refactor: pin MinerU 2.7.4 and retire standalone workbench

develop-refact
panyy 2026-10-27 10:54:53 +08:00
parent 8b501924f7
commit 71e4cb4254
87 changed files with 1992 additions and 10972 deletions

662
LICENSE.md 100644
View File

@ -0,0 +1,662 @@
GNU AFFERO GENERAL PUBLIC LICENSE
Version 3, 19 November 2007
Copyright (C) 2007 Free Software Foundation, Inc. <https://fsf.org/>
Everyone is permitted to copy and distribute verbatim copies
of this license document, but changing it is not allowed.
Preamble
The GNU Affero General Public License is a free, copyleft license for
software and other kinds of works, specifically designed to ensure
cooperation with the community in the case of network server software.
The licenses for most software and other practical works are designed
to take away your freedom to share and change the works. By contrast,
our General Public Licenses are intended to guarantee your freedom to
share and change all versions of a program--to make sure it remains free
software for all its users.
When we speak of free software, we are referring to freedom, not
price. Our General Public Licenses are designed to make sure that you
have the freedom to distribute copies of free software (and charge for
them if you wish), that you receive source code or can get it if you
want it, that you can change the software or use pieces of it in new
free programs, and that you know you can do these things.
Developers that use our General Public Licenses protect your rights
with two steps: (1) assert copyright on the software, and (2) offer
you this License which gives you legal permission to copy, distribute
and/or modify the software.
A secondary benefit of defending all users' freedom is that
improvements made in alternate versions of the program, if they
receive widespread use, become available for other developers to
incorporate. Many developers of free software are heartened and
encouraged by the resulting cooperation. However, in the case of
software used on network servers, this result may fail to come about.
The GNU General Public License permits making a modified version and
letting the public access it on a server without ever releasing its
source code to the public.
The GNU Affero General Public License is designed specifically to
ensure that, in such cases, the modified source code becomes available
to the community. It requires the operator of a network server to
provide the source code of the modified version running there to the
users of that server. Therefore, public use of a modified version, on
a publicly accessible server, gives the public access to the source
code of the modified version.
An older license, called the Affero General Public License and
published by Affero, was designed to accomplish similar goals. This is
a different license, not a version of the Affero GPL, but Affero has
released a new version of the Affero GPL which permits relicensing under
this license.
The precise terms and conditions for copying, distribution and
modification follow.
TERMS AND CONDITIONS
0. Definitions.
"This License" refers to version 3 of the GNU Affero General Public License.
"Copyright" also means copyright-like laws that apply to other kinds of
works, such as semiconductor masks.
"The Program" refers to any copyrightable work licensed under this
License. Each licensee is addressed as "you". "Licensees" and
"recipients" may be individuals or organizations.
To "modify" a work means to copy from or adapt all or part of the work
in a fashion requiring copyright permission, other than the making of an
exact copy. The resulting work is called a "modified version" of the
earlier work or a work "based on" the earlier work.
A "covered work" means either the unmodified Program or a work based
on the Program.
To "propagate" a work means to do anything with it that, without
permission, would make you directly or secondarily liable for
infringement under applicable copyright law, except executing it on a
computer or modifying a private copy. Propagation includes copying,
distribution (with or without modification), making available to the
public, and in some countries other activities as well.
To "convey" a work means any kind of propagation that enables other
parties to make or receive copies. Mere interaction with a user through
a computer network, with no transfer of a copy, is not conveying.
An interactive user interface displays "Appropriate Legal Notices"
to the extent that it includes a convenient and prominently visible
feature that (1) displays an appropriate copyright notice, and (2)
tells the user that there is no warranty for the work (except to the
extent that warranties are provided), that licensees may convey the
work under this License, and how to view a copy of this License. If
the interface presents a list of user commands or options, such as a
menu, a prominent item in the list meets this criterion.
1. Source Code.
The "source code" for a work means the preferred form of the work
for making modifications to it. "Object code" means any non-source
form of a work.
A "Standard Interface" means an interface that either is an official
standard defined by a recognized standards body, or, in the case of
interfaces specified for a particular programming language, one that
is widely used among developers working in that language.
The "System Libraries" of an executable work include anything, other
than the work as a whole, that (a) is included in the normal form of
packaging a Major Component, but which is not part of that Major
Component, and (b) serves only to enable use of the work with that
Major Component, or to implement a Standard Interface for which an
implementation is available to the public in source code form. A
"Major Component", in this context, means a major essential component
(kernel, window system, and so on) of the specific operating system
(if any) on which the executable work runs, or a compiler used to
produce the work, or an object code interpreter used to run it.
The "Corresponding Source" for a work in object code form means all
the source code needed to generate, install, and (for an executable
work) run the object code and to modify the work, including scripts to
control those activities. However, it does not include the work's
System Libraries, or general-purpose tools or generally available free
programs which are used unmodified in performing those activities but
which are not part of the work. For example, Corresponding Source
includes interface definition files associated with source files for
the work, and the source code for shared libraries and dynamically
linked subprograms that the work is specifically designed to require,
such as by intimate data communication or control flow between those
subprograms and other parts of the work.
The Corresponding Source need not include anything that users
can regenerate automatically from other parts of the Corresponding
Source.
The Corresponding Source for a work in source code form is that
same work.
2. Basic Permissions.
All rights granted under this License are granted for the term of
copyright on the Program, and are irrevocable provided the stated
conditions are met. This License explicitly affirms your unlimited
permission to run the unmodified Program. The output from running a
covered work is covered by this License only if the output, given its
content, constitutes a covered work. This License acknowledges your
rights of fair use or other equivalent, as provided by copyright law.
You may make, run and propagate covered works that you do not
convey, without conditions so long as your license otherwise remains
in force. You may convey covered works to others for the sole purpose
of having them make modifications exclusively for you, or provide you
with facilities for running those works, provided that you comply with
the terms of this License in conveying all material for which you do
not control copyright. Those thus making or running the covered works
for you must do so exclusively on your behalf, under your direction
and control, on terms that prohibit them from making any copies of
your copyrighted material outside their relationship with you.
Conveying under any other circumstances is permitted solely under
the conditions stated below. Sublicensing is not allowed; section 10
makes it unnecessary.
3. Protecting Users' Legal Rights From Anti-Circumvention Law.
No covered work shall be deemed part of an effective technological
measure under any applicable law fulfilling obligations under article
11 of the WIPO copyright treaty adopted on 20 December 1996, or
similar laws prohibiting or restricting circumvention of such
measures.
When you convey a covered work, you waive any legal power to forbid
circumvention of technological measures to the extent such circumvention
is effected by exercising rights under this License with respect to
the covered work, and you disclaim any intention to limit operation or
modification of the work as a means of enforcing, against the work's
users, your or third parties' legal rights to forbid circumvention of
technological measures.
4. Conveying Verbatim Copies.
You may convey verbatim copies of the Program's source code as you
receive it, in any medium, provided that you conspicuously and
appropriately publish on each copy an appropriate copyright notice;
keep intact all notices stating that this License and any
non-permissive terms added in accord with section 7 apply to the code;
keep intact all notices of the absence of any warranty; and give all
recipients a copy of this License along with the Program.
You may charge any price or no price for each copy that you convey,
and you may offer support or warranty protection for a fee.
5. Conveying Modified Source Versions.
You may convey a work based on the Program, or the modifications to
produce it from the Program, in the form of source code under the
terms of section 4, provided that you also meet all of these conditions:
a) The work must carry prominent notices stating that you modified
it, and giving a relevant date.
b) The work must carry prominent notices stating that it is
released under this License and any conditions added under section
7. This requirement modifies the requirement in section 4 to
"keep intact all notices".
c) You must license the entire work, as a whole, under this
License to anyone who comes into possession of a copy. This
License will therefore apply, along with any applicable section 7
additional terms, to the whole of the work, and all its parts,
regardless of how they are packaged. This License gives no
permission to license the work in any other way, but it does not
invalidate such permission if you have separately received it.
d) If the work has interactive user interfaces, each must display
Appropriate Legal Notices; however, if the Program has interactive
interfaces that do not display Appropriate Legal Notices, your
work need not make them do so.
A compilation of a covered work with other separate and independent
works, which are not by their nature extensions of the covered work,
and which are not combined with it such as to form a larger program,
in or on a volume of a storage or distribution medium, is called an
"aggregate" if the compilation and its resulting copyright are not
used to limit the access or legal rights of the compilation's users
beyond what the individual works permit. Inclusion of a covered work
in an aggregate does not cause this License to apply to the other
parts of the aggregate.
6. Conveying Non-Source Forms.
You may convey a covered work in object code form under the terms
of sections 4 and 5, provided that you also convey the
machine-readable Corresponding Source under the terms of this License,
in one of these ways:
a) Convey the object code in, or embodied in, a physical product
(including a physical distribution medium), accompanied by the
Corresponding Source fixed on a durable physical medium
customarily used for software interchange.
b) Convey the object code in, or embodied in, a physical product
(including a physical distribution medium), accompanied by a
written offer, valid for at least three years and valid for as
long as you offer spare parts or customer support for that product
model, to give anyone who possesses the object code either (1) a
copy of the Corresponding Source for all the software in the
product that is covered by this License, on a durable physical
medium customarily used for software interchange, for a price no
more than your reasonable cost of physically performing this
conveying of source, or (2) access to copy the
Corresponding Source from a network server at no charge.
c) Convey individual copies of the object code with a copy of the
written offer to provide the Corresponding Source. This
alternative is allowed only occasionally and noncommercially, and
only if you received the object code with such an offer, in accord
with subsection 6b.
d) Convey the object code by offering access from a designated
place (gratis or for a charge), and offer equivalent access to the
Corresponding Source in the same way through the same place at no
further charge. You need not require recipients to copy the
Corresponding Source along with the object code. If the place to
copy the object code is a network server, the Corresponding Source
may be on a different server (operated by you or a third party)
that supports equivalent copying facilities, provided you maintain
clear directions next to the object code saying where to find the
Corresponding Source. Regardless of what server hosts the
Corresponding Source, you remain obligated to ensure that it is
available for as long as needed to satisfy these requirements.
e) Convey the object code using peer-to-peer transmission, provided
you inform other peers where the object code and Corresponding
Source of the work are being offered to the general public at no
charge under subsection 6d.
A separable portion of the object code, whose source code is excluded
from the Corresponding Source as a System Library, need not be
included in conveying the object code work.
A "User Product" is either (1) a "consumer product", which means any
tangible personal property which is normally used for personal, family,
or household purposes, or (2) anything designed or sold for incorporation
into a dwelling. In determining whether a product is a consumer product,
doubtful cases shall be resolved in favor of coverage. For a particular
product received by a particular user, "normally used" refers to a
typical or common use of that class of product, regardless of the status
of the particular user or of the way in which the particular user
actually uses, or expects or is expected to use, the product. A product
is a consumer product regardless of whether the product has substantial
commercial, industrial or non-consumer uses, unless such uses represent
the only significant mode of use of the product.
"Installation Information" for a User Product means any methods,
procedures, authorization keys, or other information required to install
and execute modified versions of a covered work in that User Product from
a modified version of its Corresponding Source. The information must
suffice to ensure that the continued functioning of the modified object
code is in no case prevented or interfered with solely because
modification has been made.
If you convey an object code work under this section in, or with, or
specifically for use in, a User Product, and the conveying occurs as
part of a transaction in which the right of possession and use of the
User Product is transferred to the recipient in perpetuity or for a
fixed term (regardless of how the transaction is characterized), the
Corresponding Source conveyed under this section must be accompanied
by the Installation Information. But this requirement does not apply
if neither you nor any third party retains the ability to install
modified object code on the User Product (for example, the work has
been installed in ROM).
The requirement to provide Installation Information does not include a
requirement to continue to provide support service, warranty, or updates
for a work that has been modified or installed by the recipient, or for
the User Product in which it has been modified or installed. Access to a
network may be denied when the modification itself materially and
adversely affects the operation of the network or violates the rules and
protocols for communication across the network.
Corresponding Source conveyed, and Installation Information provided,
in accord with this section must be in a format that is publicly
documented (and with an implementation available to the public in
source code form), and must require no special password or key for
unpacking, reading or copying.
7. Additional Terms.
"Additional permissions" are terms that supplement the terms of this
License by making exceptions from one or more of its conditions.
Additional permissions that are applicable to the entire Program shall
be treated as though they were included in this License, to the extent
that they are valid under applicable law. If additional permissions
apply only to part of the Program, that part may be used separately
under those permissions, but the entire Program remains governed by
this License without regard to the additional permissions.
When you convey a copy of a covered work, you may at your option
remove any additional permissions from that copy, or from any part of
it. (Additional permissions may be written to require their own
removal in certain cases when you modify the work.) You may place
additional permissions on material, added by you to a covered work,
for which you have or can give appropriate copyright permission.
Notwithstanding any other provision of this License, for material you
add to a covered work, you may (if authorized by the copyright holders of
that material) supplement the terms of this License with terms:
a) Disclaiming warranty or limiting liability differently from the
terms of sections 15 and 16 of this License; or
b) Requiring preservation of specified reasonable legal notices or
author attributions in that material or in the Appropriate Legal
Notices displayed by works containing it; or
c) Prohibiting misrepresentation of the origin of that material, or
requiring that modified versions of such material be marked in
reasonable ways as different from the original version; or
d) Limiting the use for publicity purposes of names of licensors or
authors of the material; or
e) Declining to grant rights under trademark law for use of some
trade names, trademarks, or service marks; or
f) Requiring indemnification of licensors and authors of that
material by anyone who conveys the material (or modified versions of
it) with contractual assumptions of liability to the recipient, for
any liability that these contractual assumptions directly impose on
those licensors and authors.
All other non-permissive additional terms are considered "further
restrictions" within the meaning of section 10. If the Program as you
received it, or any part of it, contains a notice stating that it is
governed by this License along with a term that is a further
restriction, you may remove that term. If a license document contains
a further restriction but permits relicensing or conveying under this
License, you may add to a covered work material governed by the terms
of that license document, provided that the further restriction does
not survive such relicensing or conveying.
If you add terms to a covered work in accord with this section, you
must place, in the relevant source files, a statement of the
additional terms that apply to those files, or a notice indicating
where to find the applicable terms.
Additional terms, permissive or non-permissive, may be stated in the
form of a separately written license, or stated as exceptions;
the above requirements apply either way.
8. Termination.
You may not propagate or modify a covered work except as expressly
provided under this License. Any attempt otherwise to propagate or
modify it is void, and will automatically terminate your rights under
this License (including any patent licenses granted under the third
paragraph of section 11).
However, if you cease all violation of this License, then your
license from a particular copyright holder is reinstated (a)
provisionally, unless and until the copyright holder explicitly and
finally terminates your license, and (b) permanently, if the copyright
holder fails to notify you of the violation by some reasonable means
prior to 60 days after the cessation.
Moreover, your license from a particular copyright holder is
reinstated permanently if the copyright holder notifies you of the
violation by some reasonable means, this is the first time you have
received notice of violation of this License (for any work) from that
copyright holder, and you cure the violation prior to 30 days after
your receipt of the notice.
Termination of your rights under this section does not terminate the
licenses of parties who have received copies or rights from you under
this License. If your rights have been terminated and not permanently
reinstated, you do not qualify to receive new licenses for the same
material under section 10.
9. Acceptance Not Required for Having Copies.
You are not required to accept this License in order to receive or
run a copy of the Program. Ancillary propagation of a covered work
occurring solely as a consequence of using peer-to-peer transmission
to receive a copy likewise does not require acceptance. However,
nothing other than this License grants you permission to propagate or
modify any covered work. These actions infringe copyright if you do
not accept this License. Therefore, by modifying or propagating a
covered work, you indicate your acceptance of this License to do so.
10. Automatic Licensing of Downstream Recipients.
Each time you convey a covered work, the recipient automatically
receives a license from the original licensors, to run, modify and
propagate that work, subject to this License. You are not responsible
for enforcing compliance by third parties with this License.
An "entity transaction" is a transaction transferring control of an
organization, or substantially all assets of one, or subdividing an
organization, or merging organizations. If propagation of a covered
work results from an entity transaction, each party to that
transaction who receives a copy of the work also receives whatever
licenses to the work the party's predecessor in interest had or could
give under the previous paragraph, plus a right to possession of the
Corresponding Source of the work from the predecessor in interest, if
the predecessor has it or can get it with reasonable efforts.
You may not impose any further restrictions on the exercise of the
rights granted or affirmed under this License. For example, you may
not impose a license fee, royalty, or other charge for exercise of
rights granted under this License, and you may not initiate litigation
(including a cross-claim or counterclaim in a lawsuit) alleging that
any patent claim is infringed by making, using, selling, offering for
sale, or importing the Program or any portion of it.
11. Patents.
A "contributor" is a copyright holder who authorizes use under this
License of the Program or a work on which the Program is based. The
work thus licensed is called the contributor's "contributor version".
A contributor's "essential patent claims" are all patent claims
owned or controlled by the contributor, whether already acquired or
hereafter acquired, that would be infringed by some manner, permitted
by this License, of making, using, or selling its contributor version,
but do not include claims that would be infringed only as a
consequence of further modification of the contributor version. For
purposes of this definition, "control" includes the right to grant
patent sublicenses in a manner consistent with the requirements of
this License.
Each contributor grants you a non-exclusive, worldwide, royalty-free
patent license under the contributor's essential patent claims, to
make, use, sell, offer for sale, import and otherwise run, modify and
propagate the contents of its contributor version.
In the following three paragraphs, a "patent license" is any express
agreement or commitment, however denominated, not to enforce a patent
(such as an express permission to practice a patent or covenant not to
sue for patent infringement). To "grant" such a patent license to a
party means to make such an agreement or commitment not to enforce a
patent against the party.
If you convey a covered work, knowingly relying on a patent license,
and the Corresponding Source of the work is not available for anyone
to copy, free of charge and under the terms of this License, through a
publicly available network server or other readily accessible means,
then you must either (1) cause the Corresponding Source to be so
available, or (2) arrange to deprive yourself of the benefit of the
patent license for this particular work, or (3) arrange, in a manner
consistent with the requirements of this License, to extend the patent
license to downstream recipients. "Knowingly relying" means you have
actual knowledge that, but for the patent license, your conveying the
covered work in a country, or your recipient's use of the covered work
in a country, would infringe one or more identifiable patents in that
country that you have reason to believe are valid.
If, pursuant to or in connection with a single transaction or
arrangement, you convey, or propagate by procuring conveyance of, a
covered work, and grant a patent license to some of the parties
receiving the covered work authorizing them to use, propagate, modify
or convey a specific copy of the covered work, then the patent license
you grant is automatically extended to all recipients of the covered
work and works based on it.
A patent license is "discriminatory" if it does not include within
the scope of its coverage, prohibits the exercise of, or is
conditioned on the non-exercise of one or more of the rights that are
specifically granted under this License. You may not convey a covered
work if you are a party to an arrangement with a third party that is
in the business of distributing software, under which you make payment
to the third party based on the extent of your activity of conveying
the work, and under which the third party grants, to any of the
parties who would receive the covered work from you, a discriminatory
patent license (a) in connection with copies of the covered work
conveyed by you (or copies made from those copies), or (b) primarily
for and in connection with specific products or compilations that
contain the covered work, unless you entered into that arrangement,
or that patent license was granted, prior to 28 March 2007.
Nothing in this License shall be construed as excluding or limiting
any implied license or other defenses to infringement that may
otherwise be available to you under applicable patent law.
12. No Surrender of Others' Freedom.
If conditions are imposed on you (whether by court order, agreement or
otherwise) that contradict the conditions of this License, they do not
excuse you from the conditions of this License. If you cannot convey a
covered work so as to satisfy simultaneously your obligations under this
License and any other pertinent obligations, then as a consequence you may
not convey it at all. For example, if you agree to terms that obligate you
to collect a royalty for further conveying from those to whom you convey
the Program, the only way you could satisfy both those terms and this
License would be to refrain entirely from conveying the Program.
13. Remote Network Interaction; Use with the GNU General Public License.
Notwithstanding any other provision of this License, if you modify the
Program, your modified version must prominently offer all users
interacting with it remotely through a computer network (if your version
supports such interaction) an opportunity to receive the Corresponding
Source of your version by providing access to the Corresponding Source
from a network server at no charge, through some standard or customary
means of facilitating copying of software. This Corresponding Source
shall include the Corresponding Source for any work covered by version 3
of the GNU General Public License that is incorporated pursuant to the
following paragraph.
Notwithstanding any other provision of this License, you have
permission to link or combine any covered work with a work licensed
under version 3 of the GNU General Public License into a single
combined work, and to convey the resulting work. The terms of this
License will continue to apply to the part which is the covered work,
but the work with which it is combined will remain governed by version
3 of the GNU General Public License.
14. Revised Versions of this License.
The Free Software Foundation may publish revised and/or new versions of
the GNU Affero General Public License from time to time. Such new versions
will be similar in spirit to the present version, but may differ in detail to
address new problems or concerns.
Each version is given a distinguishing version number. If the
Program specifies that a certain numbered version of the GNU Affero General
Public License "or any later version" applies to it, you have the
option of following the terms and conditions either of that numbered
version or of any later version published by the Free Software
Foundation. If the Program does not specify a version number of the
GNU Affero General Public License, you may choose any version ever published
by the Free Software Foundation.
If the Program specifies that a proxy can decide which future
versions of the GNU Affero General Public License can be used, that proxy's
public statement of acceptance of a version permanently authorizes you
to choose that version for the Program.
Later license versions may give you additional or different
permissions. However, no additional obligations are imposed on any
author or copyright holder as a result of your choosing to follow a
later version.
15. Disclaimer of Warranty.
THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY
OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO,
THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM
IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF
ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
16. Limitation of Liability.
IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS
THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY
GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS),
EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
SUCH DAMAGES.
17. Interpretation of Sections 15 and 16.
If the disclaimer of warranty and limitation of liability provided
above cannot be given local legal effect according to their terms,
reviewing courts shall apply local law that most closely approximates
an absolute waiver of all civil liability in connection with the
Program, unless a warranty or assumption of liability accompanies a
copy of the Program in return for a fee.
END OF TERMS AND CONDITIONS
How to Apply These Terms to Your New Programs
If you develop a new program, and you want it to be of the greatest
possible use to the public, the best way to achieve this is to make it
free software which everyone can redistribute and change under these terms.
To do so, attach the following notices to the program. It is safest
to attach them to the start of each source file to most effectively
state the exclusion of warranty; and each file should have at least
the "copyright" line and a pointer to where the full notice is found.
<one line to give the program's name and a brief idea of what it does.>
Copyright (C) <year> <name of author>
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License as published
by the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Affero General Public License for more details.
You should have received a copy of the GNU Affero General Public License
along with this program. If not, see <https://www.gnu.org/licenses/>.
Also add information on how to contact you by electronic and paper mail.
If your software can interact with users remotely through a computer
network, you should also make sure that it provides a way for users to
get its source. For example, if your program is a web application, its
interface could display a "Source" link that leads users to an archive
of the code. There are many ways you could offer source, and different
solutions will be better for different programs; see section 13 for the
specific requirements.
You should also get your employer (if you work as a programmer) or school,
if any, to sign a "copyright disclaimer" for the program, if necessary.
For more information on this, and how to apply and follow the GNU AGPL, see
<https://www.gnu.org/licenses/>.

