type: integer
+ search_in:
+ items:
+ type: string
+ type: array
+ sort_by:
+ type: string
+ sort_order:
+ type: string
+ type: object
internal_handler.invitationLookupRequest:
properties:
token:
@@ -6597,6 +7338,10 @@ definitions:
- $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.QuestionOrigin'
description: QuestionOrigin is the knowledge source of a picked suggested
question.
+ reasoning_effort:
+ description: ReasoningEffort overrides thinking for this request; empty inherits
+ the agent configuration.
+ type: string
skill_names:
description: Per-request Skills selected via @mention
items:
@@ -6622,6 +7367,12 @@ definitions:
description:
description: Description for the session (optional)
type: string
+ project_dir:
+ description: |-
+ ProjectDir is an optional Lite host-sandbox binding. When set it must
+ be an absolute path already present in the user-approved ProjectDirs
+ list. Empty means the session gets an auto-allocated workspace.
+ type: string
title:
description: Title for the session (optional)
type: string
@@ -6688,8 +7439,28 @@ definitions:
description: '"kb", "file", "tag", "mcp", "skill"'
type: string
type: object
+ internal_handler_session.RewindSessionRequest:
+ properties:
+ message_id:
+ description: |-
+ MessageID is the message to rewind to. A user message deletes itself and
+ everything after (the client prefills that question). An assistant
+ message keeps itself and deletes everything after.
+ type: string
+ required:
+ - message_id
+ type: object
internal_handler_session.SearchKnowledgeRequest:
properties:
+ disable_keywords_match:
+ description: Vector recall only
+ type: boolean
+ disable_vector_match:
+ description: Keyword recall only
+ type: boolean
+ keyword_threshold:
+ description: Minimum keyword score
+ type: number
knowledge_base_id:
description: Single knowledge base ID (for backward compatibility)
type: string
@@ -6703,6 +7474,9 @@ definitions:
items:
type: string
type: array
+ match_count:
+ description: Number of results to return
+ type: integer
mentioned_items:
description: Optional scoped tag mentions
items:
@@ -6711,11 +7485,19 @@ definitions:
query:
description: Query text to search for
type: string
+ rerank:
+ allOf:
+ - $ref: '#/definitions/github_com_Tencent_WeKnora_internal_types.RerankOptions'
+ description: Rerank override
tag_ids:
description: Tag IDs for filtering within a single KB
items:
type: string
type: array
+ vector_threshold:
+ description: Optional overrides of the tenant retrieval config. Omitted fields
+ keep it.
+ type: number
required:
- query
type: object
@@ -7247,21 +8029,64 @@ paths:
summary: 获取智能体类型预设列表
tags:
- 智能体
- /api/v1/knowledge/{id}/spans:
- get:
- consumes:
- - application/json
- description: 返回该知识在解析流水线的 trace tree(root → stage → subspan):每段状态、耗时、input/output、错误码、langfuse_trace_id。支持
- ?attempt=N 查看历史尝试;不传则返回最新尝试。前端用于渲染时间线 + 多模态/embedding 子节点 + 一键跳转 Langfuse。
+ /artifacts:
+ delete:
+ description: 删除产物库中的一个文件。产物库一行代表一个文件的最新版本,默认连同该文件在本会话中的所有历史版本一并删除
parameters:
- - description: 知识ID
- in: path
- name: id
+ - description: 会话ID
+ in: query
+ name: session_id
required: true
type: string
- - description: 指定尝试号;省略=最新
+ - description: 消息ID
in: query
- name: attempt
+ name: message_id
+ required: true
+ type: string
+ - description: 产物在消息中的位置
+ in: query
+ name: index
+ required: true
+ type: integer
+ - description: 是否删除所有版本,默认 true
+ in: query
+ name: all_versions
+ type: boolean
+ produces:
+ - application/json
+ responses:
+ "200":
+ description: OK
+ schema:
+ additionalProperties: true
+ type: object
+ "404":
+ description: Not Found
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError'
+ security:
+ - Bearer: []
+ summary: 删除产物库中的文件
+ tags:
+ - 会话
+ get:
+ description: 跨会话列出当前用户可见会话中技能生成的文件,每个文件只返回最新版本(不含存储 URL)
+ parameters:
+ - description: 按文件名过滤
+ in: query
+ name: keyword
+ type: string
+ - description: 逗号分隔的扩展名,如 .pdf,.pptx
+ in: query
+ name: file_types
+ type: string
+ - description: 页码
+ in: query
+ name: page
+ type: integer
+ - description: 每页数量(最大 100)
+ in: query
+ name: page_size
type: integer
produces:
- application/json
@@ -7271,9 +8096,12 @@ paths:
schema:
additionalProperties: true
type: object
- summary: 获取知识文档解析的 Span 树(含历史尝试)
+ security:
+ - Bearer: []
+ - ApiKeyAuth: []
+ summary: 列出我的产物
tags:
- - 知识管理
+ - 会话
/auth/auto-setup:
post:
consumes:
@@ -7624,7 +8452,7 @@ paths:
consumes:
- application/json
description: |-
- 通过 Owner 生成的共享邀请链接 token 完成注册,绕过 invite_only 模式拦截。
+ 通过共享邀请链接 token 完成注册,仅 self_serve / invite_register 模式允许。
注册者自填邮箱(与 token 不绑定);注册成功后自动加入对应空间。
parameters:
- description: 邀请注册请求
@@ -7644,6 +8472,10 @@ paths:
description: 请求参数错误
schema:
$ref: '#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError'
+ "403":
+ description: 注册功能已禁用
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError'
"409":
description: 邮箱已注册
schema:
@@ -8509,14 +9341,14 @@ paths:
put:
consumes:
- application/json
- description: 更新指定 IM 渠道的名称、模式、知识库、凭证或启用状态
+ description: 更新指定 IM 渠道的名称、模式、回复语言、知识库、凭证或启用状态
parameters:
- description: 渠道 ID
in: path
name: id
required: true
type: string
- - description: 更新字段(name/mode/output_mode/knowledge_base_id/credentials/enabled)
+ - description: 更新字段(含 locale)