View File

@ -1,2 +1,23 @@
# Unis MinerU 解析服务
markitdown
本仓库维护官方 `mineru-2.7.4-released` 解析源码及部署配置;`mineru/` 与官方发布包一致。
思维导图工作台、用户授权、模型整理、文件格式分流和版本适配已迁至 MaxKB。旧 web_ui 不再部署。官方说明见 [README.mineru-upstream.md](README.mineru-upstream.md),基线哈希见 [2.7.4-baseline.json](docs/mineru-migration/2.7.4-baseline.json)。旧业务实现可从 Git 提交 `8b50192` 回看。
## 启动
在经过验证的独立虚拟环境安装本项目,准备原版 pipeline / VLM 模型及 mineru.json,设置 `MINERU_TOOLS_CONFIG_JSON` 和 `MINERU_MODEL_SOURCE=local`,运行:
```bash
python scripts/run_parser.py --host 127.0.0.1 --port 8100
```
仅启动解析服务。接口为 `/file_parse`,不包含旧 `/api`、`/v1/tasks`、智能整理或工作台静态页面。服务应仅供可信内网调用,业务鉴权在 MaxKB 完成。
`MINERU_GPU_MEMORY_UTILIZATION` 通过启动参数传给上游引擎;默认限制单个解析请求并发。模型文件不自动下载,也不在启动时执行数据库迁移。
## 镜像
CoreX 使用 `docker/china/Dockerfile.corex`;CUDA 使用 `docker/global/Dockerfile.cuda`。构建只安装当前源码,先固定厂商 Torch/torchvision/vLLM 版本,再解析其他依赖;冲突时停止构建。不得在厂商镜像中全量升级。
`scripts/build-images.sh` 只构建解析镜像;`scripts/start-mindmap.sh` 为兼容旧命令保留,实际仅增量启动 parser,不删除容器或修改业务目录权限。具体国产卡组合仍需在目标硬件验证。

View File

@ -0,0 +1,364 @@
<div align="center" xmlns="http://www.w3.org/1999/html">
<!-- logo -->
<p align="center">
<img src="https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docs/images/MinerU-logo.png" width="300px" style="vertical-align:middle;">
</p>
<!-- icon -->
[![stars](https://img.shields.io/github/stars/opendatalab/MinerU.svg)](https://github.com/opendatalab/MinerU)
[![forks](https://img.shields.io/github/forks/opendatalab/MinerU.svg)](https://github.com/opendatalab/MinerU)
[![open issues](https://img.shields.io/github/issues-raw/opendatalab/MinerU)](https://github.com/opendatalab/MinerU/issues)
[![issue resolution](https://img.shields.io/github/issues-closed-raw/opendatalab/MinerU)](https://github.com/opendatalab/MinerU/issues)
[![PyPI version](https://img.shields.io/pypi/v/mineru)](https://pypi.org/project/mineru/)
[![PyPI - Python Version](https://img.shields.io/pypi/pyversions/mineru)](https://pypi.org/project/mineru/)
[![Downloads](https://static.pepy.tech/badge/mineru)](https://pepy.tech/project/mineru)
[![Downloads](https://static.pepy.tech/badge/mineru/month)](https://pepy.tech/project/mineru)
[![OpenDataLab](https://img.shields.io/badge/webapp_on_mineru.net-blue?logo=data:image/svg+xml;base64,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&labelColor=white)](https://mineru.net/OpenSourceTools/Extractor?source=github)
[![HuggingFace](https://img.shields.io/badge/Demo_on_HuggingFace-yellow.svg?logo=data:image/png;base64,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&labelColor=white)](https://huggingface.co/spaces/opendatalab/MinerU)
[![ModelScope](https://img.shields.io/badge/Demo_on_ModelScope-purple?logo=data:image/svg+xml;base64,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&labelColor=white)](https://www.modelscope.cn/studios/OpenDataLab/MinerU)
[![Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/gist/myhloli/a3cb16570ab3cfeadf9d8f0ac91b4fca/mineru_demo.ipynb)
[![arXiv](https://img.shields.io/badge/MinerU-Technical%20Report-b31b1b.svg?logo=arXiv)](https://arxiv.org/abs/2409.18839)
[![arXiv](https://img.shields.io/badge/MinerU2.5-Technical%20Report-b31b1b.svg?logo=arXiv)](https://arxiv.org/abs/2509.22186)
[![Ask DeepWiki](https://deepwiki.com/badge.svg)](https://deepwiki.com/opendatalab/MinerU)
<a href="https://trendshift.io/repositories/11174" target="_blank"><img src="https://trendshift.io/api/badge/repositories/11174" alt="opendatalab%2FMinerU | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
<!-- language -->
[English](README.md) | [简体中文](README_zh-CN.md)
<!-- hot link -->
<p align="center">
🚀<a href="https://mineru.net/?source=github">Access MinerU Now→✅ Zero-Install Web Version ✅ Full-Featured Desktop Client ✅ Instant API Access; Skip deployment headaches – get all product formats in one click. Developers, dive in!</a>
</p>
<!-- join us -->
<p align="center">
👋 join us on <a href="https://discord.gg/Tdedn9GTXq" target="_blank">Discord</a> and <a href="https://mineru.net/community-portal/?aliasId=3c430f94" target="_blank">WeChat</a>
</p>
</div>
# Changelog
- 2026/01/30 2.7.4 Release
- Added support for domestic computing platforms IluvatarCorex and Cambricon. Currently, the officially supported domestic computing platforms include:
- [Ascend](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/Ascend/)
- [T-Head](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/THead/)
- [METAX](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/METAX/)
- [Hygon](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/Hygon/)
- [Enflame](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/Enflame/)
- [MooreThreads](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/MooreThreads/)
- [IluvatarCorex](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/IluvatarCorex/)
- [Cambricon](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/Cambricon/)
- MinerU continues to ensure compatibility with domestic hardware platforms, supporting mainstream chip architectures. With secure and reliable technology, we empower researchers, government, and enterprises to reach new heights in document digitization!
- 2026/01/23 2.7.2 Release
- Added support for domestic computing platforms Hygon, Enflame, and Moore Threads.
- Cross-page table merging optimization, improving merge success rate and merge quality.
- 2026/01/06 2.7.1 Release
- fix bug: #4300
- Updated pdfminer.six dependency version to resolve [CVE-2025-64512](https://github.com/advisories/GHSA-wf5f-4jwr-ppcp)
- Support automatic correction of input image exif orientation to improve OCR recognition accuracy #4283
- 2025/12/30 2.7.0 Release
- Simplified installation process. No need to separately install `vlm` acceleration engine dependencies. Using `uv pip install mineru[all]` during installation will install all optional backend dependencies.
- Added new `hybrid` backend, which combines the advantages of `pipeline` and `vlm` backends. Built on vlm, it integrates some capabilities of pipeline, adding extra extensibility on top of high accuracy:
- Directly extracts text from text PDFs, natively supports multi-language recognition in text PDF scenarios, and greatly reduces parsing hallucinations;
- Supports text recognition in 109 languages for scanned PDF scenarios by specifying OCR language;
- Independent inline formula recognition switch, which can be disabled separately when inline formula recognition is not needed, improving the visual effect of parsing results.
- Simplified engine selection logic for `vlm/hybrid` backends. Users only need to specify the backend as `*-auto-engine`, and the system will automatically select the appropriate engine for inference acceleration based on the current environment, improving usability.
- Switched default parsing backend from `pipeline` to `hybrid-auto-engine`, improving out-of-the-box result consistency for new users and avoiding cognitive differences in parsing results.
- Added i18n support to gradio application, supporting switching between Chinese and English languages.
> 📝 View the complete [Changelog](https://opendatalab.github.io/MinerU/reference/changelog/) for more historical version information
# MinerU
## Project Introduction
MinerU is a tool that converts PDFs into machine-readable formats (e.g., markdown, JSON), allowing for easy extraction into any format.
MinerU was born during the pre-training process of [InternLM](https://github.com/InternLM/InternLM). We focus on solving symbol conversion issues in scientific literature and hope to contribute to technological development in the era of large models.
Compared to well-known commercial products, MinerU is still young. If you encounter any issues or if the results are not as expected, please submit an issue on [issue](https://github.com/opendatalab/MinerU/issues) and **attach the relevant PDF**.
https://github.com/user-attachments/assets/4bea02c9-6d54-4cd6-97ed-dff14340982c
## Key Features
- Remove headers, footers, footnotes, page numbers, etc., to ensure semantic coherence.
- Output text in human-readable order, suitable for single-column, multi-column, and complex layouts.
- Preserve the structure of the original document, including headings, paragraphs, lists, etc.
- Extract images, image descriptions, tables, table titles, and footnotes.
- Automatically recognize and convert formulas in the document to LaTeX format.
- Automatically recognize and convert tables in the document to HTML format.
- Automatically detect scanned PDFs and garbled PDFs and enable OCR functionality.
- OCR supports detection and recognition of 109 languages.
- Supports multiple output formats, such as multimodal and NLP Markdown, JSON sorted by reading order, and rich intermediate formats.
- Supports various visualization results, including layout visualization and span visualization, for efficient confirmation of output quality.
- Supports running in a pure CPU environment, and also supports GPU(CUDA)/NPU(CANN)/MPS acceleration
- Compatible with Windows, Linux, and Mac platforms.
# Quick Start
If you encounter any installation issues, please first consult the <a href="#faq">FAQ</a>. </br>
If the parsing results are not as expected, refer to the <a href="#known-issues">Known Issues</a>. </br>
## Online Experience
### Official online web application
The official online version has the same functionality as the client, with a beautiful interface and rich features, requires login to use
- [![OpenDataLab](https://img.shields.io/badge/webapp_on_mineru.net-blue?logo=data:image/svg+xml;base64,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&labelColor=white)](https://mineru.net/OpenSourceTools/Extractor?source=github)
### Gradio-based online demo
A WebUI developed based on Gradio, with a simple interface and only core parsing functionality, no login required
- [![ModelScope](https://img.shields.io/badge/Demo_on_ModelScope-purple?logo=data:image/svg+xml;base64,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&labelColor=white)](https://www.modelscope.cn/studios/OpenDataLab/MinerU)
- [![HuggingFace](https://img.shields.io/badge/Demo_on_HuggingFace-yellow.svg?logo=data:image/png;base64,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&labelColor=white)](https://huggingface.co/spaces/opendatalab/MinerU)
## Local Deployment
> [!WARNING]
> **Pre-installation Notice—Hardware and Software Environment Support**
>
> To ensure the stability and reliability of the project, we only optimize and test for specific hardware and software environments during development. This ensures that users deploying and running the project on recommended system configurations will get the best performance with the fewest compatibility issues.
>
> By focusing resources on the mainline environment, our team can more efficiently resolve potential bugs and develop new features.
>
> In non-mainline environments, due to the diversity of hardware and software configurations, as well as third-party dependency compatibility issues, we cannot guarantee 100% project availability. Therefore, for users who wish to use this project in non-recommended environments, we suggest carefully reading the documentation and FAQ first. Most issues already have corresponding solutions in the FAQ. We also encourage community feedback to help us gradually expand support.
<table>
<thead>
<tr>
<th rowspan="2">Parsing Backend</th>
<th rowspan="2">pipeline</th>
<th colspan="2">*-auto-engine</th>
<th colspan="2">*-http-client</th>
</tr>
<tr>
<th>hybrid</th>
<th>vlm</th>
<th>hybrid</th>
<th>vlm</th>
</tr>
</thead>
<tbody>
<tr>
<th>Backend Features</th>
<td >Good Compatibility</td>
<td colspan="2">High Hardware Requirements</td>
<td colspan="2">For OpenAI Compatible Servers<sup>2</sup></td>
</tr>
<tr>
<th>Accuracy<sup>1</sup></th>
<td style="text-align:center;">82+</td>
<td colspan="4" style="text-align:center;">90+</td>
</tr>
<tr>
<th>Operating System</th>
<td colspan="5" style="text-align:center;">Linux<sup>3</sup> / Windows<sup>4</sup> / macOS<sup>5</sup></td>
</tr>
<tr>
<th>Pure CPU Support</th>
<td style="text-align:center;">✅</td>
<td colspan="2" style="text-align:center;">❌</td>
<td colspan="2" style="text-align:center;">✅</td>
</tr>
<tr>
<th>GPU Acceleration</th>
<td colspan="4" style="text-align:center;">Volta and later architecture GPUs or Apple Silicon</td>
<td rowspan="2">Not Required</td>
</tr>
<tr>
<th>Min VRAM</th>
<td style="text-align:center;">6GB</td>
<td style="text-align:center;">10GB</td>
<td style="text-align:center;">8GB</td>
<td style="text-align:center;">3GB</td>
</tr>
<tr>
<th>RAM</th>
<td colspan="3" style="text-align:center;">Min 16GB, Recommended 32GB or more</td>
<td colspan="2" style="text-align:center;">Min 8GB</td>
</tr>
<tr>
<th>Disk Space</th>
<td colspan="3" style="text-align:center;">Min 20GB, SSD Recommended</td>
<td colspan="2" style="text-align:center;">Min 2GB</td>
</tr>
<tr>
<th>Python Version</th>
<td colspan="5" style="text-align:center;">3.10-3.13</td>
</tr>
</tbody>
</table>
<sup>1</sup> Accuracy metrics are the End-to-End Evaluation Overall scores from OmniDocBench (v1.5), based on the latest version of `MinerU`.
<sup>2</sup> Servers compatible with OpenAI API, such as local model servers or remote model services deployed via inference frameworks like `vLLM`/`SGLang`/`LMDeploy`.
<sup>3</sup> Linux only supports distributions from 2019 and later.
<sup>4</sup> Since the key dependency `ray` does not support Python 3.13 on Windows, only versions 3.10~3.12 are supported.
<sup>5</sup> macOS requires version 14.0 or later.
### Install MinerU
#### Install MinerU using pip or uv
```bash
pip install --upgrade pip
pip install uv
uv pip install -U "mineru[all]"
```
#### Install MinerU from source code
```bash
git clone https://github.com/opendatalab/MinerU.git
cd MinerU
uv pip install -e .[all]
```
> [!TIP]
> `mineru[all]` includes all core features, compatible with Windows / Linux / macOS systems, suitable for most users.
> If you need to specify the inference framework for the VLM model, or only intend to install a lightweight client on an edge device, please refer to the documentation [Extension Modules Installation Guide](https://opendatalab.github.io/MinerU/quick_start/extension_modules/).
---
#### Deploy MinerU using Docker
MinerU provides a convenient Docker deployment method, which helps quickly set up the environment and solve some tricky environment compatibility issues.
You can get the [Docker Deployment Instructions](https://opendatalab.github.io/MinerU/quick_start/docker_deployment/) in the documentation.
---
### Using MinerU
If your device meets the GPU acceleration requirements in the table above, you can use a simple command line for document parsing:
```bash
mineru -p <input_path> -o <output_path>
```
If your device does not meet the GPU acceleration requirements, you can specify the backend as `pipeline` to run in a pure CPU environment:
```bash
mineru -p <input_path> -o <output_path> -b pipeline
```
You can use MinerU for PDF parsing through various methods such as command line, API, and WebUI. For detailed instructions, please refer to the [Usage Guide](https://opendatalab.github.io/MinerU/usage/).
# TODO
- [x] Reading order based on the model
- [x] Recognition of `index` and `list` in the main text
- [x] Table recognition
- [x] Heading Classification
- [x] Handwritten Text Recognition
- [x] Vertical Text Recognition
- [x] Latin Accent Mark Recognition
- [x] Code block recognition in the main text
- [x] [Chemical formula recognition](docs/chemical_knowledge_introduction/introduction.pdf)(mineru.net)
- [ ] Geometric shape recognition
# Known Issues
- Reading order is determined by the model based on the spatial distribution of readable content, and may be out of order in some areas under extremely complex layouts.
- Limited support for vertical text.
- Tables of contents and lists are recognized through rules, and some uncommon list formats may not be recognized.
- Code blocks are not yet supported in the layout model.
- Comic books, art albums, primary school textbooks, and exercises cannot be parsed well.
- Table recognition may result in row/column recognition errors in complex tables.
- OCR recognition may produce inaccurate characters in PDFs of lesser-known languages (e.g., diacritical marks in Latin script, easily confused characters in Arabic script).
- Some formulas may not render correctly in Markdown.
# FAQ
- If you encounter any issues during usage, you can first check the [FAQ](https://opendatalab.github.io/MinerU/faq/) for solutions.
- If your issue remains unresolved, you may also use [DeepWiki](https://deepwiki.com/opendatalab/MinerU) to interact with an AI assistant, which can address most common problems.
- If you still cannot resolve the issue, you are welcome to join our community via [Discord](https://discord.gg/Tdedn9GTXq) or [WeChat](https://mineru.net/community-portal/?aliasId=3c430f94) to discuss with other users and developers.
# All Thanks To Our Contributors
<a href="https://github.com/opendatalab/MinerU/graphs/contributors">
<img src="https://contrib.rocks/image?repo=opendatalab/MinerU" />
</a>
# License Information
[LICENSE.md](LICENSE.md)
Currently, some models in this project are trained based on YOLO. However, since YOLO follows the AGPL license, it may impose restrictions on certain use cases. In future iterations, we plan to explore and replace these with models under more permissive licenses to enhance user-friendliness and flexibility.
# Acknowledgments
- [PDF-Extract-Kit](https://github.com/opendatalab/PDF-Extract-Kit)
- [DocLayout-YOLO](https://github.com/opendatalab/DocLayout-YOLO)
- [UniMERNet](https://github.com/opendatalab/UniMERNet)
- [RapidTable](https://github.com/RapidAI/RapidTable)
- [TableStructureRec](https://github.com/RapidAI/TableStructureRec)
- [PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR)
- [PaddleOCR2Pytorch](https://github.com/frotms/PaddleOCR2Pytorch)
- [layoutreader](https://github.com/ppaanngggg/layoutreader)
- [xy-cut](https://github.com/Sanster/xy-cut)
- [fast-langdetect](https://github.com/LlmKira/fast-langdetect)
- [pypdfium2](https://github.com/pypdfium2-team/pypdfium2)
- [pdftext](https://github.com/datalab-to/pdftext)
- [pdfminer.six](https://github.com/pdfminer/pdfminer.six)
- [pypdf](https://github.com/py-pdf/pypdf)
- [magika](https://github.com/google/magika)
- [vLLM](https://github.com/vllm-project/vllm)
- [LMDeploy](https://github.com/InternLM/lmdeploy)
# Citation
```bibtex
@misc{niu2025mineru25decoupledvisionlanguagemodel,
title={MinerU2.5: A Decoupled Vision-Language Model for Efficient High-Resolution Document Parsing},
author={Junbo Niu and Zheng Liu and Zhuangcheng Gu and Bin Wang and Linke Ouyang and Zhiyuan Zhao and Tao Chu and Tianyao He and Fan Wu and Qintong Zhang and Zhenjiang Jin and Guang Liang and Rui Zhang and Wenzheng Zhang and Yuan Qu and Zhifei Ren and Yuefeng Sun and Yuanhong Zheng and Dongsheng Ma and Zirui Tang and Boyu Niu and Ziyang Miao and Hejun Dong and Siyi Qian and Junyuan Zhang and Jingzhou Chen and Fangdong Wang and Xiaomeng Zhao and Liqun Wei and Wei Li and Shasha Wang and Ruiliang Xu and Yuanyuan Cao and Lu Chen and Qianqian Wu and Huaiyu Gu and Lindong Lu and Keming Wang and Dechen Lin and Guanlin Shen and Xuanhe Zhou and Linfeng Zhang and Yuhang Zang and Xiaoyi Dong and Jiaqi Wang and Bo Zhang and Lei Bai and Pei Chu and Weijia Li and Jiang Wu and Lijun Wu and Zhenxiang Li and Guangyu Wang and Zhongying Tu and Chao Xu and Kai Chen and Yu Qiao and Bowen Zhou and Dahua Lin and Wentao Zhang and Conghui He},
year={2025},
eprint={2509.22186},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2509.22186},
}
@misc{wang2024mineruopensourcesolutionprecise,
title={MinerU: An Open-Source Solution for Precise Document Content Extraction},
author={Bin Wang and Chao Xu and Xiaomeng Zhao and Linke Ouyang and Fan Wu and Zhiyuan Zhao and Rui Xu and Kaiwen Liu and Yuan Qu and Fukai Shang and Bo Zhang and Liqun Wei and Zhihao Sui and Wei Li and Botian Shi and Yu Qiao and Dahua Lin and Conghui He},
year={2024},
eprint={2409.18839},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2409.18839},
}
@article{he2024opendatalab,
title={Opendatalab: Empowering general artificial intelligence with open datasets},
author={He, Conghui and Li, Wei and Jin, Zhenjiang and Xu, Chao and Wang, Bin and Lin, Dahua},
journal={arXiv preprint arXiv:2407.13773},
year={2024}
}
```
# Star History
<a>
<picture>
<source media="(prefers-color-scheme: dark)" srcset="https://api.star-history.com/svg?repos=opendatalab/MinerU&type=Date&theme=dark" />
<source media="(prefers-color-scheme: light)" srcset="https://api.star-history.com/svg?repos=opendatalab/MinerU&type=Date" />
<img alt="Star History Chart" src="https://api.star-history.com/svg?repos=opendatalab/MinerU&type=Date" />
</picture>
</a>
# Links
- [Easy Data Preparation with latest LLMs-based Operators and Pipelines](https://github.com/OpenDCAI/DataFlow)
- [Vis3 (OSS browser based on s3)](https://github.com/opendatalab/Vis3)
- [LabelU (A Lightweight Multi-modal Data Annotation Tool)](https://github.com/opendatalab/labelU)
- [LabelLLM (An Open-source LLM Dialogue Annotation Platform)](https://github.com/opendatalab/LabelLLM)
- [PDF-Extract-Kit (A Comprehensive Toolkit for High-Quality PDF Content Extraction)](https://github.com/opendatalab/PDF-Extract-Kit)
- [OmniDocBench (A Comprehensive Benchmark for Document Parsing and Evaluation)](https://github.com/opendatalab/OmniDocBench)
- [Magic-HTML (Mixed web page extraction tool)](https://github.com/opendatalab/magic-html)
- [Magic-Doc (Fast speed ppt/pptx/doc/docx/pdf extraction tool)](https://github.com/InternLM/magic-doc)
- [Dingo: A Comprehensive AI Data Quality Evaluation Tool](https://github.com/MigoXLab/dingo)