in: body
name: request
required: true
@@ -9852,7 +10684,8 @@ paths:
get:
consumes:
- application/json
- description: 在知识库中执行向量和关键词混合搜索。推荐使用 POST;GET 携带 JSON 请求体仍受支持(兼容旧客户端)。
+ description: 底层召回:向量+关键词混合检索,默认不 rerank(可用 rerank 字段开启);一般检索请用 /knowledge-search。推荐
+ POST,GET 带 JSON 体仅兼容旧客户端。
parameters:
- description: 知识库ID
in: path
@@ -9894,7 +10727,8 @@ paths:
post:
consumes:
- application/json
- description: 在知识库中执行向量和关键词混合搜索。推荐使用 POST;GET 携带 JSON 请求体仍受支持(兼容旧客户端)。
+ description: 底层召回:向量+关键词混合检索,默认不 rerank(可用 rerank 字段开启);一般检索请用 /knowledge-search。推荐
+ POST,GET 带 JSON 体仅兼容旧客户端。
parameters:
- description: 知识库ID
in: path
@@ -9933,6 +10767,109 @@ paths:
summary: 混合搜索
tags:
- 知识库
+ /knowledge-bases/{id}/images:
+ get:
+ consumes:
+ - application/json
+ description: 浏览某知识库内的全部图片资产,支持关键字搜索(caption/OCR 文本)、按图片属性筛选(attr_filters / attr_rules)与排序分页。
+ parameters:
+ - description: 知识库 ID
+ in: path
+ name: id
+ required: true
+ type: string
+ - description: 页码,默认 1
+ in: query
+ name: page
+ type: integer
+ - description: 每页数量,默认 20,最大 1000
+ in: query
+ name: page_size
+ type: integer
+ - description: 关键字,对 search_in 指定字段做不区分大小写的子串匹配
+ in: query
+ name: keyword
+ type: string
+ - description: 参与搜索的属性 ID(逗号分隔);须为 in_searchfield=true
+ in: query
+ name: search_in
+ type: string
+ - description: 排序属性 ID;须为 in_sortfield=true,默认 created_at
+ in: query
+ name: sort_by
+ type: string
+ - description: 排序方向:asc / desc,默认 desc
+ in: query
+ name: sort_order
+ type: string
+ - description: 仅包含启用状态的图片
+ in: query
+ name: is_enabled
+ type: boolean
+ - description: 来源文档 ID 列表(逗号分隔,最多 50 个),仅返回这些文档的图片
+ in: query
+ name: knowledge_ids
+ type: string
+ - description: 属性筛选 JSON,同一属性多值 OR、属性间 AND
+ in: query
+ name: attr_filters
+ type: string
+ - description: 属性逐值裁决 JSON(off 隐藏 / on 强制显示,on 优先)
+ in: query
+ name: attr_rules
+ type: string
+ produces:
+ - application/json
+ responses:
+ "200":
+ description: 图片资产分页列表
+ schema:
+ additionalProperties: true
+ type: object
+ "400":
+ description: 请求参数错误
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError'
+ security:
+ - Bearer: []
+ - ApiKeyAuth: []
+ summary: 列出知识库图片资产
+ tags:
+ - knowledge
+ post:
+ consumes:
+ - application/json
+ description: 与 GET /images 同义,筛选参数改经 JSON body 传递:knowledge_ids 不再受 URL 长度限制(上限
+ 1000),供来源文档全选使用。
+ parameters:
+ - description: 知识库 ID
+ in: path
+ name: id
+ required: true
+ type: string
+ - description: 筛选参数(与 GET /images 同名参数同义)
+ in: body
+ name: body
+ schema:
+ $ref: '#/definitions/internal_handler.imageListRequestBody'
+ produces:
+ - application/json
+ responses:
+ "200":
+ description: 图片资产分页列表
+ schema:
+ additionalProperties: true
+ type: object
+ "400":
+ description: 请求参数错误
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError'
+ security:
+ - Bearer: []
+ - ApiKeyAuth: []
+ summary: 列出知识库图片资产(body 版)
+ tags:
+ - knowledge
/knowledge-bases/{id}/knowledge:
delete:
consumes:
@@ -10021,6 +10958,14 @@ paths:
in: query
name: folder_recursive
type: boolean
+ - description: '排序字段: updated_at/created_at/file_name,默认 created_at'
+ in: query
+ name: sort_by
+ type: string
+ - description: '排序方向: asc/desc,默认 desc'
+ in: query
+ name: sort_order
+ type: string
produces:
- application/json
responses:
@@ -10369,6 +11314,33 @@ paths:
summary: 置顶/取消置顶知识库
tags:
- 知识库
+ /knowledge-bases/{id}/profile/generate:
+ post:
+ description: 基于文档画像聚合,立即重新生成知识库的 AI 描述(不覆盖手写描述)
+ parameters:
+ - description: 知识库ID
+ in: path
+ name: id
+ required: true
+ type: string
+ produces:
+ - application/json
+ responses:
+ "200":
+ description: 生成的知识库画像
+ schema:
+ additionalProperties: true
+ type: object
+ "400":
+ description: 知识库类型不支持或未配置模型
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError'
+ security:
+ - Bearer: []
+ - ApiKeyAuth: []
+ summary: 生成知识库描述
+ tags:
+ - 知识库
/knowledge-bases/{id}/shares:
get:
description: 获取知识库的所有共享记录
@@ -10764,7 +11736,8 @@ paths:
post:
consumes:
- application/json
- description: 在知识库中搜索(不使用LLM总结)
+ description: 在知识库中搜索(不使用LLM总结)。与产品内问答使用同一检索流程(召回、rerank、合并),外部检索首选;可覆盖召回参数与
+ rerank
parameters:
- description: 搜索请求
in: body
@@ -11030,6 +12003,36 @@ paths:
summary: 重新解析知识
tags:
- 知识管理
+ /knowledge/{id}/spans:
+ get:
+ consumes:
+ - application/json
+ description: 返回该知识在解析流水线的 trace tree(root → stage → subspan):每段状态、耗时、input/output、错误码、langfuse_trace_id。支持
+ ?attempt=N 查看历史尝试;不传则返回最新尝试。前端用于渲染时间线 + 多模态/embedding 子节点 + 一键跳转 Langfuse。
+ parameters:
+ - description: 知识ID
+ in: path
+ name: id
+ required: true
+ type: string
+ - description: 指定尝试号;省略=最新
+ in: query
+ name: attempt
+ type: integer
+ produces:
+ - application/json
+ responses:
+ "200":
+ description: OK
+ schema:
+ additionalProperties: true
+ type: object
+ security:
+ - Bearer: []
+ - ApiKeyAuth: []
+ summary: 获取知识文档解析的 Span 树(含历史尝试)
+ tags:
+ - 知识管理
/knowledge/batch:
get:
consumes:
@@ -12305,6 +13308,15 @@ paths:
consumes:
- application/json
description: 获取当前空间的所有MCP服务(含已保存工具目录数量)
+ parameters:
+ - description: Agent ID; needs agent_source_tenant_id
+ in: query
+ name: agent_id
+ type: string
+ - description: Shared agent source workspace
+ in: query
+ name: agent_source_tenant_id
+ type: integer
produces:
- application/json
responses:
@@ -12317,6 +13329,10 @@ paths:
description: 请求参数错误
schema:
$ref: '#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError'
+ "403":
+ description: 无权使用该共享智能体
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError'
security:
- Bearer: []
- ApiKeyAuth: []
@@ -13768,13 +14784,89 @@ paths:
summary: 更新模型
tags:
- 模型管理
+ /models/{id}/copy:
+ post:
+ consumes:
+ - application/json
+ description: 复制模型配置。name 与源模型一致,凭证从已存记录复制,不接受客户端重传。
+ parameters:
+ - description: 源模型ID
+ in: path
+ name: id
+ required: true
+ type: string
+ - description: 副本展示名
+ in: body
+ name: request
+ required: true
+ schema:
+ $ref: '#/definitions/internal_handler.CopyModelRequest'
+ produces:
+ - application/json
+ responses:
+ "201":
+ description: 复制的模型
+ schema:
+ additionalProperties: true
+ type: object
+ "400":
+ description: 请求参数错误
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError'
+ "404":
+ description: 模型不存在
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError'
+ security:
+ - Bearer: []
+ - ApiKeyAuth: []
+ summary: 复制模型
+ tags:
+ - 模型管理
+ /models/catalog/resolve:
+ get:
+ consumes:
+ - application/json
+ description: 根据厂商、模型名、Base URL 与 extra_config 返回目录解析结果(协议、思考等级、上下文等),供模型编辑器实时展示
+ parameters:
+ - description: 厂商标识
+ in: query
+ name: provider
+ required: true
+ type: string
+ - description: 模型名
+ in: query
+ name: model
+ type: string
+ - description: Base URL
+ in: query
+ name: base_url
+ type: string
+ - description: 模型类型
+ in: query
+ name: model_type
+ type: string
+ produces:
+ - application/json
+ responses:
+ "200":
+ description: 解析结果
+ schema:
+ additionalProperties: true
+ type: object
+ security:
+ - Bearer: []
+ - ApiKeyAuth: []
+ summary: 解析模型的有效接入配置
+ tags:
+ - 模型管理
/models/providers:
get:
consumes:
- application/json
- description: 根据模型类型获取支持的厂商列表及配置信息
+ description: 根据模型类型获取支持的厂商定义(含图标、默认地址、额外字段、内置模型目录与思考能力)
parameters:
- - description: 模型类型 (chat, embedding, rerank, vllm)
+ - description: 模型类型 (chat, embedding, rerank, vllm, asr)
in: query
name: model_type
type: string
@@ -15519,6 +16611,46 @@ paths:
/sessions/{session_id}/messages/{message_id}/artifacts:
get:
responses: {}
+ /sessions/{session_id}/messages/{message_id}/artifacts/{index}:
+ delete:
+ description: 删除某条 assistant 消息生成的单个文件;all_versions=true 时连同本会话中该文件的所有历史版本一并删除
+ parameters:
+ - description: 会话ID
+ in: path
+ name: session_id
+ required: true
+ type: string
+ - description: 消息ID
+ in: path
+ name: message_id
+ required: true
+ type: string
+ - description: 产物在消息中的位置
+ in: path
+ name: index
+ required: true
+ type: integer
+ - description: 是否删除该文件的所有版本
+ in: query
+ name: all_versions
+ type: boolean
+ produces:
+ - application/json
+ responses:
+ "200":
+ description: OK
+ schema:
+ additionalProperties: true
+ type: object
+ "404":
+ description: Not Found
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError'
+ security:
+ - Bearer: []
+ summary: 删除会话中的产物文件
+ tags:
+ - 会话
/sessions/{session_id}/messages/{message_id}/artifacts/{index}/download:
get:
responses: {}
@@ -15615,6 +16747,53 @@ paths:
summary: 置顶会话
tags:
- 会话
+ /sessions/{session_id}/rewind:
+ post:
+ consumes:
+ - application/json
+ description: 回滚当前会话到指定用户或助手消息:删除其后消息,可到达时 git reset 工作区。
+ parameters:
+ - description: 会话 ID
+ in: path
+ name: session_id
+ required: true
+ type: string
+ - description: 回滚请求
+ in: body
+ name: request
+ required: true
+ schema:
+ $ref: '#/definitions/internal_handler_session.RewindSessionRequest'
+ produces:
+ - application/json
+ responses:
+ "200":
+ description: 回滚结果
+ schema:
+ additionalProperties: true
+ type: object
+ "400":
+ description: 请求参数错误 / 回滚点角色不支持
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError'
+ "404":
+ description: 会话或消息不存在
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError'
+ "409":
+ description: 会话正在生成中 / 无检查点 / 沙箱已更换
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError'
+ "500":
+ description: 工作区回滚失败
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError'
+ security:
+ - Bearer: []
+ - ApiKeyAuth: []
+ summary: 回滚会话
+ tags:
+ - 会话
/sessions/{session_id}/steer:
post:
consumes:
@@ -15909,10 +17088,18 @@ paths:
- application/json
description: 返回指定沙箱配置镜像内、智能体实际能调用的已安装技能(ready 且启用)。不传 sandbox_config_id 时列表为空。
parameters:
- - description: Sandbox config ID
+ - description: Sandbox config ID; ignored for a shared agent
in: query
name: sandbox_config_id
type: string
+ - description: Agent ID; needs agent_source_tenant_id
+ in: query
+ name: agent_id
+ type: string
+ - description: Shared agent source workspace
+ in: query
+ name: agent_source_tenant_id
+ type: integer
produces:
- application/json
responses:
@@ -15921,6 +17108,10 @@ paths:
schema:
additionalProperties: true
type: object
+ "403":
+ description: 无权使用该共享智能体
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError'
security:
- Bearer: []
- ApiKeyAuth: []
@@ -16494,6 +17685,141 @@ paths:
summary: List all system administrators
tags:
- System Admin
+ /system/admin/model-catalog:
+ get:
+ description: |-
+ System administrator JWT only; API keys are not accepted. Returns built-in, deployment and effective
+ catalogs, the current overlay, version/baseline tokens, and the latest 20 published revisions.