View File

@ -1,108 +1,14 @@
# ==========================================
# 阶段 1:前端构建 (Node.js Builder)
# ==========================================
FROM node:18-alpine AS frontend-builder
WORKDIR /build
# 利用镜像缓存:先拷贝 package.json
COPY web_ui/package*.json ./
RUN npm config set registry https://registry.npmmirror.com && \
npm install
# 拷贝源码并构建
COPY web_ui/ .
RUN npm run build
# ==========================================
# 阶段 2:最终运行环境 (CoreX Runtime)
# ==========================================
FROM crpi-vofi3w62lkohhxsp.cn-shanghai.personal.cr.aliyuncs.com/opendatalab-mineru/corex:4.4.0_torch2.7.1_vllm0.11.2_py3.10
ENV DEBIAN_FRONTEND=noninteractive
ENV MINERU_MODEL_SOURCE=local
# 1. 系统级依赖:仅安装运行时必需的字体和工具
RUN apt-get update && apt-get install -y \
fonts-noto-core \
fonts-noto-cjk \
fontconfig \
libgl1-mesa-glx \
&& fc-cache -fv \
&& apt-get clean && rm -rf /var/lib/apt/lists/*
# 2. Python 依赖安装策略:
# 使用 --no-deps 确保不覆盖 CoreX 预装的 torch 2.7.1 和 vllm 0.11.2
RUN python3 -m pip install -U pip -i https://mirrors.aliyun.com/pypi/simple && \
python3 -m pip install 'mineru[core]>=2.7.4' --no-deps -i https://mirrors.aliyun.com/pypi/simple && \
# 手动补齐 MinerU 运行必需但基础镜像可能缺失的轻量依赖(不会破坏 torch)
python3 -m pip install \
"numpy==1.26.4" \
"opencv-python==4.11.0.86" \
"pydantic>=2.12,<3" \
"modelscope" \
"magic-pdf" \
boto3 \
"pdfminer.six>=20251230" \
pypdfium2 \
pypdf \
reportlab \
pdftext \
huggingface-hub \
json-repair \
fast-langdetect \
scikit-image \
openai \
beautifulsoup4 \
magika \
mineru-vl-utils \
qwen-vl-utils \
matplotlib \
ultralytics \
doclayout_yolo==0.0.4 \
dill \
PyYAML \
ftfy \
shapely \
pyclipper \
omegaconf \
onnxruntime \
python-pptx \
python-docx \
openpyxl \
xlrd \
xlwt \
pylatexenc \
mammoth \
-i https://mirrors.aliyun.com/pypi/simple && \
python3 -m pip cache purge
ENV MINERU_MODEL_SOURCE=local MINERU_MODEL_DIR=/models PYTHONPATH=/app
RUN apt-get update && apt-get install -y --no-install-recommends libgl1 libglib2.0-0 \
&& rm -rf /var/lib/apt/lists/*
WORKDIR /app
# 3. 拷贝项目源码(排除 web_ui 源码以减小体积)
COPY . .
# ... 前面步骤保持不变 ...
# 4. 从阶段 1 拷贝前端构建产物
RUN mkdir -p mineru/cli/static/web
COPY --from=frontend-builder /build/dist/ ./mineru/cli/static/web/
# 5. 配置文件优化:必须先于模型下载执行
RUN mkdir -p /root/.cache/modelscope/hub/models && \
echo '{ \
"models-dir": "/root/.cache/modelscope/hub/models", \
"device-mode": "gpu", \
"vlm-config": { \
"kind": "vllm", \
"precision": "fp16" \
} \
}' > /root/magic-pdf.json
# 6. 模型预下载
RUN /bin/bash -c "export MINERU_MODEL_SOURCE=local && mineru-models-download -s modelscope -m all"
# 7. 入口点
ENTRYPOINT ["/bin/bash", "-c", "exec \"$@\"", "--"]
CMD ["python3", "-m", "mineru.cli.main"]
COPY pyproject.toml README.mineru-upstream.md LICENSE.md ./
COPY mineru ./mineru
COPY scripts/run_parser.py ./scripts/run_parser.py
# Pin vendor binaries before resolving Python dependencies. Conflicts fail the build rather than replacing vendor Torch.
RUN python3 -c "import importlib.metadata as m; from pathlib import Path; Path('/tmp/vendor-constraints.txt').write_text(chr(10).join(n+'=='+m.version(n) for n in ('torch','torchvision','vllm')))" \
&& python3 -m pip install --no-cache-dir -c /tmp/vendor-constraints.txt '.[vlm,pipeline,api]'
COPY docker/entrypoint.mineru-api.sh /entrypoint.sh
ENTRYPOINT ["sh", "/entrypoint.sh"]
CMD ["python3", "scripts/run_parser.py", "--host", "0.0.0.0", "--port", "8000"]

View File

@ -1,65 +1,14 @@
FROM crpi-vofi3w62lkohhxsp.cn-shanghai.personal.cr.aliyuncs.com/opendatalab-mineru/corex:4.4.0_torch2.7.1_vllm0.11.2_py3.10
ENV DEBIAN_FRONTEND=noninteractive
ENV MINERU_MODEL_SOURCE=local
ENV MINERU_MODEL_DIR=/models
ENV MINERU_TOOLS_CONFIG_JSON=/root/mineru.json
RUN apt-get update && apt-get install -y \
fonts-noto-core \
fonts-noto-cjk \
fontconfig \
libgl1-mesa-glx \
&& fc-cache -fv \
&& apt-get clean && rm -rf /var/lib/apt/lists/*
RUN python3 -m pip install -U pip -i https://mirrors.aliyun.com/pypi/simple && \
python3 -m pip install 'mineru[core]>=2.7.4' --no-deps -i https://mirrors.aliyun.com/pypi/simple && \
python3 -m pip install \
"numpy==1.26.4" \
"opencv-python==4.11.0.86" \
"pydantic>=2.12,<3" \
"modelscope" \
"magic-pdf" \
boto3 \
"pdfminer.six>=20251230" \
pypdfium2 \
pypdf \
reportlab \
pdftext \
huggingface-hub \
json-repair \
fast-langdetect \
scikit-image \
openai \
beautifulsoup4 \
magika \
mineru-vl-utils \
qwen-vl-utils \
matplotlib \
ultralytics \
doclayout_yolo==0.0.4 \
dill \
PyYAML \
ftfy \
shapely \
pyclipper \
omegaconf \
onnxruntime \
python-pptx \
python-docx \
openpyxl \
xlrd \
xlwt \
pylatexenc \
mammoth \
-i https://mirrors.aliyun.com/pypi/simple && \
python3 -m pip cache purge
ENV MINERU_MODEL_SOURCE=local MINERU_MODEL_DIR=/models PYTHONPATH=/app
RUN apt-get update && apt-get install -y --no-install-recommends libgl1 libglib2.0-0 \
&& rm -rf /var/lib/apt/lists/*
WORKDIR /app
COPY . .
COPY docker/entrypoint.mineru-api.sh /usr/local/bin/entrypoint.mineru-api.sh
RUN chmod +x /usr/local/bin/entrypoint.mineru-api.sh
ENTRYPOINT ["/usr/local/bin/entrypoint.mineru-api.sh"]
CMD ["mineru-api", "--host", "0.0.0.0", "--port", "8000"]
COPY pyproject.toml README.mineru-upstream.md LICENSE.md ./
COPY mineru ./mineru
COPY scripts/run_parser.py ./scripts/run_parser.py
# Pin vendor binaries before resolving Python dependencies. Conflicts fail the build rather than replacing vendor Torch.
RUN python3 -c "import importlib.metadata as m; from pathlib import Path; Path('/tmp/vendor-constraints.txt').write_text(chr(10).join(n+'=='+m.version(n) for n in ('torch','torchvision','vllm')))" \
&& python3 -m pip install --no-cache-dir -c /tmp/vendor-constraints.txt '.[vlm,pipeline,api]'
COPY docker/entrypoint.mineru-api.sh /entrypoint.sh
ENTRYPOINT ["sh", "/entrypoint.sh"]
CMD ["python3", "scripts/run_parser.py", "--host", "0.0.0.0", "--port", "8000"]

View File

@ -1,87 +1,16 @@
services:
mineru-openai-server:
image: mineru:latest
container_name: mineru-openai-server
restart: always
profiles: ["openai-server"]
ports:
- 30000:30000
environment:
MINERU_MODEL_SOURCE: local
entrypoint: mineru-openai-server
command:
--host 0.0.0.0
--port 30000
# --data-parallel-size 2 # If using multiple GPUs, increase throughput using vllm's multi-GPU parallel mode
# --gpu-memory-utilization 0.5 # If running on a single GPU and encountering VRAM shortage, reduce the KV cache size by this parameter, if VRAM issues persist, try lowering it further to `0.4` or below.
ulimits:
memlock: -1
stack: 67108864
ipc: host
healthcheck:
test: ["CMD-SHELL", "curl -f http://localhost:30000/health || exit 1"]
deploy:
resources:
reservations:
devices:
- driver: nvidia
device_ids: ["0"] # Modify for multiple GPUs: ["0", "1"]
capabilities: [gpu]
mineru-api:
image: mineru:latest
container_name: mineru-api
restart: always
profiles: ["api"]
image: unis-mineru:2.7.4
restart: unless-stopped
ports:
- 8000:8000
- "127.0.0.1:${MINERU_PORT:-8100}:8000"
environment:
MINERU_MODEL_SOURCE: local
entrypoint: mineru-api
command:
--host 0.0.0.0
--port 8000
# parameters for vllm-engine
# --data-parallel-size 2 # If using multiple GPUs, increase throughput using vllm's multi-GPU parallel mode
# --gpu-memory-utilization 0.5 # If running on a single GPU and encountering VRAM shortage, reduce the KV cache size by this parameter, if VRAM issues persist, try lowering it further to `0.4` or below.
ulimits:
memlock: -1
stack: 67108864
ipc: host
deploy:
resources:
reservations:
devices:
- driver: nvidia
device_ids: ["0"] # Modify for multiple GPUs: ["0", "1"]
capabilities: [gpu]
mineru-gradio:
image: mineru:latest
container_name: mineru-gradio
restart: always
profiles: ["gradio"]
ports:
- 7860:7860
environment:
MINERU_MODEL_SOURCE: local
entrypoint: mineru-gradio
command:
--server-name 0.0.0.0
--server-port 7860
# --enable-api false # If you want to disable the API, set this to false
# --max-convert-pages 20 # If you want to limit the number of pages for conversion, set this to a specific number
# parameters for vllm-engine
# --data-parallel-size 2 # If using multiple GPUs, increase throughput using vllm's multi-GPU parallel mode
# --gpu-memory-utilization 0.5 # If running on a single GPU and encountering VRAM shortage, reduce the KV cache size by this parameter, if VRAM issues persist, try lowering it further to `0.4` or below.
ulimits:
memlock: -1
stack: 67108864
ipc: host
deploy:
resources:
reservations:
devices:
- driver: nvidia
device_ids: ["0"] # Modify for multiple GPUs: ["0", "1"]
capabilities: [gpu]
MINERU_MODEL_DIR: /models
MINERU_GPU_MEMORY_UTILIZATION: ${MINERU_GPU_MEMORY_UTILIZATION:-0.5}
volumes:
- "${MINERU_MODEL_DIR:?Set MINERU_MODEL_DIR}:/models:ro"
- mineru-output:/app/output
command: ["python3", "scripts/run_parser.py", "--host", "0.0.0.0", "--port", "8000"]
volumes:
mineru-output:

View File

@ -6,10 +6,11 @@ PIPELINE_MODEL_DIR=${MINERU_PIPELINE_MODEL_DIR:-${MODEL_DIR}/pipeline}
VLM_MODEL_DIR=${MINERU_VLM_MODEL_DIR:-${MODEL_DIR}/vlm}
CONFIG_PATH=${MINERU_TOOLS_CONFIG_JSON:-/root/mineru.json}
mkdir -p "$PIPELINE_MODEL_DIR" "$VLM_MODEL_DIR" "$(dirname "$CONFIG_PATH")"
mkdir -p "$(dirname "$CONFIG_PATH")"
export CONFIG_PATH PIPELINE_MODEL_DIR VLM_MODEL_DIR
if [ ! -f "$CONFIG_PATH" ]; then
python3 - <<'PY'
import json
import os
@ -39,6 +40,7 @@ config = {
}
config_path.write_text(json.dumps(config, ensure_ascii=False, indent=2), encoding="utf-8")
PY
fi
export MINERU_TOOLS_CONFIG_JSON="$CONFIG_PATH"
export MINERU_MODEL_SOURCE=${MINERU_MODEL_SOURCE:-local}