+ produces:
+ - application/json
+ responses:
+ "200":
+ description: OK
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogState'
+ "401":
+ description: Unauthorized
+ schema:
+ additionalProperties: true
+ type: object
+ "403":
+ description: Forbidden
+ schema:
+ additionalProperties: true
+ type: object
+ "503":
+ description: Service Unavailable
+ schema:
+ additionalProperties: true
+ type: object
+ security:
+ - Bearer: []
+ summary: Get the system model catalog
+ tags:
+ - 系统管理
+ put:
+ consumes:
+ - application/json
+ description: |-
+ System administrator JWT only; API keys are not accepted. Validates and publishes a new overlay
+ revision using the current version and baseline. Stale values return 409. To restore a historical
+ configuration, submit its overlay with the current version and baseline.
+ parameters:
+ - description: Version, baseline and overlay object (max 1 MiB)
+ in: body
+ name: request
+ required: true
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogUpdate'
+ produces:
+ - application/json
+ responses:
+ "200":
+ description: OK
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogState'
+ "400":
+ description: Bad Request
+ schema:
+ additionalProperties: true
+ type: object
+ "401":
+ description: Unauthorized
+ schema:
+ additionalProperties: true
+ type: object
+ "403":
+ description: Forbidden
+ schema:
+ additionalProperties: true
+ type: object
+ "409":
+ description: Conflict
+ schema:
+ additionalProperties: true
+ type: object
+ "500":
+ description: Internal Server Error
+ schema:
+ additionalProperties: true
+ type: object
+ security:
+ - Bearer: []
+ summary: Publish a model catalog overlay
+ tags:
+ - 系统管理
+ /system/admin/model-catalog/preview:
+ post:
+ consumes:
+ - application/json
+ description: |-
+ System administrator JWT only; API keys are not accepted. Validates the overlay without publishing
+ it. Supply the current version and baseline; stale values return 409. The response contains the
+ candidate effective catalog and normalized overlay; history, builtin and deployment are null.
+ parameters:
+ - description: Version, baseline and overlay object (max 1 MiB)
+ in: body
+ name: request
+ required: true
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogUpdate'
+ produces:
+ - application/json
+ responses:
+ "200":
+ description: OK
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_application_service.CatalogState'
+ "400":
+ description: Bad Request
+ schema:
+ additionalProperties: true
+ type: object
+ "401":
+ description: Unauthorized
+ schema:
+ additionalProperties: true
+ type: object
+ "403":
+ description: Forbidden
+ schema:
+ additionalProperties: true
+ type: object
+ "409":
+ description: Conflict
+ schema:
+ additionalProperties: true
+ type: object
+ "500":
+ description: Internal Server Error
+ schema:
+ additionalProperties: true
+ type: object
+ security:
+ - Bearer: []
+ summary: Preview a model catalog overlay
+ tags:
+ - 系统管理
/system/admin/promote:
post:
consumes:
@@ -16808,6 +18134,38 @@ paths:
summary: Apply the default storage quota to every existing workspace
tags:
- System Admin
+ /system/admin/users:
+ get:
+ description: |-
+ Retrieve a paginated list of every user across tenants (SystemAdmin only).
+ Supports `offset` (default 0) and `limit` (default 50, max 200).
+ Password/hashes are never serialised — only the redacted UserInfo shape.
+ parameters:
+ - default: 0
+ description: Page offset
+ in: query
+ name: offset
+ type: integer
+ - default: 50
+ description: Page size (max 200)
+ in: query
+ name: limit
+ type: integer
+ produces:
+ - application/json
+ responses:
+ "200":
+ description: Users retrieved successfully
+ schema:
+ $ref: '#/definitions/internal_handler.ListAllUsersResponse'
+ "403":
+ description: 'Forbidden: not a system admin'
+ schema:
+ additionalProperties: true
+ type: object
+ summary: List all platform users
+ tags:
+ - System Admin
/system/admin/users/create:
post:
consumes:
@@ -16900,6 +18258,41 @@ paths:
summary: Reset another user's password
tags:
- System Admin
+ /system/admin/users/status:
+ post:
+ consumes:
+ - application/json
+ description: |-
+ Disable revokes the target's active sessions immediately.
+ Self-disable and disabling the last active system administrator are rejected.
+ parameters:
+ - description: Account status request
+ in: body
+ name: request
+ required: true
+ schema:
+ $ref: '#/definitions/internal_handler.SetUserStatusRequest'
+ produces:
+ - application/json
+ responses:
+ "200":
+ description: Status updated
+ schema:
+ additionalProperties: true
+ type: object
+ "400":
+ description: Self-disable / last-admin guard
+ schema:
+ additionalProperties: true
+ type: object
+ "404":
+ description: User not found
+ schema:
+ additionalProperties: true
+ type: object
+ summary: Enable or disable a user account
+ tags:
+ - System Admin
/system/capabilities:
get:
description: 返回当前部署版本及实际注册的后端路由所对应的功能能力;仅 supported=false 表示入口应隐藏
@@ -18381,6 +19774,37 @@ paths:
summary: 保存 WeKnoraCloud 凭证
tags:
- WeKnoraCloud
+ /wiki-search:
+ post:
+ consumes:
+ - application/json
+ description: Cross-KB wiki search using the same POSIX regex ranking as single-KB
+ wiki search
+ parameters:
+ - description: Wiki search request
+ in: body
+ name: request
+ required: true
+ schema:
+ $ref: '#/definitions/internal_handler.SearchWikiRequest'
+ produces:
+ - application/json
+ responses:
+ "200":
+ description: OK
+ schema:
+ additionalProperties: true
+ type: object
+ "400":
+ description: Bad Request
+ schema:
+ $ref: '#/definitions/github_com_Tencent_WeKnora_internal_errors.AppError'
+ security:
+ - Bearer: []
+ - ApiKeyAuth: []
+ summary: Cross-KB wiki search
+ tags:
+ - Wiki
securityDefinitions:
ApiKeyAuth:
description: API Key 认证:空间 Key 固定访问所属空间;平台 Key 调用空间接口时需同时传 X-Tenant-ID
diff --git a/docs/swagger_contract_test.go b/docs/swagger_contract_test.go