View File

@ -1,39 +1,14 @@
# Use the official vllm image for gpu with Ampere、Ada Lovelace、Hopper architecture (8.0 <= Compute Capability <= 9.0)
# Compute Capability version query (https://developer.nvidia.com/cuda-gpus)
# only support x86_64 architecture
FROM vllm/vllm-openai:v0.10.1.1
# Use the official vllm image for gpu with Volta、Turing、Blackwell architecture (7.0 < Compute Capability < 8.0 or Compute Capability >= 10.0)
# support x86_64 architecture and ARM(AArch64) architecture
# FROM vllm/vllm-openai:v0.11.0
# Install libgl for opencv support & Noto fonts for Chinese characters
RUN apt-get update && \
apt-get install -y \
fonts-noto-core \
fonts-noto-cjk \
fontconfig \
libgl1 && \
fc-cache -fv && \
apt-get clean && \
rm -rf /var/lib/apt/lists/*
# Install LibreOffice for Word to PDF conversion (minimal installation)
RUN apt-get update && \
apt-get install -y --no-install-recommends \
libreoffice-writer \
libreoffice-core \
fonts-noto-core \
fonts-noto-cjk && \
apt-get clean && \
rm -rf /var/lib/apt/lists/*
# Install mineru latest
RUN python3 -m pip install -U 'mineru[core]>=2.7.0' --break-system-packages && \
python3 -m pip cache purge
# Download models and update the configuration file
RUN /bin/bash -c "mineru-models-download -s huggingface -m all"
# Set the entry point to activate the virtual environment and run the command line tool
ENTRYPOINT ["/bin/bash", "-c", "export MINERU_MODEL_SOURCE=local && exec \"$@\"", "--"]
FROM pytorch/pytorch:2.7.1-cuda12.6-cudnn9-runtime
ENV MINERU_MODEL_SOURCE=local MINERU_MODEL_DIR=/models PYTHONPATH=/app
RUN apt-get update && apt-get install -y --no-install-recommends libgl1 libglib2.0-0 \
&& rm -rf /var/lib/apt/lists/*
WORKDIR /app
COPY pyproject.toml README.mineru-upstream.md LICENSE.md ./
COPY mineru ./mineru
COPY scripts/run_parser.py ./scripts/run_parser.py
# Pin vendor binaries before resolving Python dependencies. Conflicts fail the build rather than replacing vendor Torch.
RUN python3 -c "import importlib.metadata as m; from pathlib import Path; Path('/tmp/vendor-constraints.txt').write_text(chr(10).join(n+'=='+m.version(n) for n in ('torch','torchvision')))" \
&& python3 -m pip install --no-cache-dir -c /tmp/vendor-constraints.txt '.[vlm,pipeline,api]'
COPY docker/entrypoint.mineru-api.sh /entrypoint.sh
ENTRYPOINT ["sh", "/entrypoint.sh"]
CMD ["python3", "scripts/run_parser.py", "--host", "0.0.0.0", "--port", "8000"]

View File

@ -1,67 +1,14 @@
FROM pytorch/pytorch:2.7.1-cuda12.6-cudnn9-runtime
ENV DEBIAN_FRONTEND=noninteractive
ENV MINERU_MODEL_SOURCE=local
ENV MINERU_MODEL_DIR=/models
ENV MINERU_TOOLS_CONFIG_JSON=/root/mineru.json
RUN apt-get update && apt-get install -y \
fonts-noto-core \
fonts-noto-cjk \
fontconfig \
libgl1 \
libglib2.0-0 \
git \
&& fc-cache -fv \
&& apt-get clean && rm -rf /var/lib/apt/lists/*
RUN ln -sf "$(command -v python)" /usr/local/bin/python3 && \
python -m pip install -U pip -i https://mirrors.aliyun.com/pypi/simple && \
python -m pip install \
'mineru[core]>=2.7.4' \
'vllm==0.11.2' \
'pydantic>=2.12,<3' \
modelscope \
magic-pdf \
boto3 \
'pdfminer.six>=20251230' \
pypdfium2 \
pypdf \
reportlab \
pdftext \
huggingface-hub \
json-repair \
fast-langdetect \
scikit-image \
openai \
beautifulsoup4 \
magika \
mineru-vl-utils \
qwen-vl-utils \
matplotlib \
ultralytics \
doclayout_yolo==0.0.4 \
dill \
PyYAML \
ftfy \
shapely \
pyclipper \
omegaconf \
onnxruntime-gpu \
python-pptx \
python-docx \
openpyxl \
xlrd \
xlwt \
pylatexenc \
mammoth \
-i https://mirrors.aliyun.com/pypi/simple && \
python -m pip cache purge
ENV MINERU_MODEL_SOURCE=local MINERU_MODEL_DIR=/models PYTHONPATH=/app
RUN apt-get update && apt-get install -y --no-install-recommends libgl1 libglib2.0-0 \
&& rm -rf /var/lib/apt/lists/*
WORKDIR /app
COPY . .
COPY docker/entrypoint.mineru-api.sh /usr/local/bin/entrypoint.mineru-api.sh
RUN chmod +x /usr/local/bin/entrypoint.mineru-api.sh
ENTRYPOINT ["/usr/local/bin/entrypoint.mineru-api.sh"]
CMD ["mineru-api", "--host", "0.0.0.0", "--port", "8000"]
COPY pyproject.toml README.mineru-upstream.md LICENSE.md ./
COPY mineru ./mineru
COPY scripts/run_parser.py ./scripts/run_parser.py
# Pin vendor binaries before resolving Python dependencies. Conflicts fail the build rather than replacing vendor Torch.
RUN python3 -c "import importlib.metadata as m; from pathlib import Path; Path('/tmp/vendor-constraints.txt').write_text(chr(10).join(n+'=='+m.version(n) for n in ('torch','torchvision')))" \
&& python3 -m pip install --no-cache-dir -c /tmp/vendor-constraints.txt '.[vlm,pipeline,api]'
COPY docker/entrypoint.mineru-api.sh /entrypoint.sh
ENTRYPOINT ["sh", "/entrypoint.sh"]
CMD ["python3", "scripts/run_parser.py", "--host", "0.0.0.0", "--port", "8000"]

View File

@ -0,0 +1,6 @@
{
"baseline": "PyPI mineru==2.7.4",
"url": "https://files.pythonhosted.org/packages/af/e4/60bbe4af8f4f7fa1cef4fd25e6adfb44589172830ab246a673398651f520/mineru-2.7.4-py3-none-any.whl",
"sha256": "e802a59352e4e5ea7782a903c35053581c3539f55a0ebaa95d9afd473bc757bc",
"files": 187
}

View File

@ -1,3 +1,5 @@
> 状态更新:下文为历史差异盘点,4.0.12 合并方案已被替代。当前固定官方 2.7.4,业务功能迁入 MaxKB,参见仓库 README。
# MinerU 2.7.1 定制盘点与 4.0.12 合并方案
核验日期:2026-10-22。此次完成源码与发布包对比,尚未替换旧内核、安装新依赖或切换运行服务。

View File

@ -424,7 +424,8 @@ def __tie_up_category_by_index(blocks, subject_block_type, object_block_type):
# 调用通用方法
return tie_up_category_by_index(
get_subjects,
get_objects
get_objects,
object_block_type=object_block_type
)

View File

@ -297,7 +297,14 @@ def ocr_det_batch_setting(device):
# 检测torch的版本号
import torch
from packaging import version
if version.parse(torch.__version__) >= version.parse("2.8.0") or str(device).startswith('mps'):
device_type = os.getenv("MINERU_LMDEPLOY_DEVICE", "")
if (
version.parse(torch.__version__) >= version.parse("2.8.0")
or str(device).startswith('mps')
or device_type.lower() in ["corex"]
):
enable_ocr_det_batch = False
else:
enable_ocr_det_batch = True

View File

@ -193,7 +193,12 @@ def batch_image_analyze(
# 检测torch的版本号
import torch
from packaging import version
if version.parse(torch.__version__) >= version.parse("2.8.0") or str(device).startswith('mps'):
device_type = os.getenv("MINERU_LMDEPLOY_DEVICE", "")
if (
version.parse(torch.__version__) >= version.parse("2.8.0")
or str(device).startswith('mps')
or device_type.lower() in ["corex"]
):
enable_ocr_det_batch = False
else:
enable_ocr_det_batch = True

View File

@ -18,6 +18,12 @@ def enable_custom_logits_processors() -> bool:
compute_capability = f"{major}.{minor}"
elif hasattr(torch, 'npu') and torch.npu.is_available():
compute_capability = "8.0"
elif hasattr(torch, 'gcu') and torch.gcu.is_available():
compute_capability = "8.0"
elif hasattr(torch, 'musa') and torch.musa.is_available():
compute_capability = "8.0"
elif hasattr(torch, 'mlu') and torch.mlu.is_available():
compute_capability = "8.0"
else:
logger.info("CUDA not available, disabling custom_logits_processors")
return False
@ -71,16 +77,6 @@ def set_lmdeploy_backend(device_type: str) -> str:
def set_default_gpu_memory_utilization() -> float:
override = os.getenv("MINERU_GPU_MEMORY_UTILIZATION")
if override:
try:
value = float(override)
except ValueError:
raise ValueError("MINERU_GPU_MEMORY_UTILIZATION must be a float between 0 and 1")
if not 0 < value < 1:
raise ValueError("MINERU_GPU_MEMORY_UTILIZATION must be a float between 0 and 1")
return value
from vllm import __version__ as vllm_version
device = get_device()
gpu_memory = get_vram(device)
@ -106,4 +102,4 @@ def set_default_batch_size() -> int:
except Exception as e:
logger.warning(f'Error determining VRAM: {e}, using default batch_ratio: 1')
batch_size = 1
return batch_size
return batch_size

View File

@ -1,6 +1,7 @@
# Copyright (c) Opendatalab. All rights reserved.
import os
import time
import json
from loguru import logger
@ -99,6 +100,37 @@ class ModelSingleton:
import vllm
except ImportError:
raise ImportError("Please install vllm to use the vllm-engine backend.")
# musa vllm v1 引擎特殊配置
# device = get_device()
# if device_type.startswith("musa"):
# import torch
# if torch.musa.is_available():
# compilation_config = {
# "cudagraph_capture_sizes": [1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 14, 16, 18, 20, 24, 28, 30],
# "simple_cuda_graph": True
# }
# block_size = 32
# kwargs["compilation_config"] = compilation_config
# kwargs["block_size"] = block_size
# corex vllm v1 引擎特殊配置
device_type = os.getenv("MINERU_LMDEPLOY_DEVICE", "")
if device_type.lower() == "corex":
compilation_config = {
"cudagraph_mode": "FULL_DECODE_ONLY",
"level": 0
}
kwargs["compilation_config"] = compilation_config
if "compilation_config" in kwargs:
if isinstance(kwargs["compilation_config"], str):
try:
kwargs["compilation_config"] = json.loads(kwargs["compilation_config"])
except json.JSONDecodeError:
logger.warning(
f"Failed to parse compilation_config as JSON: {kwargs['compilation_config']}")
del kwargs["compilation_config"]
if "gpu_memory_utilization" not in kwargs:
kwargs["gpu_memory_utilization"] = set_default_gpu_memory_utilization()
if "model" not in kwargs:
@ -110,11 +142,48 @@ class ModelSingleton:
vllm_llm = vllm.LLM(**kwargs)
elif backend == "vllm-async-engine":
try:
import inspect
from vllm.engine.arg_utils import AsyncEngineArgs
from vllm.v1.engine.async_llm import AsyncLLM
from vllm.config import CompilationConfig
except ImportError:
raise ImportError("Please install vllm to use the vllm-async-engine backend.")
# musa vllm v1 引擎特殊配置
# device = get_device()
# if device.startswith("musa"):
# import torch
# if torch.musa.is_available():
# compilation_config = CompilationConfig(
# cudagraph_capture_sizes=[1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 14, 16, 18, 20, 24, 28, 30],
# simple_cuda_graph=True
# )
# block_size = 32
# kwargs["compilation_config"] = compilation_config
# kwargs["block_size"] = block_size
# corex vllm v1 引擎特殊配置
device_type = os.getenv("MINERU_LMDEPLOY_DEVICE", "")
if device_type.lower() == "corex":
compilation_config = CompilationConfig(
cudagraph_mode="FULL_DECODE_ONLY",
level=0
)
kwargs["compilation_config"] = compilation_config
if "compilation_config" in kwargs:
if isinstance(kwargs["compilation_config"], dict):
# 如果是字典,转换为 CompilationConfig 对象
kwargs["compilation_config"] = CompilationConfig(**kwargs["compilation_config"])
elif isinstance(kwargs["compilation_config"], str):
# 如果是 JSON 字符串,先解析再转换
try:
config_dict = json.loads(kwargs["compilation_config"])
kwargs["compilation_config"] = CompilationConfig(**config_dict)
except (json.JSONDecodeError, TypeError) as e:
logger.warning(
f"Failed to parse compilation_config: {kwargs['compilation_config']}, error: {e}")
del kwargs["compilation_config"]
if "gpu_memory_utilization" not in kwargs:
kwargs["gpu_memory_utilization"] = set_default_gpu_memory_utilization()
if "model" not in kwargs:
@ -123,8 +192,6 @@ class ModelSingleton:
from mineru_vl_utils import MinerULogitsProcessor
kwargs["logits_processors"] = [MinerULogitsProcessor]
# 使用kwargs为 vllm初始化参数
supported_args = set(inspect.signature(AsyncEngineArgs).parameters)
kwargs = {k: v for k, v in kwargs.items() if k in supported_args}
vllm_async_llm = AsyncLLM.from_engine_args(AsyncEngineArgs(**kwargs))
elif backend == "lmdeploy-engine":
try:

View File

@ -349,7 +349,8 @@ def __tie_up_category_by_index(blocks, subject_block_type, object_block_type):
# 调用通用方法
return tie_up_category_by_index(
get_subjects,
get_objects
get_objects,
object_block_type=object_block_type
)

View File

@ -26,7 +26,6 @@ if os.getenv("MINERU_LMDEPLOY_DEVICE", "") == "maca":
pdf_suffixes = ["pdf"]
image_suffixes = ["png", "jpeg", "jp2", "webp", "gif", "bmp", "jpg", "tiff"]
word_suffixes = ["docx", "doc"]
os.environ["TOKENIZERS_PARALLELISM"] = "false"
@ -40,109 +39,6 @@ def read_fn(path):
return images_bytes_to_pdf_bytes(file_bytes)
elif file_suffix in pdf_suffixes:
return file_bytes
elif file_suffix in word_suffixes:
# 处理 Word 文件,转换为 PDF(保持原格式)
from io import BytesIO
import tempfile
import os
try:
# 方法 1: 使用 LibreOffice 进行高质量转换(推荐)
# 创建临时文件
with tempfile.NamedTemporaryFile(suffix=f'.{file_suffix}', delete=False) as temp_word:
temp_word.write(file_bytes)
temp_word_path = temp_word.name
pdf_fd, pdf_path = tempfile.mkstemp(suffix='.pdf')
os.close(pdf_fd)
try:
# 使用 LibreOffice 进行转换(保持原格式,包括图片、表格等)
import subprocess
result = subprocess.run(
[
'libreoffice', '--headless', '--convert-to', 'pdf',
'--outdir', os.path.dirname(pdf_path),
temp_word_path
],
capture_output=True,
timeout=120
)
if result.returncode == 0:
# LibreOffice 会在同目录下生成 PDF
base_name = os.path.basename(temp_word_path).rsplit('.', 1)[0]
generated_pdf = os.path.join(os.path.dirname(temp_word_path), f"{base_name}.pdf")
if os.path.exists(generated_pdf):
with open(generated_pdf, 'rb') as f:
pdf_bytes_result = f.read()
os.remove(generated_pdf)
os.remove(temp_word_path)
return pdf_bytes_result
logger.warning(f"LibreOffice conversion failed: {result.stderr}")
except (subprocess.TimeoutExpired, FileNotFoundError) as libreoffice_error:
logger.warning(f"LibreOffice not available or failed: {libreoffice_error}")
finally:
# 清理临时文件
if os.path.exists(temp_word_path):
os.remove(temp_word_path)
if os.path.exists(pdf_path):
os.remove(pdf_path)
# 方法 2: 如果 LibreOffice 不可用,使用 python-docx + reportlab 作为备选
if file_suffix == "docx":
from docx import Document
from reportlab.lib.pagesizes import letter
from reportlab.pdfgen import canvas
# 读取 Word 文件
doc = Document(BytesIO(file_bytes))
# 创建 PDF(简单文本转换,不保留复杂格式)
output = BytesIO()
c = canvas.Canvas(output, pagesize=letter)
width, height = letter
y = height - 50
for para in doc.paragraphs:
text = para.text
if text:
# 处理长文本自动换行
max_width = width - 100
words = text.split(' ')
line = ''
for word in words:
test_line = f"{line} {word}".strip()
if c.stringWidth(test_line, 'Helvetica', 12) < max_width:
line = test_line
else:
if y < 50:
c.showPage()
y = height - 50
c.drawString(50, y, line)
y -= 15
line = word
# 写入最后一行
if y < 50:
c.showPage()
y = height - 50
c.drawString(50, y, line)
y -= 15
c.save()
output.seek(0)
return output.read()
raise Exception("Word to PDF conversion failed with all methods")
except Exception as e:
logger.exception(f"Error converting Word file to PDF: {e}")
raise Exception(f"Failed to convert Word file to PDF: {str(e)}. Please install LibreOffice for better conversion: sudo apt-get install libreoffice")
else:
raise Exception(f"Unknown file suffix: {file_suffix}")
@ -341,7 +237,6 @@ async def _async_process_vlm(
f_dump_content_list,
f_make_md_mode,
server_url=None,
progress_callback=None,
**kwargs,
):
"""异步处理VLM后端逻辑"""
@ -350,23 +245,14 @@ async def _async_process_vlm(
if not backend.endswith("client"):
server_url = None
total_files = len(pdf_bytes_list)
for idx, pdf_bytes in enumerate(pdf_bytes_list):
pdf_file_name = pdf_file_names[idx]
local_image_dir, local_md_dir = prepare_env(output_dir, pdf_file_name, parse_method)
image_writer, md_writer = FileBasedDataWriter(local_image_dir), FileBasedDataWriter(local_md_dir)
# 更新进度
if progress_callback:
progress_callback((idx / total_files) * 50, f"处理文件 {idx+1}/{total_files}")
middle_json, infer_result = await aio_vlm_doc_analyze(
pdf_bytes, image_writer=image_writer, backend=backend, server_url=server_url, **kwargs,
)
# 更新进度
if progress_callback:
progress_callback(((idx + 1) / total_files) * 100, f"完成文件 {idx+1}/{total_files}")
pdf_info = middle_json["pdf_info"]
@ -489,7 +375,6 @@ async def _async_process_hybrid(
f_dump_content_list,
f_make_md_mode,
server_url=None,
progress_callback=None,
**kwargs,
):
from mineru.backend.hybrid.hybrid_analyze import aio_doc_analyze as aio_hybrid_doc_analyze
@ -497,15 +382,10 @@ async def _async_process_hybrid(
if not backend.endswith("client"):
server_url = None
total_files = len(pdf_bytes_list)
for idx, (pdf_bytes, lang) in enumerate(zip(pdf_bytes_list, h_lang_list)):
pdf_file_name = pdf_file_names[idx]
local_image_dir, local_md_dir = prepare_env(output_dir, pdf_file_name, f"hybrid_{parse_method}")
image_writer, md_writer = FileBasedDataWriter(local_image_dir), FileBasedDataWriter(local_md_dir)
# 更新进度
if progress_callback:
progress_callback((idx / total_files) * 50, f"处理文件 {idx+1}/{total_files}")
middle_json, infer_result, _vlm_ocr_enable = await aio_hybrid_doc_analyze(
pdf_bytes,
@ -517,10 +397,6 @@ async def _async_process_hybrid(
server_url=server_url,
**kwargs,
)
# 更新进度
if progress_callback:
progress_callback(((idx + 1) / total_files) * 100, f"完成文件 {idx+1}/{total_files}")
pdf_info = middle_json["pdf_info"]
@ -627,7 +503,6 @@ async def aio_do_parse(
f_make_md_mode=MakeMode.MM_MD,
start_page_id=0,
end_page_id=None,
progress_callback=None,
**kwargs,
):
# 预处理PDF字节数据
@ -654,12 +529,11 @@ async def aio_do_parse(
os.environ['MINERU_VLM_FORMULA_ENABLE'] = str(formula_enable)
os.environ['MINERU_VLM_TABLE_ENABLE'] = str(table_enable)
# 传递进度回调
await _async_process_vlm(
output_dir, pdf_file_names, pdf_bytes_list, backend,
f_draw_layout_bbox, f_draw_span_bbox, f_dump_md, f_dump_middle_json,
f_dump_model_output, f_dump_orig_pdf, f_dump_content_list, f_make_md_mode,
server_url, progress_callback=progress_callback, **kwargs,
server_url, **kwargs,
)
elif backend.startswith("hybrid-"):
backend = backend[7:]
@ -673,12 +547,11 @@ async def aio_do_parse(
os.environ['MINERU_VLM_TABLE_ENABLE'] = str(table_enable)
os.environ['MINERU_VLM_FORMULA_ENABLE'] = "true"
# 传递进度回调
await _async_process_hybrid(
output_dir, pdf_file_names, pdf_bytes_list, p_lang_list, parse_method, formula_enable, backend,
f_draw_layout_bbox, f_draw_span_bbox, f_dump_md, f_dump_middle_json,
f_dump_model_output, f_dump_orig_pdf, f_dump_content_list, f_make_md_mode,
server_url, progress_callback=progress_callback, **kwargs,
server_url, **kwargs,
)

File diff suppressed because it is too large Load Diff

View File

@ -23,143 +23,26 @@ from mineru.utils.engine_utils import get_vlm_engine
from mineru.utils.hash_utils import str_sha256
# --- 新增:标准的树状思维导图生成函数 ---
def md_to_markmap_html(md_content):
"""
将 Markdown 转换为标准的树状思维导图 (Markmap)
"""
if not md_content:
return ""
# 转义 Markdown 中的反引号和符号,防止破坏 JS 字符串
safe_md = md_content.replace('`', '\\`').replace('$', '\\$')
# 完整的 HTML + Markmap 渲染引擎
full_html = f"""
<!DOCTYPE html>
<html>
<head>
<meta charset="UTF-8">
<style>
html, body, #mindmap {{ width: 100%; height: 100%; margin: 0; padding: 0; overflow: hidden; background-color: white; }}
</style>
<script src="https://cdn.jsdelivr.net/npm/d3@7"></script>
<script src="https://cdn.jsdelivr.net/npm/markmap-view"></script>
<script src="https://cdn.jsdelivr.net/npm/markmap-toolbar"></script>
<script src="https://cdn.jsdelivr.net/npm/markmap-lib"></script>
</head>
<body>
<svg id="mindmap"></svg>
<script>
// 等待页面完全加载后再初始化
document.addEventListener('DOMContentLoaded', function() {{
try {{
const {{ Markmap, loadCSS, loadJS, Transformer }} = window.markmap;
const transformer = new Transformer();
const {{ root, features }} = transformer.transform(`{safe_md}`);
const {{ styles, scripts }} = transformer.getAssets();
if (styles) loadCSS(styles);
if (scripts) loadJS(scripts);
// 增加延迟确保资源加载完成
setTimeout(() => {{
const mm = Markmap.create('#mindmap', {{
autoFit: true,
fitRatio: 0.9,
initialExpandLevel: -1
}}, root);
// 添加错误处理
if (mm) {{
console.log("Markmap created successfully");
}} else {{
console.error("Failed to create Markmap");
}}
}}, 500); // 延迟500毫秒以确保资源加载
}} catch (error) {{
console.error('Error initializing markmap:', error);
}}
}});
</script>
</body>
</html>
"""
# 使用 iframe 封装,彻底解决渲染失效问题
iframe_srcdoc = full_html.replace('"', '&quot;')
iframe_code = f"""
<iframe srcdoc="{iframe_srcdoc}" style="width: 100%; height: 800px; border: 1px solid #ddd; border-radius: 8px;" sandbox="allow-scripts"></iframe>
"""
return iframe_code
# ────────────── 新增:根据上一级标题自动补全下一级标题 ──────────────
def auto_promote_paragraphs_to_subheading(text):
lines = text.splitlines()
result = []
in_section = False
empty_count = 0
for line in lines:
stripped = line.strip()
if stripped.startswith('# '):
result.append(line)
in_section = True
empty_count = 0
continue
if stripped.startswith('#'):
result.append(line)
in_section = False
empty_count = 0
continue
if not stripped:
result.append(line)
empty_count += 1
if empty_count >= 2:
in_section = False
continue
# 跳过图片、列表、代码等特殊行
if (
stripped.startswith('![') or
stripped.startswith('>') or
stripped.startswith('```') or
re.match(r'^[-*+] ', stripped) or
re.match(r'^\d+\. ', stripped)
):
result.append(line)
empty_count = 0
continue
empty_count = 0
if in_section:
result.append('## ' + stripped)
else:
result.append(line)
return '\n'.join(result)
async def parse_pdf(doc_path, output_dir, end_page_id, is_ocr, formula_enable, table_enable, language, backend, url):
os.makedirs(output_dir, exist_ok=True)
try:
file_name = f'{safe_stem(Path(doc_path).stem)}_{time.strftime("%y%m%d_%H%M%S")}'
pdf_data = read_fn(doc_path)
# 根据 backend 确定 parse_method
if backend.startswith("vlm"):
parse_method = "vlm"
else:
parse_method = 'ocr' if is_ocr else 'auto'
# 根据 backend 类型准备环境目录
if backend.startswith("hybrid"):
env_name = f"hybrid_{parse_method}"
else:
env_name = parse_method
local_image_dir, local_md_dir = prepare_env(output_dir, file_name, env_name)
await aio_do_parse(
output_dir=output_dir,
pdf_file_names=[file_name],
@ -179,12 +62,22 @@ async def parse_pdf(doc_path, output_dir, end_page_id, is_ocr, formula_enable, t
def compress_directory_to_zip(directory_path, output_zip_path):
"""压缩指定目录到一个 ZIP 文件。
:param directory_path: 要压缩的目录路径
:param output_zip_path: 输出的 ZIP 文件路径
"""
try:
with zipfile.ZipFile(output_zip_path, 'w', zipfile.ZIP_DEFLATED) as zipf:
# 遍历目录中的所有文件和子目录
for root, dirs, files in os.walk(directory_path):
for file in files:
# 构建完整的文件路径
file_path = os.path.join(root, file)
# 计算相对路径
arcname = os.path.relpath(file_path, directory_path)
# 添加文件到 ZIP 文件
zipf.write(file_path, arcname)
return 0
except Exception as e:
@ -198,32 +91,31 @@ def image_to_base64(image_path):
def replace_image_with_base64(markdown_text, image_dir_path):
# 匹配Markdown中的图片标签
pattern = r'\!\[(?:[^\]]*)\]\(([^)]+)\)'
# 替换图片链接
def replace(match):
relative_path = match.group(1)
# 只处理以.jpg结尾的图片
if relative_path.endswith('.jpg'):
full_path = os.path.join(image_dir_path, relative_path)
base64_image = image_to_base64(full_path)
return f'![{relative_path}](data:image/jpeg;base64,{base64_image})'
return match.group(0)
else:
# 其他格式的图片保持原样
return match.group(0)
# 应用替换
return re.sub(pattern, replace, markdown_text)
async def to_markdown(file_path, end_pages=10, is_ocr=False, formula_enable=True, table_enable=True, language="ch",
backend="pipeline", url=None):
async def to_markdown(file_path, end_pages=10, is_ocr=False, formula_enable=True, table_enable=True, language="ch", backend="pipeline", url=None):
# 如果language包含(),则提取括号前的内容作为实际语言
if '(' in language and ')' in language:
language = language.split('(')[0].strip()
file_path = to_pdf(file_path)
# 打印请求参数日志
logger.info(f"parse_pdf 请求参数: file_path={file_path}, output_dir='./output', end_page_id={end_pages - 1}, "
f"is_ocr={is_ocr}, formula_enable={formula_enable}, table_enable={table_enable}, "
f"language='{language}', backend='{backend}', url={url}")
# 获取识别的md文件以及压缩包文件路径
local_md_dir, file_name = await parse_pdf(file_path, './output', end_pages - 1, is_ocr, formula_enable,
table_enable, language, backend, url)
local_md_dir, file_name = await parse_pdf(file_path, './output', end_pages - 1, is_ocr, formula_enable, table_enable, language, backend, url)
archive_zip_path = os.path.join('./output', str_sha256(local_md_dir) + '.zip')
zip_archive_success = compress_directory_to_zip(local_md_dir, archive_zip_path)
if zip_archive_success == 0:
@ -233,55 +125,74 @@ async def to_markdown(file_path, end_pages=10, is_ocr=False, formula_enable=True
md_path = os.path.join(local_md_dir, file_name + '.md')
with open(md_path, 'r', encoding='utf-8') as f:
txt_content = f.read()
# ────────────── 自动补全:根据 # 标题补全后续段落为 ## ──────────────
txt_content = auto_promote_paragraphs_to_subheading(txt_content)
# ────────────────────────────────────────────────────────────────
md_content = replace_image_with_base64(txt_content, local_md_dir)
# 生成思维导图HTML - 使用新的实现
mind_map_html = md_to_markmap_html(txt_content)
# 返回转换后的PDF路径
new_pdf_path = os.path.join(local_md_dir, file_name + '_layout.pdf')
return md_content, txt_content, archive_zip_path, new_pdf_path, mind_map_html
return md_content, txt_content, archive_zip_path, new_pdf_path
latex_delimiters_type_all = [
{'left': '$$', 'right': '$$', 'display': True},
{'left': '$', 'right': '$', 'display': False},
{'left': '\\(', 'right': '\\)', 'display': False},
{'left': '\\[', 'right': '\\]', 'display': True},
]
latex_delimiters_type_a = [
{'left': '$$', 'right': '$$', 'display': True},
{'left': '$', 'right': '$', 'display': False},
]
latex_delimiters_type_b = [
{'left': '\\(', 'right': '\\)', 'display': False},
{'left': '\\[', 'right': '\\]', 'display': True},
]
latex_delimiters_type_all = latex_delimiters_type_a + latex_delimiters_type_b
other_lang = ['ch (Chinese, English, Chinese Traditional)', 'en (English)', 'korean', 'japan']
all_lang = [*other_lang]
header_path = os.path.join(os.path.dirname(os.path.dirname(__file__)), 'resources', 'header.html')
with open(header_path, mode='r', encoding='utf-8') as header_file:
header = header_file.read()
other_lang = [
'ch (Chinese, English, Chinese Traditional)',
'ch_lite (Chinese, English, Chinese Traditional, Japanese)',
'ch_server (Chinese, English, Chinese Traditional, Japanese)',
'en (English)',
'korean (Korean, English)',
'japan (Chinese, English, Chinese Traditional, Japanese)',
'chinese_cht (Chinese, English, Chinese Traditional, Japanese)',
'ta (Tamil, English)',
'te (Telugu, English)',
'ka (Kannada)',
'el (Greek, English)',
'th (Thai, English)'
]
add_lang = [
'latin (French, German, Afrikaans, Italian, Spanish, Bosnian, Portuguese, Czech, Welsh, Danish, Estonian, Irish, Croatian, Uzbek, Hungarian, Serbian (Latin), Indonesian, Occitan, Icelandic, Lithuanian, Maori, Malay, Dutch, Norwegian, Polish, Slovak, Slovenian, Albanian, Swedish, Swahili, Tagalog, Turkish, Latin, Azerbaijani, Kurdish, Latvian, Maltese, Pali, Romanian, Vietnamese, Finnish, Basque, Galician, Luxembourgish, Romansh, Catalan, Quechua)',
'arabic (Arabic, Persian, Uyghur, Urdu, Pashto, Kurdish, Sindhi, Balochi, English)',
'east_slavic (Russian, Belarusian, Ukrainian, English)',
'cyrillic (Russian, Belarusian, Ukrainian, Serbian (Cyrillic), Bulgarian, Mongolian, Abkhazian, Adyghe, Kabardian, Avar, Dargin, Ingush, Chechen, Lak, Lezgin, Tabasaran, Kazakh, Kyrgyz, Tajik, Macedonian, Tatar, Chuvash, Bashkir, Malian, Moldovan, Udmurt, Komi, Ossetian, Buryat, Kalmyk, Tuvan, Sakha, Karakalpak, English)',
'devanagari (Hindi, Marathi, Nepali, Bihari, Maithili, Angika, Bhojpuri, Magahi, Santali, Newari, Konkani, Sanskrit, Haryanvi, English)'
]
all_lang = [*other_lang, *add_lang]
def safe_stem(file_path):
stem = Path(file_path).stem
# 只保留字母、数字、下划线和点,其他字符替换为下划线
return re.sub(r'[^\w.]', '_', stem)
def to_pdf(file_path):
if file_path is None: return None
if file_path is None:
return None
pdf_bytes = read_fn(file_path)
# unique_filename = f'{uuid.uuid4()}.pdf'
unique_filename = f'{safe_stem(file_path)}.pdf'
# 构建完整的文件路径
tmp_file_path = os.path.join(os.path.dirname(file_path), unique_filename)
# 将字节数据写入文件
with open(tmp_file_path, 'wb') as tmp_pdf_file:
tmp_pdf_file.write(pdf_bytes)
return tmp_file_path
@ -339,35 +250,13 @@ def to_pdf(file_path):
default='all',
)
def main(ctx,
example_enable,
http_client_enable,
api_enable, max_convert_pages,
server_name, server_port, latex_delimiters_type, **kwargs
):
# 检测系统语言环境,默认为中文
import locale
import os
example_enable,
http_client_enable,
api_enable, max_convert_pages,
server_name, server_port, latex_delimiters_type, **kwargs
):
def detect_language():
# 检查环境变量
lang = os.getenv('LANG', '')
if 'zh' in lang.lower() or 'chinese' in lang.lower():
return 'zh'
# 检查系统默认locale
try:
default_locale = locale.getdefaultlocale()[0]
if default_locale and 'zh' in default_locale.lower():
return 'zh'
except:
pass
# 默认返回中文
return 'zh'
detected_lang = detect_language()
# 创建 i18n 实例,支持中英文,默认为中文
# 创建 i18n 实例,支持中英文
i18n = gr.I18n(
en={
"upload_file": "Please upload a PDF or image",
@ -375,7 +264,7 @@ def main(ctx,
"backend": "Backend",
"server_url": "Server URL",
"server_url_info": "OpenAI-compatible server URL for http-client backend.",
"recognition_options": "Recognition Options",
"recognition_options": "**Recognition Options:**",
"table_enable": "Enable table recognition",
"table_info": "If disabled, tables will be shown as images.",
"formula_label_vlm": "Enable display formula recognition",
@ -395,7 +284,6 @@ def main(ctx,
"convert_result": "Convert result",
"md_rendering": "Markdown rendering",
"md_text": "Markdown text",
"mind_map": "Mind Map", # 新增
"backend_info_vlm": "High-precision parsing via VLM, supports Chinese and English documents only.",
"backend_info_pipeline": "Traditional Multi-model pipeline parsing, supports multiple languages, hallucination-free.",
"backend_info_hybrid": "High-precision hybrid parsing, supports multiple languages.",
@ -407,7 +295,7 @@ def main(ctx,
"backend": "解析后端",
"server_url": "服务器地址",
"server_url_info": "http-client 后端的 OpenAI 兼容服务器地址。",
"recognition_options": "识别选项",
"recognition_options": "**识别选项:**",
"table_enable": "启用表格识别",
"table_info": "禁用后,表格将显示为图片。",
"formula_label_vlm": "启用行间公式识别",
@ -427,7 +315,6 @@ def main(ctx,
"convert_result": "转换结果",
"md_rendering": "Markdown 渲染",
"md_text": "Markdown 文本",
"mind_map": "思维导图", # 新增
"backend_info_vlm": "多模态大模型高精度解析,仅支持中英文文档。",
"backend_info_pipeline": "传统多模型管道解析,支持多语言,无幻觉。",
"backend_info_hybrid": "高精度混合解析,支持多语言。",
@ -481,6 +368,7 @@ def main(ctx,
return client_options_update, ocr_options_update, formula_label_update, backend_info_update
kwargs.update(arg_parse(ctx))
if latex_delimiters_type == 'a':
@ -511,10 +399,8 @@ def main(ctx,
logger.exception(e)
suffixes = [f".{suffix}" for suffix in pdf_suffixes + image_suffixes]
with gr.Blocks(title="多模态思维导图助手",
fill_height=True) as demo:
# gr.HTML(header)
gr.HTML("<h1 style='text-align: center;'>多模态思维导图助手</h1>")
with gr.Blocks() as demo:
gr.HTML(header)
with gr.Row():
with gr.Column(variant='panel', scale=5):
with gr.Row():
@ -526,21 +412,16 @@ def main(ctx,
preferred_option = "hybrid-auto-engine"
if http_client_enable:
drop_list.extend(["vlm-http-client", "hybrid-http-client"])
backend = gr.Dropdown(drop_list, label=i18n("backend"), value=preferred_option,
info=get_backend_info(preferred_option))
backend = gr.Dropdown(drop_list, label=i18n("backend"), value=preferred_option, info=get_backend_info(preferred_option))
with gr.Row(visible=False) as client_options:
url = gr.Textbox(label=i18n("server_url"), value='http://localhost:30000',
placeholder='http://localhost:30000', info=i18n("server_url_info"))
url = gr.Textbox(label=i18n("server_url"), value='http://localhost:30000', placeholder='http://localhost:30000', info=i18n("server_url_info"))
with gr.Row(equal_height=True):
with gr.Column():
gr.Markdown(i18n("recognition_options"))
table_enable = gr.Checkbox(label=i18n("table_enable"), value=True, info=i18n("table_info"))
formula_enable = gr.Checkbox(label=get_formula_label(preferred_option), value=True,
info=get_formula_info(preferred_option))
formula_enable = gr.Checkbox(label=get_formula_label(preferred_option), value=True, info=get_formula_info(preferred_option))
with gr.Column(visible=False) as ocr_options:
language = gr.Dropdown(all_lang, label=i18n("ocr_language"),
value='ch (Chinese, English, Chinese Traditional)',
info=i18n("ocr_language_info"))
language = gr.Dropdown(all_lang, label=i18n("ocr_language"), value='ch (Chinese, English, Chinese Traditional)', info=i18n("ocr_language_info"))
is_ocr = gr.Checkbox(label=i18n("force_ocr"), value=False, info=i18n("force_ocr_info"))
with gr.Row():
change_bu = gr.Button(i18n("convert"))
@ -563,46 +444,50 @@ def main(ctx,
md = gr.Markdown(
label=i18n("md_rendering"),
height=1200,
show_copy_button=True,
# buttons=["copy"], # gradio 6 以上版本使用
show_copy_button=True, # gradio 6 以下版本使用
latex_delimiters=latex_delimiters,
line_breaks=True
)
with gr.Tab(i18n("md_text")):
md_text = gr.TextArea(
lines=45,
show_copy_button=True,
# buttons=["copy"], # gradio 6 以上版本使用
show_copy_button=True, # gradio 6 以下版本使用
label=i18n("md_text")
)
with gr.Tab(i18n("mind_map")): # 新增的思维导图tab
mind_map = gr.HTML(label=i18n("mind_map")) # 使用HTML组件来渲染markmap
# 添加事件处理
backend.change(
fn=update_interface,
inputs=[backend],
outputs=[client_options, ocr_options, formula_enable, backend],
api_name=False
# api_visibility="private" # gradio 6 以上版本使用
api_name=False # gradio 6 以下版本使用
)
# 添加demo.load事件,在页面加载时触发一次界面更新
demo.load(
fn=update_interface,
inputs=[backend],
outputs=[client_options, ocr_options, formula_enable, backend],
api_name=False
# api_visibility="private" # gradio 6 以上版本使用
api_name=False # gradio 6 以下版本使用
)
clear_bu.add([input_file, md, pdf_show, md_text, output_file, is_ocr, mind_map])
clear_bu.add([input_file, md, pdf_show, md_text, output_file, is_ocr])
input_file.change(
fn=to_pdf,
inputs=input_file,
outputs=pdf_show,
api_name="to_pdf" if api_enable else False
api_name="to_pdf" if api_enable else False, # gradio 6 以下版本使用
# api_visibility="public" if api_enable else "private" # gradio 6 以上版本使用
)
change_bu.click(
fn=to_markdown,
inputs=[input_file, max_pages, is_ocr, formula_enable, table_enable, language, backend, url],
outputs=[md, md_text, output_file, pdf_show, mind_map], # 添加mind_map输出
api_name="to_markdown" if api_enable else False
outputs=[md, md_text, output_file, pdf_show],
api_name="to_markdown" if api_enable else False, # gradio 6 以下版本使用
# api_visibility="public" if api_enable else "private" # gradio 6 以上版本使用
)
footer_links = ["gradio", "settings"]
@ -611,12 +496,11 @@ def main(ctx,
demo.launch(
server_name=server_name,
server_port=server_port,
show_api=api_enable,
i18n=i18n,
# title = "多模态思维导图助手",
# favicon_path = "logo.png"
# footer_links=footer_links, # gradio 6 以上版本使用
show_api=api_enable, # gradio 6 以下版本使用
i18n=i18n
)
if __name__ == '__main__':
main()
main()