index 302d9c98..1be752cf 100644
--- a/docs/swagger_contract_test.go
+++ b/docs/swagger_contract_test.go
@@ -3,6 +3,7 @@ package docs_test
import (
"encoding/json"
"os"
+ "strings"
"testing"
docs "github.com/Tencent/WeKnora/docs"
@@ -63,6 +64,55 @@ func TestKnowledgeSearchRouteContract(t *testing.T) {
}
}
+func TestReleaseAPIContracts(t *testing.T) {
+ for _, tt := range swaggerDocuments() {
+ t.Run(tt.name, func(t *testing.T) {
+ var spec struct {
+ BasePath string `json:"basePath" yaml:"basePath"`
+ Paths map[string]map[string]any `json:"paths" yaml:"paths"`
+ Definitions map[string]struct {
+ Properties map[string]struct {
+ Type string `json:"type" yaml:"type"`
+ } `json:"properties" yaml:"properties"`
+ } `json:"definitions" yaml:"definitions"`
+ }
+ if err := tt.parse(tt.loadSpec(t), &spec); err != nil {
+ t.Fatal(err)
+ }
+ if spec.BasePath != "/api/v1" {
+ t.Fatalf("unexpected API base path %q", spec.BasePath)
+ }
+ for path := range spec.Paths {
+ if strings.HasPrefix(path, spec.BasePath+"/") {
+ t.Errorf("route %s repeats the API base path", path)
+ }
+ }
+ for path, methods := range map[string][]string{
+ "/knowledge/{id}/spans": {"get"},
+ "/knowledge-bases/{id}/images": {"get", "post"},
+ "/wiki-search": {"post"},
+ "/agents/tools": {"get"},
+ "/system/admin/users": {"get"},
+ "/system/admin/users/status": {"post"},
+ "/system/admin/model-catalog": {"get", "put"},
+ "/system/admin/model-catalog/preview": {"post"},
+ } {
+ for _, method := range methods {
+ if _, ok := spec.Paths[path][method]; !ok {
+ t.Errorf("missing %s %s", method, path)
+ }
+ }
+ }
+ for _, name := range []string{"CatalogUpdate", "CatalogState", "CatalogRevision"} {
+ definition := "github_com_Tencent_WeKnora_internal_application_service." + name
+ if spec.Definitions[definition].Properties["overlay"].Type != "object" {
+ t.Errorf("%s.overlay must be a JSON object", name)
+ }
+ }
+ })
+ }
+}
+
func TestModelDeleteUsageContract(t *testing.T) {
for _, tt := range swaggerDocuments() {
t.Run(tt.name, func(t *testing.T) {
diff --git a/frontend/package-lock.json b/frontend/package-lock.json
index 0f70a6ed..5522746b 100644
--- a/frontend/package-lock.json
+++ b/frontend/package-lock.json
@@ -1,12 +1,12 @@
{
"name": "knowledage-base",
- "version": "0.8.2",
+ "version": "0.8.3",
"lockfileVersion": 3,
"requires": true,
"packages": {
"": {
"name": "knowledage-base",
- "version": "0.8.2",
+ "version": "0.8.3",
"dependencies": {
"@microsoft/fetch-event-source": "^2.0.1",
"@novnc/novnc": "^1.7.0",
diff --git a/frontend/package.json b/frontend/package.json
index ee38c482..d0b8f196 100644
--- a/frontend/package.json
+++ b/frontend/package.json
@@ -1,6 +1,6 @@
{
"name": "knowledage-base",
- "version": "0.8.2",
+ "version": "0.8.3",
"private": true,
"type": "module",
"scripts": {
diff --git a/helm/Chart.yaml b/helm/Chart.yaml
index ea9f2c47..ebeeafb4 100644
--- a/helm/Chart.yaml
+++ b/helm/Chart.yaml
@@ -5,7 +5,7 @@ description: |
with document parsing, vector search, and LLM integration.
type: application
version: 0.1.0
-appVersion: "v0.8.2"
+appVersion: "v0.8.3"
kubeVersion: ">=1.25.0-0"
home: https://github.com/Tencent/WeKnora
icon: https://raw.githubusercontent.com/Tencent/WeKnora/main/docs/images/logo.png
diff --git a/internal/application/service/model_catalog.go b/internal/application/service/model_catalog.go
index a3d06d07..9ec5a144 100644
--- a/internal/application/service/model_catalog.go
+++ b/internal/application/service/model_catalog.go
@@ -51,7 +51,7 @@ type CatalogProviderView struct {
// CatalogRevision is one earlier published overlay.
type CatalogRevision struct {
Version uint64 `json:"version"`
- Overlay json.RawMessage `json:"overlay"`
+ Overlay json.RawMessage `json:"overlay" swaggertype:"object"`
UpdatedBy string `json:"updated_by"`
UpdatedAt time.Time `json:"updated_at"`
}
@@ -62,7 +62,7 @@ type CatalogState struct {
AppliedVersion uint64 `json:"applied_version"`
Version uint64 `json:"version"`
Baseline string `json:"baseline"`
- Overlay json.RawMessage `json:"overlay"`
+ Overlay json.RawMessage `json:"overlay" swaggertype:"object"`
History []CatalogRevision `json:"history"`
Builtin []CatalogProviderView `json:"builtin"`
Deployment []CatalogProviderView `json:"deployment"`
@@ -74,7 +74,7 @@ type CatalogState struct {
type CatalogUpdate struct {
Version uint64 `json:"version"`
Baseline string `json:"baseline"`
- Overlay json.RawMessage `json:"overlay"`
+ Overlay json.RawMessage `json:"overlay" swaggertype:"object"`
}
// ModelCatalogService persists console overlays independently of the mounted
diff --git a/internal/handler/knowledge.go b/internal/handler/knowledge.go
index 857bf3d7..9ef245b3 100644
--- a/internal/handler/knowledge.go
+++ b/internal/handler/knowledge.go
@@ -673,7 +673,9 @@ func (h *KnowledgeHandler) GetKnowledge(c *gin.Context) {
// @Param id path string true "知识ID"
// @Param attempt query int false "指定尝试号;省略=最新"
// @Success 200 {object} map[string]interface{}
-// @Router /api/v1/knowledge/{id}/spans [get]
+// @Security Bearer
+// @Security ApiKeyAuth
+// @Router /knowledge/{id}/spans [get]
//
// Always returns the canonical 5-stage timeline; missing stage rows are
// synthesized as "pending" so the frontend timeline always renders five
diff --git a/internal/handler/system_model_catalog.go b/internal/handler/system_model_catalog.go
index 58b85eb4..a2c3f2e9 100644
--- a/internal/handler/system_model_catalog.go
+++ b/internal/handler/system_model_catalog.go
@@ -10,6 +10,17 @@ import (
)
// GetModelCatalog returns the catalog layers, overlay and version history.
+// @Summary Get the system model catalog
+// @Description System administrator JWT only; API keys are not accepted. Returns built-in, deployment and effective
+// @Description catalogs, the current overlay, version/baseline tokens, and the latest 20 published revisions.