File diff suppressed because one or more lines are too long

File diff suppressed because one or more lines are too long

File diff suppressed because one or more lines are too long

File diff suppressed because one or more lines are too long

File diff suppressed because one or more lines are too long

File diff suppressed because one or more lines are too long

File diff suppressed because one or more lines are too long

File diff suppressed because one or more lines are too long

View File

@ -1,28 +0,0 @@
<!DOCTYPE html>
<html lang="zh-CN">
<head>
<meta charset="UTF-8">
<link rel="icon" href="/favicon.ico">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>思维导图</title>
<style>
html, body {
height: 100%;
margin: 0;
padding: 0;
box-sizing: border-box;
}
#app {
height: 100%;
}
</style>
<script type="module" crossorigin src="/assets/index-zcV97yfy.js"></script>
<link rel="modulepreload" crossorigin href="/assets/vendor-DSv9k51x.js">
<link rel="modulepreload" crossorigin href="/assets/markdown-_Dy2zPzC.js">
<link rel="stylesheet" crossorigin href="/assets/index-B_viQAxH.css">
</head>
<body>
<div id="app"></div>
</body>
</html>

File diff suppressed because one or more lines are too long

File diff suppressed because one or more lines are too long

File diff suppressed because one or more lines are too long

File diff suppressed because one or more lines are too long

View File

@ -1,5 +1,7 @@
import os
from typing import List, Union
import torch
from tqdm import tqdm
from ultralytics import YOLO
import numpy as np
@ -18,8 +20,8 @@ class YOLOv8MFDModel:
conf: float = 0.25,
iou: float = 0.45,
):
self.model = YOLO(weight).to(device)
self.device = device
self.device = torch.device(device)
self.model = YOLO(weight).to(self.device)
self.imgsz = imgsz
self.conf = conf
self.iou = iou

View File

@ -23,12 +23,12 @@ class MathDataset(Dataset):
class UnimernetModel(object):
def __init__(self, weight_dir, _device_="cpu"):
from .unimernet_hf import UnimernetModel
if _device_.startswith("mps") or _device_.startswith("npu"):
if _device_.startswith("mps") or _device_.startswith("npu") or _device_.startswith("musa"):
self.model = UnimernetModel.from_pretrained(weight_dir, attn_implementation="eager")
else:
self.model = UnimernetModel.from_pretrained(weight_dir)
self.device = _device_
self.model.to(_device_)
self.device = torch.device(_device_)
self.model.to(self.device)
if not _device_.startswith("cpu"):
self.model = self.model.to(dtype=torch.float16)
self.model.eval()

View File

@ -4,6 +4,8 @@ import cv2
import numpy as np
from scipy.spatial import distance as dist
from skimage import measure
from skimage import __version__ as skimage_version
from packaging import version
def transform_preds(coords, center, scale, output_size, rot=0):
@ -295,7 +297,11 @@ def min_area_rect_box(
"""
boxes = []
for region in regions:
if region.bbox_area > H * W * 3 / 4: # 过滤大的单元格
if version.parse(skimage_version) >= version.parse("0.26.0"):
region_bbox_area = region.area_bbox
else:
region_bbox_area = region.bbox_area
if region_bbox_area > H * W * 3 / 4: # 过滤大的单元格
continue
rect = cv2.minAreaRect(region.coords[:, ::-1])

View File

@ -2,6 +2,7 @@ import os
import sys
from mineru.backend.vlm.utils import set_default_gpu_memory_utilization, enable_custom_logits_processors
from mineru.utils.config_reader import get_device
from mineru.utils.models_download_utils import auto_download_and_get_model_root_path
from vllm.entrypoints.cli.main import main as vllm_main
@ -13,6 +14,8 @@ def main():
has_port_arg = False
has_gpu_memory_utilization_arg = False
has_logits_processors_arg = False
has_block_size_arg = False
has_compilation_config = False
model_path = None
model_arg_indices = []
@ -24,6 +27,10 @@ def main():
has_gpu_memory_utilization_arg = True
if arg == "--logits-processors" or arg.startswith("--logits-processors="):
has_logits_processors_arg = True
if arg == "--block-size" or arg.startswith("--block-size="):
has_block_size_arg = True
if arg == "--compilation-config" or arg.startswith("--compilation-config="):
has_compilation_config = True
if arg == "--model":
if i + 1 < len(args):
model_path = args[i + 1]
@ -50,6 +57,22 @@ def main():
if (not has_logits_processors_arg) and custom_logits_processors:
args.extend(["--logits-processors", "mineru_vl_utils:MinerULogitsProcessor"])
# musa vllm v1 引擎特殊配置
# device = get_device()
# if device.startswith("musa"):
# import torch
# if torch.musa.is_available():
# if not has_block_size_arg:
# args.extend(["--block-size", "32"])
# if not has_compilation_config:
# args.extend(["--compilation-config", '{"cudagraph_capture_sizes": [1,2,3,4,5,6,7,8,10,12,14,16,18,20,24,28,30], "simple_cuda_graph": true}'])
# corex vllm v1 引擎特殊配置
device_type = os.getenv("MINERU_LMDEPLOY_DEVICE", "")
if device_type.lower() == "corex":
if not has_compilation_config:
args.extend(["--compilation-config", '{"cudagraph_mode": "FULL_DECODE_ONLY", "level": 0}'])
# 重构参数,将模型路径作为位置参数
sys.argv = [sys.argv[0]] + ["serve", model_path] + args

View File

@ -186,6 +186,22 @@ def model_init(model_name: str):
bf_16_support = True
elif device_name.startswith("mps"):
bf_16_support = True
elif device_name.startswith("gcu"):
if hasattr(torch, 'gcu') and torch.gcu.is_available():
if torch.gcu.is_bf16_supported():
bf_16_support = True
elif device_name.startswith("musa"):
if hasattr(torch, 'musa') and torch.musa.is_available():
if torch.musa.is_bf16_supported():
bf_16_support = True
elif device_name.startswith("npu"):
if hasattr(torch, 'npu') and torch.npu.is_available():
if torch.npu.is_bf16_supported():
bf_16_support = True
elif device_name.startswith("mlu"):
if hasattr(torch, 'mlu') and torch.mlu.is_available():
if torch.mlu.is_bf16_supported():
bf_16_support = True
if model_name == 'layoutreader':
# 检测modelscope的缓存目录是否存在

View File

@ -86,7 +86,19 @@ def get_device():
if torch_npu.npu.is_available():
return "npu"
except Exception as e:
pass
try:
if torch.gcu.is_available():
return "gcu"
except Exception as e:
try:
if torch.musa.is_available():
return "musa"
except Exception as e:
try:
if torch.mlu.is_available():
return "mlu"
except Exception as e:
pass
return "cpu"

View File

@ -2,6 +2,8 @@
包含两个MagicModel类中重复使用的方法和逻辑
"""
from typing import List, Dict, Any, Callable
from loguru import logger
from mineru.utils.boxbase import bbox_distance, bbox_center_distance, is_in
@ -172,11 +174,15 @@ def tie_up_category_by_index(
get_subjects_func: Callable,
get_objects_func: Callable,
extract_subject_func: Callable = None,
extract_object_func: Callable = None
extract_object_func: Callable = None,
object_block_type: str = "object",
):
"""
基于index的类别关联方法,用于将主体对象与客体对象进行关联
客体优先匹配给index最接近的主体,index差值相同时使用bbox中心点距离作为tiebreaker
客体优先匹配给index最接近的主体,匹配优先级为:
1. index差值(最高优先级)
2. bbox边缘距离(相邻边距离)
3. bbox中心点距离(最低优先级,作为最终tiebreaker)
参数:
get_subjects_func: 函数,提取主体对象
@ -207,6 +213,29 @@ def tie_up_category_by_index(
"sub_idx": i,
}
# 提取所有客体的index集合,用于计算有效index差值
object_indices = set(obj["index"] for obj in objects)
def calc_effective_index_diff(obj_index: int, sub_index: int) -> int:
"""
计算有效的index差值
有效差值 = 绝对差值 - 区间内其他客体的数量
即:如果obj_index和sub_index之间的差值是由其他客体造成的,则应该扣除这部分差值
"""
if obj_index == sub_index:
return 0
start, end = min(obj_index, sub_index), max(obj_index, sub_index)
abs_diff = end - start
# 计算区间(start, end)内有多少个其他客体的index
other_objects_count = 0
for idx in range(start + 1, end):
if idx in object_indices:
other_objects_count += 1
return abs_diff - other_objects_count
# 为每个客体找到最匹配的主体
for obj in objects:
if len(subjects) == 0:
@ -217,10 +246,10 @@ def tie_up_category_by_index(
min_index_diff = float("inf")
best_subject_indices = []
# 找出index差值最小的所有主体
# 找出有效index差值最小的所有主体
for i, subject in enumerate(subjects):
sub_index = subject["index"]
index_diff = abs(obj_index - sub_index)
index_diff = calc_effective_index_diff(obj_index, sub_index)
if index_diff < min_index_diff:
min_index_diff = index_diff
@ -228,18 +257,37 @@ def tie_up_category_by_index(
elif index_diff == min_index_diff:
best_subject_indices.append(i)
# 如果有多个主体的index差值相同,使用中心点距离作为tiebreaker
if len(best_subject_indices) > 1:
min_center_dist = float("inf")
if len(best_subject_indices) == 1:
best_subject_idx = best_subject_indices[0]
# 如果有多个主体的index差值相同(最多两个),根据边缘距离进行筛选
elif len(best_subject_indices) == 2:
# 计算所有候选主体的边缘距离
edge_distances = [(idx, bbox_distance(obj["bbox"], subjects[idx]["bbox"])) for idx in best_subject_indices]
edge_dist_diff = abs(edge_distances[0][1] - edge_distances[1][1])
for idx in best_subject_indices:
center_dist = bbox_center_distance(obj["bbox"], subjects[idx]["bbox"])
if center_dist < min_center_dist:
min_center_dist = center_dist
best_subject_idx = idx
for idx, edge_dist in edge_distances:
logger.debug(f"Obj index: {obj_index}, Sub index: {subjects[idx]['index']}, Edge distance: {edge_dist}")
if edge_dist_diff > 2:
# 边缘距离差值大于2,匹配边缘距离更小的主体
best_subject_idx = min(edge_distances, key=lambda x: x[1])[0]
logger.debug(f"Obj index: {obj_index}, edge_dist_diff > 2, matching to subject with min edge distance, index: {subjects[best_subject_idx]['index']}")
elif object_block_type == "table_caption":
# 边缘距离差值<=2且为table_caption,匹配index更大的主体
best_subject_idx = max(best_subject_indices, key=lambda idx: subjects[idx]["index"])
logger.debug(f"Obj index: {obj_index}, edge_dist_diff <= 2 and table_caption, matching to later subject with index: {subjects[best_subject_idx]['index']}")
elif object_block_type.endswith("footnote"):
# 边缘距离差值<=2且为footnote,匹配index更小的主体
best_subject_idx = min(best_subject_indices, key=lambda idx: subjects[idx]["index"])
logger.debug(f"Obj index: {obj_index}, edge_dist_diff <= 2 and footnote, matching to earlier subject with index: {subjects[best_subject_idx]['index']}")
else:
# 边缘距离差值<=2 且不适用特殊匹配规则,使用中心点距离匹配
center_distances = [(idx, bbox_center_distance(obj["bbox"], subjects[idx]["bbox"])) for idx in best_subject_indices]
for idx, center_dist in center_distances:
logger.debug(f"Obj index: {obj_index}, Sub index: {subjects[idx]['index']}, Center distance: {center_dist}")
best_subject_idx = min(center_distances, key=lambda x: x[1])[0]
else:
best_subject_idx = best_subject_indices[0]
raise ValueError("More than two subjects have the same minimal index difference, which is unexpected.")
# 将客体添加到最佳主体的obj_bboxes中
result_dict[best_subject_idx]["obj_bboxes"].append(extract_object_func(obj))

View File

@ -414,7 +414,7 @@ def get_res_list_from_layout_res(layout_res, iou_threshold=0.7, overlap_threshol
def clean_memory(device='cuda'):
if device == 'cuda':
if str(device).startswith("cuda"):
if torch.cuda.is_available():
torch.cuda.empty_cache()
torch.cuda.ipc_collect()
@ -423,6 +423,15 @@ def clean_memory(device='cuda'):
torch_npu.npu.empty_cache()
elif str(device).startswith("mps"):
torch.mps.empty_cache()
elif str(device).startswith("gcu"):
if torch.gcu.is_available():
torch.gcu.empty_cache()
elif str(device).startswith("musa"):
if torch.musa.is_available():
torch.musa.empty_cache()
elif str(device).startswith("mlu"):
if torch.mlu.is_available():
torch.mlu.empty_cache()
gc.collect()
@ -458,5 +467,14 @@ def get_vram(device) -> int:
elif str(device).startswith("npu"):
if torch_npu.npu.is_available():
total_memory = round(torch_npu.npu.get_device_properties(device).total_memory / (1024 ** 3)) # 转为 GB
elif str(device).startswith("gcu"):
if torch.gcu.is_available():
total_memory = round(torch.gcu.get_device_properties(device).total_memory / (1024 ** 3)) # 转为 GB
elif str(device).startswith("musa"):
if torch.musa.is_available():
total_memory = round(torch.musa.get_device_properties(device).total_memory / (1024 ** 3)) # 转为 GB
elif str(device).startswith("mlu"):
if torch.mlu.is_available():
total_memory = round(torch.mlu.get_device_properties(device).total_memory / (1024 ** 3)) # 转为 GB
return total_memory