+// @Tags 系统管理
+// @Produce json
+// @Success 200 {object} service.CatalogState
+// @Failure 401 {object} map[string]interface{}
+// @Failure 403 {object} map[string]interface{}
+// @Failure 503 {object} map[string]interface{}
+// @Security Bearer
+// @Router /system/admin/model-catalog [get]
func (h *SystemHandler) GetModelCatalog(c *gin.Context) {
state, err := h.modelCatalogSvc.State(c.Request.Context())
if err != nil {
@@ -20,9 +31,41 @@ func (h *SystemHandler) GetModelCatalog(c *gin.Context) {
}
// PreviewModelCatalog validates an overlay and returns the candidate catalog.
+// @Summary Preview a model catalog overlay
+// @Description System administrator JWT only; API keys are not accepted. Validates the overlay without publishing
+// @Description it. Supply the current version and baseline; stale values return 409. The response contains the
+// @Description candidate effective catalog and normalized overlay; history, builtin and deployment are null.
+// @Tags 系统管理
+// @Accept json
+// @Produce json
+// @Param request body service.CatalogUpdate true "Version, baseline and overlay object (max 1 MiB)"
+// @Success 200 {object} service.CatalogState
+// @Failure 400 {object} map[string]interface{}
+// @Failure 401 {object} map[string]interface{}
+// @Failure 403 {object} map[string]interface{}
+// @Failure 409 {object} map[string]interface{}
+// @Failure 500 {object} map[string]interface{}
+// @Security Bearer
+// @Router /system/admin/model-catalog/preview [post]
func (h *SystemHandler) PreviewModelCatalog(c *gin.Context) { h.changeModelCatalog(c, false) }
// PublishModelCatalog validates, persists and applies an overlay.
+// @Summary Publish a model catalog overlay
+// @Description System administrator JWT only; API keys are not accepted. Validates and publishes a new overlay
+// @Description revision using the current version and baseline. Stale values return 409. To restore a historical
+// @Description configuration, submit its overlay with the current version and baseline.
+// @Tags 系统管理
+// @Accept json
+// @Produce json
+// @Param request body service.CatalogUpdate true "Version, baseline and overlay object (max 1 MiB)"
+// @Success 200 {object} service.CatalogState
+// @Failure 400 {object} map[string]interface{}
+// @Failure 401 {object} map[string]interface{}
+// @Failure 403 {object} map[string]interface{}
+// @Failure 409 {object} map[string]interface{}
+// @Failure 500 {object} map[string]interface{}
+// @Security Bearer
+// @Router /system/admin/model-catalog [put]
func (h *SystemHandler) PublishModelCatalog(c *gin.Context) { h.changeModelCatalog(c, true) }
func (h *SystemHandler) changeModelCatalog(c *gin.Context, publish bool) {
diff --git a/website-docs/.vitepress/theme/SiteHeader.vue b/website-docs/.vitepress/theme/SiteHeader.vue
index 816dae65..d6d450fd 100644
--- a/website-docs/.vitepress/theme/SiteHeader.vue
+++ b/website-docs/.vitepress/theme/SiteHeader.vue
@@ -12,7 +12,7 @@ const docsNavigation = [
{ label: 'API', href: '/docs/04-api/01-api-overview', section: '/docs/04-api/' },
{ label: '客户端', href: '/docs/05-clients/01-frontend', section: '/docs/05-clients/' },
{ label: '开发', href: '/docs/06-development/01-dev-guide', section: '/docs/06-development/' },
- { label: '版本', href: '/docs/07-releases/v0.8.2', section: '/docs/07-releases/' },
+ { label: '版本', href: '/docs/07-releases/v0.8.3', section: '/docs/07-releases/' },
]
const open = ref(false)
watch(() => route.path, () => open.value = false)
diff --git a/website-docs/01-getting-started/01-introduction.md b/website-docs/01-getting-started/01-introduction.md
index f6e99b18..46099b49 100644
--- a/website-docs/01-getting-started/01-introduction.md
+++ b/website-docs/01-getting-started/01-introduction.md
@@ -113,7 +113,7 @@ flowchart TB
- **检索**:向量 + BM25 混合检索、RRF 融合、Rerank 重排、查询改写与扩展、意图识别(greeting/chitchat/web_search 等,见 `config/prompt_templates/intent_prompts.yaml`)。
- **问答与 Agent**:流式 SSE 问答、多轮上下文压缩、引用溯源;ReAct Agent(工具:`search_knowledge`、`read_document`、`list_documents`、`wiki_search`、`data_analysis` 等)、MCP 外部工具、Agent Skills(Docker、Cube 或 E2B 沙箱执行脚本)、Web 搜索。
- **对话体验**:回答进行中追加要求、从任意历史提问分叉或原地回滚(带沙箱工作区检查点)、按会话调整思考强度,「产物」页跨会话汇总智能体生成的文件,详见[会话体验](../03-features/18-chat-experience.md)。
-- **多租户与安全**:RBAC 角色鉴权(默认开启,`WEKNORA_TENANT_ENABLE_RBAC`)、审计日志(默认保留 90 天)、邀请制注册(`auth.registration_mode=invite_only`,也可用旧变量 `DISABLE_REGISTRATION=true`)、OIDC 单点登录、SSRF 防护(可选仅白名单出站 `SSRF_DNS_WHITELIST_ONLY`)、敏感字段 AES-256 加密。界面提供简体中文、英文、日文、韩文和俄文。
+- **多租户与安全**:RBAC 角色鉴权(默认开启,`WEKNORA_TENANT_ENABLE_RBAC`)、审计日志(默认保留 90 天)、三档注册模式(开放注册、`invite_register` 仅限邀请注册、`invite_only` 禁止注册;旧变量 `DISABLE_REGISTRATION=true` 对应禁止注册)、OIDC 单点登录、SSRF 防护(可选仅白名单出站 `SSRF_DNS_WHITELIST_ONLY`)、敏感字段 AES-256 加密。界面提供简体中文、繁体中文、英文、日文、韩文和俄文。
- **可观测性**:Langfuse 全链路追踪(LLM/Embedding/Rerank/VLM/ASR 调用与 token 统计)、健康检查、Swagger API 文档(`GIN_MODE=debug` 时)。
- **生态**:REST API(`/api/v1`)+ API Key、内置 MCP Server(按空间创建端点,把 WeKnora 作为工具暴露给其他 Agent,见[MCP 集成](../03-features/08-mcp.md))、CLI(`cli/`)、微信小程序(`miniprogram/`)、浏览器插件渠道、[本机浏览器](../05-clients/09-local-browser.md)(智能体通过 Chrome/Edge 扩展操作用户浏览器)。Lite 桌面版(macOS)可在本机沙箱中执行智能体命令。
diff --git a/website-docs/01-getting-started/02-installation.md b/website-docs/01-getting-started/02-installation.md
index a05182c6..beb91ad3 100644
--- a/website-docs/01-getting-started/02-installation.md
+++ b/website-docs/01-getting-started/02-installation.md
@@ -74,7 +74,7 @@ docker compose ps # 等所有服务变成 healthy/running
> 如果数据库仍为 ParadeDB `v0.22.2-pg17`,先按 [ParadeDB 升级说明](06-paradedb-upgrade.md) 停止写入、备份、保留数据卷更换镜像并完成 `pg_search` 扩展升级,再恢复应用。仅替换镜像不会更新已有数据库的扩展 SQL;迁移 `000099` 会处理 WeKnora 库中符合条件的 `0.22.2–0.22.5`,其他数据库仍需单独检查。
```bash
-# 在 .env 中将 WEKNORA_VERSION 设为目标版本(如 0.7.0),或保持 latest
+# 在 .env 中将 WEKNORA_VERSION 设为目标版本(如 0.8.3),或保持 latest
docker compose pull
docker compose up -d
```
@@ -190,7 +190,7 @@ PLAYWRIGHT_WEBKIT_DOWNLOAD_HOST=https://your-playwright-mirror.example.com \
## 六、Helm 部署(helm/)
-`helm/Chart.yaml`:apiVersion v2,chart 名 `weknora`,appVersion 跟随版本(如 v0.8.2),要求 Kubernetes >= 1.25.0。
+`helm/Chart.yaml`:apiVersion v2,chart 名 `weknora`,appVersion 跟随版本(如 v0.8.3),要求 Kubernetes >= 1.25.0。
Chart 内包含五个组件:`app`(`wechatopenai/weknora-app`)、`frontend`(`wechatopenai/weknora-ui`)、`docreader`、`postgresql`(ParadeDB 镜像,chart 默认 `paradedb/paradedb:v0.18.9-pg17`,与 Compose 的版本不同)、`redis`(`redis:7-alpine`),并可选启用 `minio` 与 `neo4j`。
diff --git a/website-docs/01-getting-started/03-quickstart.md b/website-docs/01-getting-started/03-quickstart.md
index 73211b9d..3939074d 100644
--- a/website-docs/01-getting-started/03-quickstart.md
+++ b/website-docs/01-getting-started/03-quickstart.md
@@ -22,7 +22,7 @@
注册要求与部署差异:
- 用户名 2–50 个字符;密码为 8–32 位,至少包含字母和数字。启用复杂密码策略后,还须同时包含大小写字母和特殊字符,界面与 API 使用同一策略;
-- 团队部署可关闭公开注册,之后通过邀请链接添加成员。可设置 `DISABLE_REGISTRATION=true`(启动时把注册模式强制为 `invite_only`),或由系统管理员在「设置 → 系统」将 `auth.registration_mode` 改成 `invite_only`(立即生效,不用重启);
+- 团队部署需要通过邀请链接让新成员注册时,由系统管理员在「设置 → 系统设置」把 `auth.registration_mode` 设为 `invite_register`(立即生效)。`invite_only` / `DISABLE_REGISTRATION=true` 表示禁止密码注册,邀请也不能创建新账号,只允许已有账号登录后加入;
- 如果部署把默认空间策略设成了 `tenantless`(`auth.default_tenant_mode`),注册后**不会**自动建空间,而是被引导到 `/onboarding/workspace`,需要先自建或接受邀请加入一个空间才能继续;
- 桌面应用免注册,启动即自动创建本地账号并登录;Lite 单二进制在浏览器中访问时仍需注册和登录。
diff --git a/website-docs/01-getting-started/04-configuration.md b/website-docs/01-getting-started/04-configuration.md
index 6780be47..19487738 100644
--- a/website-docs/01-getting-started/04-configuration.md
+++ b/website-docs/01-getting-started/04-configuration.md
@@ -119,7 +119,7 @@ flowchart LR
| 段 | 结构体 | 关键字段与默认值 |
| --- | --- | --- |
-| `auth` | `AuthConfig` | `registration_mode`:`self_serve`(默认)/ `invite_only`(`DISABLE_REGISTRATION=true` 时强制);`default_tenant_mode`:`create_personal`(默认)/ `tenantless` |
+| `auth` | `AuthConfig` | `registration_mode`:`self_serve`(默认,开放注册)/ `invite_register`(仅限邀请注册)/ `invite_only`(禁止注册,`DISABLE_REGISTRATION=true` 时使用);`default_tenant_mode`:`create_personal`(默认)/ `tenantless` |
| `audit` | `AuditConfig` | `retention_days`:审计日志保留天数,段落省略时默认 90;0 禁用清理;<0 校验报错(env `WEKNORA_AUDIT_RETENTION_DAYS`) |