View File

@ -17,6 +17,7 @@ CONTINUATION_END_MARKERS = [
"(cont.)",
"(cont’d)",
"(…continued)",
"续表",
]
CONTINUATION_INLINE_MARKERS = [
@ -69,6 +70,69 @@ def calculate_table_total_columns(soup):
return max_cols
def build_table_occupied_matrix(soup):
"""构建表格的占用矩阵,返回每行的有效列数
Args:
soup: BeautifulSoup解析的表格
Returns:
dict: {row_idx: effective_columns} 每行的有效列数(考虑rowspan占用)
"""
rows = soup.find_all("tr")
if not rows:
return {}
occupied = {} # {row_idx: {col_idx: True}}
row_effective_cols = {} # {row_idx: effective_columns}
for row_idx, row in enumerate(rows):
col_idx = 0
cells = row.find_all(["td", "th"])
if row_idx not in occupied:
occupied[row_idx] = {}
for cell in cells:
# 找到下一个未被占用的列位置
while col_idx in occupied[row_idx]:
col_idx += 1
colspan = int(cell.get("colspan", 1))
rowspan = int(cell.get("rowspan", 1))
# 标记被这个单元格占用的所有位置
for r in range(row_idx, row_idx + rowspan):
if r not in occupied:
occupied[r] = {}
for c in range(col_idx, col_idx + colspan):
occupied[r][c] = True
col_idx += colspan
# 该行的有效列数为已占用的最大列索引+1
if occupied[row_idx]:
row_effective_cols[row_idx] = max(occupied[row_idx].keys()) + 1
else:
row_effective_cols[row_idx] = 0
return row_effective_cols
def calculate_row_effective_columns(soup, row_idx):
"""计算指定行的有效列数(考虑rowspan占用)
Args:
soup: BeautifulSoup解析的表格
row_idx: 行索引
Returns:
int: 该行的有效列数
"""
row_effective_cols = build_table_occupied_matrix(soup)
return row_effective_cols.get(row_idx, 0)
def calculate_row_columns(row):
"""
计算表格行的实际列数,考虑colspan属性
@ -118,6 +182,10 @@ def detect_table_headers(soup1, soup2, max_header_rows=5):
rows1 = soup1.find_all("tr")
rows2 = soup2.find_all("tr")
# 构建两个表格的有效列数矩阵
effective_cols1 = build_table_occupied_matrix(soup1)
effective_cols2 = build_table_occupied_matrix(soup2)
min_rows = min(len(rows1), len(rows2), max_header_rows)
header_rows = 0
headers_match = True
@ -135,20 +203,24 @@ def detect_table_headers(soup1, soup2, max_header_rows=5):
if len(cells1) != len(cells2):
structure_match = False
else:
# 然后检查单元格的属性和内容
for cell1, cell2 in zip(cells1, cells2):
colspan1 = int(cell1.get("colspan", 1))
rowspan1 = int(cell1.get("rowspan", 1))
colspan2 = int(cell2.get("colspan", 1))
rowspan2 = int(cell2.get("rowspan", 1))
# 检查有效列数是否一致(考虑rowspan影响)
if effective_cols1.get(i, 0) != effective_cols2.get(i, 0):
structure_match = False
else:
# 然后检查单元格的属性和内容
for cell1, cell2 in zip(cells1, cells2):
colspan1 = int(cell1.get("colspan", 1))
rowspan1 = int(cell1.get("rowspan", 1))
colspan2 = int(cell2.get("colspan", 1))
rowspan2 = int(cell2.get("rowspan", 1))
# 去除所有空白字符(包括空格、换行、制表符等)
text1 = ''.join(full_to_half(cell1.get_text()).split())
text2 = ''.join(full_to_half(cell2.get_text()).split())
# 去除所有空白字符(包括空格、换行、制表符等)
text1 = ''.join(full_to_half(cell1.get_text()).split())
text2 = ''.join(full_to_half(cell2.get_text()).split())
if colspan1 != colspan2 or rowspan1 != rowspan2 or text1 != text2:
structure_match = False
break
if colspan1 != colspan2 or rowspan1 != rowspan2 or text1 != text2:
structure_match = False
break
if structure_match:
header_rows += 1
@ -158,7 +230,54 @@ def detect_table_headers(soup1, soup2, max_header_rows=5):
headers_match = header_rows > 0 # 只有当至少匹配了一行时,才认为表头匹配
break
# 如果没有找到匹配的表头行,则返回失败
# 如果严格匹配失败,尝试视觉一致性匹配(只比较文本内容)
if header_rows == 0:
header_rows, headers_match, header_texts = _detect_table_headers_visual(soup1, soup2, rows1, rows2, max_header_rows)
return header_rows, headers_match, header_texts
def _detect_table_headers_visual(soup1, soup2, rows1, rows2, max_header_rows=5):
"""
基于视觉一致性检测表头(只比较文本内容,忽略colspan/rowspan差异)
Args:
soup1: 第一个表格的BeautifulSoup对象
soup2: 第二个表格的BeautifulSoup对象
rows1: 第一个表格的行列表
rows2: 第二个表格的行列表
max_header_rows: 最大可能的表头行数
Returns:
tuple: (表头行数, 表头是否一致, 表头文本列表)
"""
# 构建两个表格的有效列数矩阵
effective_cols1 = build_table_occupied_matrix(soup1)
effective_cols2 = build_table_occupied_matrix(soup2)
min_rows = min(len(rows1), len(rows2), max_header_rows)
header_rows = 0
headers_match = True
header_texts = []
for i in range(min_rows):
cells1 = rows1[i].find_all(["td", "th"])
cells2 = rows2[i].find_all(["td", "th"])
# 提取每行的文本内容列表(去除空白字符)
texts1 = [''.join(full_to_half(cell.get_text()).split()) for cell in cells1]
texts2 = [''.join(full_to_half(cell.get_text()).split()) for cell in cells2]
# 检查视觉一致性:文本内容完全相同,且有效列数一致
effective_cols_match = effective_cols1.get(i, 0) == effective_cols2.get(i, 0)
if texts1 == texts2 and effective_cols_match:
header_rows += 1
row_texts = [full_to_half(cell.get_text().strip()) for cell in cells1]
header_texts.append(row_texts)
else:
headers_match = header_rows > 0
break
if header_rows == 0:
headers_match = False
@ -243,34 +362,44 @@ def check_rows_match(soup1, soup2):
if not (rows1 and rows2):
return False
# 获取第一个表的最后一行数据行
# 获取第一个表的最后一行数据行索引
last_row_idx = None
last_row = None
for row in reversed(rows1):
if row.find_all(["td", "th"]):
last_row = row
for idx in range(len(rows1) - 1, -1, -1):
if rows1[idx].find_all(["td", "th"]):
last_row_idx = idx
last_row = rows1[idx]
break
# 检测表头行数,以便获取第二个表的首个数据行
header_count, _, _ = detect_table_headers(soup1, soup2)
# 获取第二个表的首个数据行
first_data_row_idx = None
first_data_row = None
if len(rows2) > header_count:
first_data_row_idx = header_count
first_data_row = rows2[header_count] # 第一个非表头行
if not (last_row and first_data_row):
return False
# 计算实际列数(考虑colspan)和视觉列数
# 计算有效列数(考虑rowspan和colspan)
last_row_effective_cols = calculate_row_effective_columns(soup1, last_row_idx)
first_row_effective_cols = calculate_row_effective_columns(soup2, first_data_row_idx)
# 计算实际列数(仅考虑colspan)和视觉列数
last_row_cols = calculate_row_columns(last_row)
first_row_cols = calculate_row_columns(first_data_row)
last_row_visual_cols = calculate_visual_columns(last_row)
first_row_visual_cols = calculate_visual_columns(first_data_row)
# logger.debug(f"行列数 - 前表最后一行: {last_row_cols}(视觉列数:{last_row_visual_cols}), 当前表首行: {first_row_cols}(视觉列数:{first_row_visual_cols})")
# logger.debug(f"行列数 - 前表最后一行: {last_row_cols}(有效列数:{last_row_effective_cols}, 视觉列数:{last_row_visual_cols}), 当前表首行: {first_row_cols}(有效列数:{first_row_effective_cols}, 视觉列数:{first_row_visual_cols})")
# 同时考虑实际列数匹配和视觉列数匹配
return last_row_cols == first_row_cols or last_row_visual_cols == first_row_visual_cols
# 同时考虑有效列数匹配、实际列数匹配和视觉列数匹配
return (last_row_effective_cols == first_row_effective_cols or
last_row_cols == first_row_cols or
last_row_visual_cols == first_row_visual_cols)
def check_row_columns_match(row1, row2):
@ -287,12 +416,13 @@ def check_row_columns_match(row1, row2):
return True
def adjust_table_rows_colspan(rows, start_idx, end_idx,
def adjust_table_rows_colspan(soup, rows, start_idx, end_idx,
reference_structure, reference_visual_cols,
target_cols, current_cols, reference_row):
"""调整表格行的colspan属性以匹配目标列数
Args:
soup: BeautifulSoup解析的表格对象(用于计算有效列数)
rows: 表格行列表
start_idx: 起始行索引
end_idx: 结束行索引(不包含)
@ -304,14 +434,21 @@ def adjust_table_rows_colspan(rows, start_idx, end_idx,
"""
reference_row_copy = deepcopy(reference_row)
# 构建有效列数矩阵
effective_cols_matrix = build_table_occupied_matrix(soup)
for i in range(start_idx, end_idx):
row = rows[i]
cells = row.find_all(["td", "th"])
if not cells:
continue
# 使用有效列数(考虑rowspan)判断是否需要调整
current_row_effective_cols = effective_cols_matrix.get(i, 0)
current_row_cols = calculate_row_columns(row)
if current_row_cols >= target_cols:
# 如果有效列数或实际列数已经达到目标,则跳过
if current_row_effective_cols >= target_cols or current_row_cols >= target_cols:
continue
# 检查是否与参考行结构匹配
@ -323,9 +460,12 @@ def adjust_table_rows_colspan(rows, start_idx, end_idx,
cell["colspan"] = str(reference_structure[j])
else:
# 扩展最后一个单元格以填补列数差异
last_cell = cells[-1]
current_last_span = int(last_cell.get("colspan", 1))
last_cell["colspan"] = str(current_last_span + (target_cols - current_cols))
# 使用有效列数来计算差异
cols_diff = target_cols - current_row_effective_cols
if cols_diff > 0:
last_cell = cells[-1]
current_last_span = int(last_cell.get("colspan", 1))
last_cell["colspan"] = str(current_last_span + cols_diff)
def perform_table_merge(soup1, soup2, previous_table_block, wait_merge_table_footnotes):
@ -356,7 +496,7 @@ def perform_table_merge(soup1, soup2, previous_table_block, wait_merge_table_foo
reference_visual_cols = calculate_visual_columns(last_row1)
# 以表1的最后一行为参考,调整表2的行
adjust_table_rows_colspan(
rows2, header_count, len(rows2),
soup2, rows2, header_count, len(rows2),
reference_structure, reference_visual_cols,
table_cols1, table_cols2, first_data_row2
)
@ -366,7 +506,7 @@ def perform_table_merge(soup1, soup2, previous_table_block, wait_merge_table_foo
reference_visual_cols = calculate_visual_columns(first_data_row2)
# 以表2的第一个数据行为参考,调整表1的行
adjust_table_rows_colspan(
rows1, 0, len(rows1),
soup1, rows1, 0, len(rows1),
reference_structure, reference_visual_cols,
table_cols2, table_cols1, last_row1
)
@ -445,4 +585,4 @@ def merge_table(page_info_list):
# 删除当前页的table
for block in current_table_block["blocks"]:
block['lines'] = []
block[SplitFlag.LINES_DELETED] = True
block[SplitFlag.LINES_DELETED] = True

View File

@ -1 +1 @@
__version__ = "2.7.1"
__version__ = "2.7.4"

View File

@ -7,7 +7,7 @@ name = "mineru"
dynamic = ["version"]
license = { text = "AGPL-3.0" }
description = "A practical tool for converting PDF to Markdown"
readme = "README.md"
readme = "README.mineru-upstream.md"
requires-python = ">=3.10,<3.14"
keywords = ["magic-pdf", "mineru", "MinerU", "convert", "pdf", "markdown"]
classifiers = [

View File

@ -1,34 +1,9 @@
#!/usr/bin/env bash
set -euo pipefail
VERSION=${VERSION:-$(date +%Y%m%d%H%M%S)}
API_FLAVOR=${API_FLAVOR:-corex}
ROOT_DIR=$(cd "$(dirname "$0")/.." && pwd)
cd "$ROOT_DIR"
case "$API_FLAVOR" in
corex)
API_DOCKERFILE=docker/china/Dockerfile.corex
;;
cuda)
API_DOCKERFILE=docker/global/Dockerfile.cuda
;;
*)
echo "Unsupported API_FLAVOR: $API_FLAVOR. Use corex or cuda." >&2
exit 1
;;
case "${API_FLAVOR:-corex}" in
corex) FILE=docker/china/Dockerfile.corex ;;
cuda) FILE=docker/global/Dockerfile.cuda ;;
*) echo "API_FLAVOR must be corex or cuda" >&2; exit 2 ;;
esac
echo "Building UnisMindMap API image (${API_FLAVOR}): unis-mindmap-api:${VERSION}"
sudo docker build -f "$API_DOCKERFILE" \
-t unis-mindmap-api:${VERSION} \
-t unis-mindmap-api:${API_FLAVOR} \
-t unis-mindmap-api:latest \
.
echo "Building UnisMindMap Web image: unis-mindmap-web:${VERSION}"
sudo docker build -f web_ui/Dockerfile.portable \
-t unis-mindmap-web:${VERSION} \
-t unis-mindmap-web:latest \
web_ui
exec docker build -f "$ROOT_DIR/$FILE" -t "unis-mineru:${VERSION:-2.7.4}" "$ROOT_DIR"

View File

@ -0,0 +1,30 @@
"""Start the pinned upstream parser; application features live in MaxKB."""
import os
from pathlib import Path
import sys
ROOT = Path(__file__).resolve().parents[1]
sys.path.insert(0, str(ROOT))
def main():
os.environ.setdefault('MINERU_MODEL_SOURCE', 'local')
os.environ.setdefault('MINERU_API_MAX_CONCURRENT_REQUESTS', '1')
from mineru.version import __version__
from mineru import __file__ as module_path
if __version__ != '2.7.4' or not Path(module_path).resolve().is_relative_to(ROOT / 'mineru'):
raise RuntimeError('解析服务版本或源码路径不匹配')
if not any(arg.startswith('--mineru-api-max-concurrent-requests') for arg in sys.argv):
sys.argv += ['--mineru-api-max-concurrent-requests', os.environ['MINERU_API_MAX_CONCURRENT_REQUESTS']]
memory = os.getenv('MINERU_GPU_MEMORY_UTILIZATION')
if memory:
if not 0 < float(memory) < 1:
raise ValueError('MINERU_GPU_MEMORY_UTILIZATION must be between 0 and 1')
if not any(arg.startswith('--gpu-memory-utilization') for arg in sys.argv):
sys.argv += ['--gpu-memory-utilization', str(float(memory))]
from mineru.cli.fast_api import main as serve
serve()
if __name__ == '__main__':
main()

View File

@ -1,33 +1,5 @@
#!/usr/bin/env bash
set -euo pipefail
NETWORK=${NETWORK:-unis-ai}
API_NAME=${API_NAME:-mineru-api}
WEB_NAME=${WEB_NAME:-mindmap-web}
WEB_PORT=${WEB_PORT:-9898}
MINERU_API_URL=${MINERU_API_URL:-http://${API_NAME}:8000}
MINERU_GPU_MEMORY_UTILIZATION=${MINERU_GPU_MEMORY_UTILIZATION:-0.5}
MINERU_MODEL_DIR=${MINERU_MODEL_DIR:-/data/unis/mineru-models}
MINERU_PYTHONPATH=${MINERU_PYTHONPATH:-/app:/usr/local/corex-4.4.0/lib64/python3/dist-packages}
sudo docker network create "$NETWORK" >/dev/null 2>&1 || true
sudo docker rm -f "$API_NAME" "$WEB_NAME" >/dev/null 2>&1 || true
sudo mkdir -p "$MINERU_MODEL_DIR"/pipeline "$MINERU_MODEL_DIR"/vlm
sudo chmod -R 777 "$MINERU_MODEL_DIR"
sudo docker run -d --name "$API_NAME" --network "$NETWORK" \
--privileged --ipc=host --shm-size=32g \
-v "${MINERU_MODEL_DIR}:/models" \
-e MINERU_MODEL_SOURCE="${MINERU_MODEL_SOURCE:-local}" \
-e MINERU_MODEL_DIR=/models \
-e MINERU_PIPELINE_MODEL_DIR=/models/pipeline \
-e MINERU_VLM_MODEL_DIR=/models/vlm \
-e PYTHONPATH="$MINERU_PYTHONPATH" \
-e MINERU_GPU_MEMORY_UTILIZATION="$MINERU_GPU_MEMORY_UTILIZATION" \
unis-mindmap-api:latest \
mineru-api --host 0.0.0.0 --port 8000
sudo docker run -d --name "$WEB_NAME" --network "$NETWORK" \
-p "${WEB_PORT}:9898" \
-e MINERU_API_URL="$MINERU_API_URL" \
unis-mindmap-web:latest
# Compatibility entrypoint: starts only the parser, never the retired workbench.
ROOT_DIR=$(cd "$(dirname "$0")/.." && pwd)
exec docker compose -f "$ROOT_DIR/docker/compose.yaml" up -d --no-deps mineru-api

View File

View File

@ -1,36 +0,0 @@
# 构建阶段
FROM node:18-alpine AS builder
# 设置工作目录
WORKDIR /app
# 使用国内镜像源加速
RUN sed -i 's/dl-cdn.alpinelinux.org/mirrors.aliyun.com/g' /etc/apk/repositories
# 复制 package.json 和 package-lock.json
COPY package*.json ./
# 安装依赖(使用 npm 淘宝镜像)
RUN npm config set registry https://registry.npmmirror.com && \
npm install
# 复制源代码
COPY . .
# 构建前端
RUN npm run build
# 运行阶段
FROM nginx:alpine
# 复制构建结果到 Nginx 静态目录
COPY --from=builder /app/dist /usr/share/nginx/html
# 复制 Nginx 配置
COPY nginx.conf /etc/nginx/conf.d/default.conf
# 暴露端口
EXPOSE 9898
# 启动 Nginx
CMD ["nginx", "-g", "daemon off;"]

View File

@ -1,17 +0,0 @@
# Portable Web UI image for UnisMindMap.
# Runtime API upstream is configured by MINERU_API_URL, defaulting to Docker DNS name mineru-api.
FROM node:18-alpine AS builder
WORKDIR /app
RUN sed -i 's/dl-cdn.alpinelinux.org/mirrors.aliyun.com/g' /etc/apk/repositories
COPY package*.json ./
RUN npm config set registry https://registry.npmmirror.com && npm install
COPY . .
RUN npm run build
FROM nginx:alpine
ENV MINERU_API_URL=http://mineru-api:8000
COPY --from=builder /app/dist /usr/share/nginx/html
COPY nginx.conf.template /etc/nginx/templates/default.conf.template
EXPOSE 9898
CMD ["nginx", "-g", "daemon off;"]