| `oidc_auth` | `OIDCAuthConfig` | `enable`、`issuer_url`、`jwks_uri`、`discovery_url`(缺省由 issuer 拼 `/.well-known/openid-configuration`)、`client_id`、`client_secret`、`authorization_endpoint`、`token_endpoint`、`user_info_endpoint`、`scopes`(默认 `openid profile email`)、`user_info_mapping.username`(默认 `name`)/`email`(默认 `email`);全部可用 `OIDC_AUTH_*` 环境变量覆盖 |
| `agent` | `AgentConfig` | `llm_call_timeout`:单次 LLM 调用超时秒数(默认 120,env `WEKNORA_AGENT_LLM_TIMEOUT`);`tool_approval_timeout_seconds`:MCP 工具人工审批等待(默认 600,env `WEKNORA_AGENT_TOOL_APPROVAL_TIMEOUT`) |
@@ -225,7 +225,7 @@ AWS S3 的 `S3_ACCESS_KEY` / `S3_SECRET_KEY` 可以**同时留空**,此时走
| `JWT_SECRET` | 空 | JWT 签名密钥(必填,可用 `openssl rand -hex 32` 生成)。留空或使用示例值时每次启动随机生成,重启后已登录用户需重新登录;多副本必须配置相同的值 |
| `SYSTEM_AES_KEY` | 空 | 敏感字段落盘加密的 AES-256 主密钥,**必须 32 字节**(可用 `openssl rand -hex 16` 生成);丢失则已加密数据(租户 API Key、模型 key、向量库凭证等)不可恢复,升级时沿用原值。v0.4.0 起取代 `TENANT_AES_KEY`/`CRYPTO_MASTER_KEY`/`CRYPTO_SALT` |
| `SYSTEM_SIGNING_KEY` | 空(回退到 `SYSTEM_AES_KEY`) | 嵌入会话与预签名文件链接的签名密钥(可用 `openssl rand -hex 32` 生成)。未设置时使用 `SYSTEM_AES_KEY`;两者都缺失、长度不足 16 或为示例值时无法签发签名链接与嵌入会话。更换后已签发的链接失效;多副本须使用同一个值 |
-| `DISABLE_REGISTRATION` | false | true 时强制 `registration_mode=invite_only` |
+| `DISABLE_REGISTRATION` | false | true 时启动配置使用 `invite_only`(禁止密码注册,包括邀请注册);数据库运行时设置仍优先 |
| `WEKNORA_AUTH_DEFAULT_TENANT_MODE` | create_personal | 注册后建空间策略(`create_personal` / `tenantless`) |
| `WEKNORA_TENANT_ENABLE_RBAC` | (默认 true) | 空间角色强制鉴权开关 |
| `WEKNORA_TENANT_ENABLE_CROSS_TENANT_ACCESS` | false | 跨空间访问 |
@@ -282,6 +282,13 @@ AWS S3 的 `S3_ACCESS_KEY` / `S3_SECRET_KEY` 可以**同时留空**,此时走
| 其余 `DOCREADER_PDF_*`(词距/边栏/隐藏文本/嵌入图/图表区等 20+ 项) | 见 `docker-compose.yml` docreader 段注释 | PDF 版式与抽取精调 |
| `DOCREADER_EXTERNAL_HTTP_PROXY` / `_HTTPS_PROXY` | 空 | docreader 出站抓取代理 |
+### 描述生成与出站连接
+
+| 环境变量 | 默认 | 作用 |
+| --- | --- | --- |
+| `KNOWLEDGE_BASE_PROFILE_MAX_TOKENS` | 2048 | 知识库描述生成的补全 token 预算;无效或非正值回退默认,推理模型出现 `no gist` 时可调大 |
+| `WEKNORA_OUTBOUND_DIAL_TIMEOUT` | 5s | SSRF 安全出站连接的拨号超时,使用 Go duration(如 `10s`);不是整个请求或模型生成的总超时 |
+
### Agent、Skills 与附件
| 名称 | 默认值 | 说明 |
@@ -290,7 +297,7 @@ AWS S3 的 `S3_ACCESS_KEY` / `S3_SECRET_KEY` 可以**同时留空**,此时走
| `WEKNORA_SANDBOX_DOCKER_ENABLED` | false | Docker 沙箱后端回退开关。系统管理员也可在「设置 → 系统设置 → 网络安全」打开(DB 优先,立即生效)。默认关闭,因为本机 `docker.sock` 等同宿主机 root |
| `WEKNORA_AGENT_LLM_TIMEOUT` | 120s | Agent 单次 LLM 调用超时(Go duration 或纯数字秒) |
| `WEKNORA_AGENT_TOOL_APPROVAL_TIMEOUT` / `_FAIL_OPEN` | 600s / fail-close | MCP 工具人工审批等待与失败策略 |
-| `WEKNORA_CHAT_ATTACHMENT_TTL_HOURS` / `_WAIT_TIMEOUT_SEC` / `_OCR_CONCURRENCY` / `_OCR_MAX_PAGES` | 24 / 60 / 8 / 8 | 聊天附件解析保留时长、等待超时与 OCR 并发/页数上限 |
+| `WEKNORA_CHAT_ATTACHMENT_TTL_HOURS` / `_WAIT_TIMEOUT_SEC` / `_OCR_CONCURRENCY` / `_OCR_MAX_PAGES` | 24 / 60 / 8 / 8 | 未被有效会话消息引用的附件保留时长、读取解析等待超时与 OCR 并发/页数上限;已发送附件在原会话或分叉会话仍引用时不因 TTL 清理 |
| `WEKNORA_HOUSEKEEPING_ENABLED` | 启用 | 回收卡在 processing 的脏数据 |
| `WEKNORA_DOCUMENT_PROCESS_TIMEOUT` / `WEKNORA_DOCREADER_CALL_TIMEOUT` | 2h / 30m | 文档处理任务与单次 RPC 超时 |
| `WEKNORA_PADDLEOCR_VL_TIMEOUT` | 1000s | 自建 PaddleOCR-VL HTTP 请求超时,支持正数 Go duration(如 `5400s`、`90m`);空值、无效值或非正数使用默认值。外层超时需留余量,例如本项 `90m`、DocReader `100m`、文档任务 `2h` |
diff --git a/website-docs/03-features/02-knowledge-base.md b/website-docs/03-features/02-knowledge-base.md
index 9af89e39..b2735ec7 100644
--- a/website-docs/03-features/02-knowledge-base.md
+++ b/website-docs/03-features/02-knowledge-base.md
@@ -54,6 +54,24 @@
caption="解析时间线:各阶段耗时、失败原因与卡住提示"
hint="展示文档解析时间线抽屉:阶段瀑布图(文档解析/分块/向量化/多模态/后处理)、失败时的错误卡片(阶段名、错误码、后端原因与重试按钮),以及「已 N 分钟没有进展」的提示条和停止解析按钮。" />
+## 浏览画廊 {#image-gallery}
+
+文档知识库的「画廊」页签汇总从文档提取的图片,以瀑布流持续加载。可搜索图片描述和 OCR 文字,按系统提供的图片属性、来源文档等筛选;点击图片可查看大图、描述、OCR 与来源。来源文档筛选支持多选,便于核对某批文件的解析结果。
+
+画廊展示已有图片资源,不要求开启图片向量检索。
+
+帮你找到答案,
并将知识付诸实践。知识问答、任务执行与自动 Wiki
整理知识用 Wiki。三种能力共享同一知识库。{mode.title}
{mode.tags.map(tag =>
了解 {mode.label} 智能体能操作浏览器,
知识库能接入其他 AI。智能体可以运行技能,
也能生成文件。图文检索、原文核对与知识组织
让智能体操作你自己的浏览器
智能体可以运行技能,也能生成文件。
数据源与工具集成
通过 IM、浏览器插件、MCP 或 API 查询和使用。数据源与工具集成
通过 IM、浏览器插件、MCP Server 或 API 查询和使用。私有化部署,
按团队需要管理权限。
查看操作记录和任务运行状态。私有化部署,
按团队需要管理权限。
查看操作记录和任务运行状态。部署与存储
空间与资源权限
审计与运行监控
选择适合你的使用方式。