View File

@ -1,169 +0,0 @@
# MinerU Vue.js Web UI
基于 Vue 3 + TypeScript + Element Plus 的现代化 Web 界面,用于替代原有的 Gradio 界面。
## 🌟 特性
- ✨ 响应式设计,支持移动端
- 🌍 国际化支持(中英文)
- 📁 文件拖拽上传
- ⚙️ 丰富的配置选项
- 📊 Markdown 渲染和源码查看
- 🧠 思维导图可视化(开发中)
- 🔗 与 FastAPI 后端无缝集成
## 🚀 快速开始
### 环境要求
- Node.js >= 16.0.0
- npm >= 8.0.0
### 安装依赖
```bash
cd web_ui
npm install
```
### 启动开发服务器
```bash
# 方式1:使用 npm
npm run dev
# 方式2:使用启动脚本
./start.sh
```
### 构建生产版本
```bash
npm run build
```
## 🏗️ 项目结构
```
web_ui/
├── src/
│ ├── components/ # 可复用组件
│ │ ├── FileUploader.vue # 文件上传组件
│ │ ├── ConfigPanel.vue # 配置面板组件
│ │ ├── ResultPanel.vue # 结果展示组件
│ │ ├── MarkdownRenderer.vue # Markdown 渲染器
│ │ └── MindMapRenderer.vue # 思维导图渲染器
│ ├── views/ # 页面视图
│ │ └── DocumentProcessor.vue # 主处理器页面
│ ├── composables/ # Vue 组合式函数
│ │ └── useDocumentProcessor.ts # 文档处理逻辑
│ ├── locales/ # 国际化文件
│ │ ├── zh.ts # 中文翻译
│ │ ├── en.ts # 英文翻译
│ │ └── index.ts # 导出文件
│ ├── api/ # API 接口
│ │ └── document.ts # 文档处理 API
│ ├── utils/ # 工具函数
│ │ └── request.ts # HTTP 请求封装
│ ├── App.vue # 根组件
│ └── main.ts # 入口文件
├── public/ # 静态资源
├── index.html # HTML 模板
├── vite.config.ts # Vite 配置
├── tsconfig.json # TypeScript 配置
├── package.json # 项目依赖
└── README.md # 本文档
```
## 🔧 技术栈
- **Vue 3** - 渐进式 JavaScript 框架
- **TypeScript** - JavaScript 的超集,提供类型安全
- **Element Plus** - Vue 3 组件库
- **Vite** - 新一代构建工具
- **Axios** - HTTP 客户端
- **Vue I18n** - 国际化解决方案
- **Markdown-it** - Markdown 解析器
## 🔄 与原 Gradio 界面的区别
| 特性 | 原 Gradio | Vue.js 版本 |
|------|-----------|-------------|
| 用户体验 | 基础交互 | 现代化 UI 设计 |
| 响应式 | 有限支持 | 完全响应式 |
| 国际化 | 有限支持 | 完善的多语言支持 |
| 组件化 | 不支持 | 高度组件化 |
| 扩展性 | 较差 | 易于扩展和维护 |
| 移动端适配 | 不支持 | 响应式设计 |
## 📋 功能对比
### ✅ 已实现功能
- [x] 文件上传(PDF/图片)
- [x] 参数配置面板
- [x] 后端选择(pipeline/VLM/hybrid)
- [x] OCR 语言选择
- [x] 表格和公式识别选项
- [x] Markdown 渲染展示
- [x] Markdown 源码查看
- [x] 结果下载
- [x] 中英文国际化
- [x] 错误处理和提示
### ⏳ 待完善功能
- [ ] 交互式思维导图
- [ ] 批量处理支持
- [ ] 处理进度显示
- [ ] 历史记录管理
- [ ] 用户偏好设置保存
- [ ] 更多主题样式
## 🛠️ 开发指南
### 添加新组件
1. 在 `src/components/` 目录下创建新的 `.vue` 文件
2. 使用 Composition API 编写组件逻辑
3. 在需要的地方导入和使用组件
### 添加国际化文本
1. 在 `src/locales/zh.ts` 和 `src/locales/en.ts` 中添加对应的翻译
2. 在组件中使用 `$t('key')` 来引用翻译文本
### API 接口扩展
1. 在 `src/api/` 目录下创建新的 API 文件
2. 使用 `request.ts` 封装的 axios 实例
3. 在 `composables` 中调用 API
## 🐛 常见问题
### 1. 启动时报错 "Port 3000 is in use"
这是正常的,Vite 会自动寻找可用端口。通常会在 3001、3002 等端口启动。
### 2. 无法连接到后端 API
确保 FastAPI 服务已在 8000 端口启动:
```bash
mineru-api --host localhost --port 8000
```
### 3. 构建失败
尝试清理缓存并重新安装依赖:
```bash
rm -rf node_modules package-lock.json
npm install
npm run build
```
## 📄 许可证
本项目遵循 AGPL-3.0 许可证。
## 🤝 贡献
欢迎提交 Issue 和 Pull Request!
---
**注意**:此 Vue.js 版本目前仍在开发中,部分功能可能不如原 Gradio 版本稳定。建议在生产环境中继续使用原版 Gradio 界面。

View File

@ -1,25 +0,0 @@
<!DOCTYPE html>
<html lang="zh-CN">
<head>
<meta charset="UTF-8">
<link rel="icon" href="/favicon.ico">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>思维导图</title>
<style>
html, body {
height: 100%;
margin: 0;
padding: 0;
box-sizing: border-box;
}
#app {
height: 100%;
}
</style>
</head>
<body>
<div id="app"></div>
<script type="module" src="/src/main.ts"></script>
</body>
</html>

View File

@ -1,47 +0,0 @@
cat > /etc/nginx/conf.d/default.conf << 'EOF'
server {
listen 9898;
server_name localhost;
root /usr/share/nginx/html;
index index.html;
access_log /var/log/nginx/access.log;
error_log /var/log/nginx/error.log;
location ~* \.(jpg|jpeg|png|gif|ico|css|js|svg|woff|woff2|ttf|eot)$ {
expires 1y;
add_header Cache-Control "public, immutable";
}
location / {
try_files $uri $uri/ /index.html;
}
# 【新增】直接代理 /file_parse 请求(无前缀)
location = /file_parse {
proxy_pass http://localhost:8000/file_parse;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto $scheme;
client_max_body_size 100M;
proxy_connect_timeout 300s;
proxy_send_timeout 300s;
proxy_read_timeout 300s;
}
# 【保留】也支持带 /api 前缀的请求
location /api/ {
proxy_pass http://localhost:8000/;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto $scheme;
client_max_body_size 100M;
proxy_connect_timeout 300s;
proxy_send_timeout 300s;
proxy_read_timeout 300s;
}
}
EOF

View File

@ -1,45 +0,0 @@
server {
listen 9898;
server_name localhost;
root /usr/share/nginx/html;
index index.html;
access_log /var/log/nginx/access.log;
error_log /var/log/nginx/error.log;
location ~* \.(jpg|jpeg|png|gif|ico|css|js|svg|woff|woff2|ttf|eot)$ {
expires 1y;
add_header Cache-Control "public, immutable";
}
location / {
try_files $uri $uri/ /index.html;
}
# 【新增】直接代理 /file_parse 请求(无前缀)
location = /file_parse {
proxy_pass http://localhost:8000/file_parse;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto $scheme;
client_max_body_size 100M;
proxy_connect_timeout 300s;
proxy_send_timeout 300s;
proxy_read_timeout 300s;
}
# 【保留】也支持带 /api 前缀的请求
location /api/ {
proxy_pass http://localhost:8000/;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto $scheme;
client_max_body_size 100M;
proxy_connect_timeout 300s;
proxy_send_timeout 300s;
proxy_read_timeout 300s;
}
}

View File

@ -1,83 +0,0 @@
server {
listen 9898;
server_name localhost;
root /usr/share/nginx/html;
index index.html;
access_log /var/log/nginx/access.log;
error_log /var/log/nginx/error.log;
location ~* ^/(assets|mindmap/assets)/.*\.(jpg|jpeg|png|gif|ico|css|js|svg|woff|woff2|ttf|eot)$ {
rewrite ^/mindmap/(.*)$ /$1 break;
expires 1y;
add_header Cache-Control "public, immutable";
try_files $uri =404;
}
location = /mindmap {
return 301 /mindmap/;
}
location = /mindmap/api/health {
proxy_pass ${MINERU_API_URL}/health;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto $scheme;
}
location = /api/health {
proxy_pass ${MINERU_API_URL}/health;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto $scheme;
}
location ^~ /mindmap/api/ {
rewrite ^/mindmap/api/(.*)$ /api/$1 break;
proxy_pass ${MINERU_API_URL};
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto $scheme;
client_max_body_size 100M;
proxy_connect_timeout 300s;
proxy_send_timeout 300s;
proxy_read_timeout 300s;
}
location = /file_parse {
proxy_pass ${MINERU_API_URL}/file_parse;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto $scheme;
client_max_body_size 100M;
proxy_connect_timeout 300s;
proxy_send_timeout 300s;
proxy_read_timeout 300s;
}
location /api/ {
proxy_pass ${MINERU_API_URL}/api/;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto $scheme;
client_max_body_size 100M;
proxy_connect_timeout 300s;
proxy_send_timeout 300s;
proxy_read_timeout 300s;
}
location ^~ /mindmap/ {
rewrite ^/mindmap/(.*)$ /$1 break;
try_files $uri $uri/ /index.html;
}
location / {
try_files $uri $uri/ /index.html;
}
}

3392
web_ui/package-lock.json generated

File diff suppressed because it is too large Load Diff

View File

@ -1,31 +0,0 @@
{
"name": "mineru-web-ui",
"version": "1.0.0",
"private": true,
"scripts": {
"dev": "vite",
"build": "vite build",
"preview": "vite preview",
"lint": "eslint . --ext .vue,.js,.jsx,.cjs,.mjs,.ts,.tsx,.cts,.mts --fix --ignore-path .gitignore"
},
"dependencies": {
"axios": "^1.6.0",
"build": "^0.1.4",
"d3": "^7.9.0",
"element-plus": "^2.4.0",
"handlebars": "^4.7.9",
"markdown-it": "^14.0.0",
"markmap-lib": "^0.18.12",
"markmap-toolbar": "^0.18.12",
"markmap-view": "^0.18.12",
"vue": "^3.4.0",
"vue-i18n": "^9.8.0"
},
"devDependencies": {
"@types/node": "^20.0.0",
"@vitejs/plugin-vue": "^5.0.0",
"typescript": "^5.2.0",
"vite": "^5.0.0",
"vue-tsc": "^1.8.0"
}
}

File diff suppressed because one or more lines are too long

File diff suppressed because one or more lines are too long

File diff suppressed because one or more lines are too long

File diff suppressed because one or more lines are too long

View File

@ -1,40 +0,0 @@
<template>
<div id="app">
<div class="container">
<DocumentProcessor />
</div>
</div>
</template>
<script setup lang="ts">
import DocumentProcessor from './views/DocumentProcessor.vue'
</script>
<style>
/* 全局样式 */
* {
margin: 0;
padding: 0;
box-sizing: border-box;
}
body {
font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, 'Helvetica Neue', Arial, sans-serif;
font-size: 14px;
line-height: 1.5;
color: #4E5969;
background-color: #F2F3F5;
}
#app {
height: 100%;
}
.container {
width: 100%;
height: 100%;
box-sizing: border-box;
padding: 0;
margin: 0;
}
</style>

View File

@ -1,81 +0,0 @@
import { hubCall } from './hub'
import { uuid } from '@/utils/uuid'
export interface ParseParams {
files: File[]
parser_resource_id: string
output_dir: string
lang_list: string
backend: string
parse_method: string
formula_enable: boolean
table_enable: boolean
server_url?: string
start_page_id: number
end_page_id: number
return_md: boolean
return_middle_json: boolean
response_format_zip: boolean
}
export interface ParseResult {
backend: string
version: string
results: Record<string, {
md_content?: string
middle_json?: string
model_output?: string
content_list?: string
images?: Record<string, string>
}>
}
export interface ParseProgress {
progress: number | null
stage: string
status: 'pending' | 'processing' | 'completed' | 'failed'
error: string | null
file_names: string
}
export interface MindmapOrganizeProgress extends ParseProgress {
result_md?: string | null
}
const ids = new Map<string, string>()
const states: Record<string, ParseProgress['status']> = { PENDING: 'pending', RUNNING: 'processing', SUCCESS: 'completed', FAILED: 'failed' }
async function progress(taskId: string): Promise<MindmapOrganizeProgress> {
const id = ids.get(taskId)
if (!id) return { progress: null, status: 'pending', stage: '准备中', error: null, file_names: '' }
const task = await hubCall('tasks', 'get', undefined, id)
return { progress: task.status === 'SUCCESS' ? 100 : task.progress?.percent ?? null, status: states[task.status], stage: task.progress?.stage || (task.status === 'PENDING' ? '等待处理' : task.status === 'RUNNING' ? '处理中' : task.status === 'SUCCESS' ? '完成' : '失败'), error: task.error, file_names: task.name, result_md: task.result?.markdown }
}
export const documentApi = {
async parseDocument(params: ParseParams, taskId: string = uuid()): Promise<ParseResult> {
if (!params.parser_resource_id) throw new Error('请选择解析服务')
const results: ParseResult['results'] = {}
for (const file of params.files) {
const task = await hubCall('tasks', 'post', { name: file.name.slice(0, 128), nodes: [{ type: 'parse', resource_id: params.parser_resource_id }, { type: 'render' }] }, undefined, file)
ids.set(taskId, task.id)
const until = Date.now() + 30 * 60 * 1000
let completed = false
while (Date.now() < until) {
const current = await hubCall('tasks', 'get', undefined, task.id)
if (current.status === 'FAILED') throw new Error(current.error || '解析失败')
if (current.status === 'SUCCESS') { results[file.name] = { md_content: current.result.markdown, middle_json: JSON.stringify(current.result.source) }; completed = true; break }
await new Promise(resolve => setTimeout(resolve, 1500))
}
if (!completed) throw new Error('任务仍在处理中,请到管理页查看')
}
return { backend: 'shared', version: '1', results }
},
async createParseTask(_taskId: string): Promise<ParseProgress> { return { progress: null, stage: '准备中', status: 'pending', error: null, file_names: '' } },
getParseProgress: progress,
async createMindmapTask(taskId: string, markdown: string, _mode = 'smart', prompt?: string, config?: { model_id: string; temperature: number }): Promise<MindmapOrganizeProgress> {
if (!config?.model_id) throw new Error('请选择整理模型')
const task = await hubCall('tasks', 'post', { name: '智能整理', markdown, nodes: [{ type: 'organize', model_id: config.model_id, prompt, temperature: config.temperature }, { type: 'render' }] })
ids.set(taskId, task.id)
return progress(taskId)
},
getMindmapProgress: progress
}

View File

@ -1,28 +0,0 @@
// The parent owns authentication. No token or provider secret is sent to this frame.
import { uuid } from '@/utils/uuid'
const pending = new Map<string, { resolve: (value: any) => void; reject: (error: Error) => void; timer: ReturnType<typeof setTimeout> }>()
const parentOrigin = (() => { try { return new URL(document.referrer).origin } catch { return '' } })()
window.addEventListener('message', event => {
if (event.source !== window.parent || event.origin !== parentOrigin || event.data?.type !== 'ai-hub-response') return
const call = pending.get(event.data.id)
if (!call) return
clearTimeout(call.timer); pending.delete(event.data.id)
if (event.data.error) call.reject(new Error(event.data.error)); else call.resolve(event.data.data)
})
export async function hubCall(kind: string, method = 'get', data?: any, recordId?: string, file?: File): Promise<any> {
if (window.parent === window || !parentOrigin) throw new Error('请从管理平台打开思维导图')
const id = uuid()
const upload = file ? { name: file.name, content: await file.arrayBuffer() } : undefined
return new Promise((resolve, reject) => {
const timer = setTimeout(() => { pending.delete(id); reject(new Error('请求超时')) }, 45000)
pending.set(id, { resolve, reject, timer })
window.parent.postMessage({ type: 'ai-hub-request', id, kind, method, data, recordId, file: upload }, parentOrigin)
})
}
export function onOpenArtifact(callback: (artifact: any) => void) {
const handler = (event: MessageEvent) => {
if (event.source === window.parent && event.origin === parentOrigin && event.data?.type === 'ai-hub-open') callback(event.data.artifact)
}
window.addEventListener('message', handler)
return () => window.removeEventListener('message', handler)
}

View File

@ -1,35 +0,0 @@
<template>
<div class="config-panel">
<el-form label-position="top">
<el-form-item label="解析服务"><el-select v-model="draft.parserResourceId" placeholder="请选择"><el-option v-for="p in resources.parsers" :key="p.id" :label="p.name" :value="p.id"/></el-select></el-form-item>
<el-form-item label="整理模型"><el-select v-model="draft.mindmapModelId" clearable placeholder="请选择"><el-option v-for="m in models" :key="m.id" :label="m.name" :value="m.id"/></el-select></el-form-item>
<el-form-item label="温度"><el-input-number v-model="draft.mindmapLlmTemperature" :min="0" :max="2" :step="0.1"/></el-form-item>
<el-form-item label="整理提示词"><el-input v-model="draft.mindmapPrompt" type="textarea" :rows="6" /></el-form-item>
<el-form-item label="导出长边"><el-select v-model="draft.exportResolution"><el-option label="1080P" :value="1080"/><el-option label="2K" :value="2560"/><el-option label="4K" :value="3840"/></el-select></el-form-item>
<div v-if="error" class="error">{{ error }}</div>
<div class="actions"><el-button @click="load">刷新资源</el-button><el-button type="primary" @click="confirm">确认</el-button></div>
</el-form>
</div>
</template>
<script setup lang="ts">
import { computed, onMounted, reactive, ref, watch } from 'vue'
import type { DocumentConfig } from '@/composables/useDocumentProcessor'
import { hubCall } from '@/api/hub'
const props = defineProps<{ modelValue: DocumentConfig; backendOptions: any[]; languageOptions: any[] }>()
const emit = defineEmits(['update:modelValue', 'backendChange', 'close'])
const draft = reactive({ ...props.modelValue })
const resources = ref<any>({ parsers: [], models: [] }), error = ref('')
const models = computed(() => resources.value.models.filter((m: any) => m.model_type === 'LLM'))
watch(() => props.modelValue, value => Object.assign(draft, value), { deep: true })
async function load() {
try {
resources.value = await hubCall('resources'); error.value = ''
if (!draft.parserResourceId && resources.value.parsers?.length) draft.parserResourceId = resources.value.parsers[0].id
if (!draft.mindmapModelId && models.value.length) draft.mindmapModelId = models.value[0].id
if (draft.parserResourceId || draft.mindmapModelId) emit('update:modelValue', { ...draft })
} catch (e: any) { error.value = e.message }
}
function confirm() { emit('update:modelValue', { ...draft }); emit('close') }
onMounted(load)
</script>
<style scoped>.config-panel { padding: 12px; } .el-select { width: 100%; } .actions { display: flex; justify-content: flex-end; gap: 8px; } .error { color: var(--el-color-danger); margin: 8px 0; }</style>

View File

@ -1,246 +0,0 @@
<template>
<div class="file-uploader">
<div
class="upload-container"
:class="{ 'drag-over': isDragging }"
@drop="handleDrop"
@dragover.prevent="isDragging = true"
@dragleave="isDragging = false"
@click="triggerUpload"
>
<div class="upload-content">
<el-icon class="upload-icon"><UploadFilled /></el-icon>
<div class="upload-text">
<div class="main-text">点击或拖拽文件到此处上传</div>
<div class="sub-text">支持 PDF、PNG、JPG、JPEG 格式,最大 100MB</div>
</div>
</div>
</div>
<!-- 隐藏的上传控件 -->
<el-upload
ref="uploadRef"
:auto-upload="false"
:show-file-list="false"
:on-change="handleFileChange"
:before-upload="beforeUpload"
multiple
accept=".pdf,.png,.jpg,.jpeg"
class="hidden-upload"
>
<el-button type="primary">上传文件</el-button>
</el-upload>
<!-- 已上传文件预览 -->
<div v-if="files.length > 0" class="file-preview">
<h3 class="preview-title">已上传文件</h3>
<div class="file-card" v-for="(file, index) in files" :key="index">
<div class="file-info">
<el-icon class="file-icon"><Document /></el-icon>
<span class="file-name">{{ file.name }}</span>
<span class="file-size">({{ formatFileSize(file.size) }})</span>
<el-button
type="danger"
:icon="Delete"
circle
size="small"
@click.stop="removeFile(index)"
class="remove-button"
/>
</div>
</div>
</div>
</div>
</template>
<script setup lang="ts">
import { ref, watch } from 'vue'
import { UploadFilled, Document, Delete, Picture } from '@element-plus/icons-vue'
import type { UploadFile, UploadInstance } from 'element-plus'
interface Props {
modelValue: File[]
}
interface Emits {
(e: 'update:modelValue', value: File[]): void
}
const props = defineProps<Props>()
const emit = defineEmits<Emits>()
const files = ref<File[]>(props.modelValue)
const uploadRef = ref<UploadInstance>()
const isDragging = ref(false)
// 监听外部值变化
watch(() => props.modelValue, (newVal) => {
files.value = newVal
})
// 监听内部值变化并通知父组件
watch(files, (newVal) => {
emit('update:modelValue', newVal)
}, { deep: true })
const triggerUpload = () => {
uploadRef.value?.handleClick()
}
const handleFileChange = (uploadFile: UploadFile) => {
if (uploadFile.raw) {
files.value.push(uploadFile.raw)
}
}
const handleDrop = (event: DragEvent) => {
event.preventDefault()
isDragging.value = false
if (event.dataTransfer?.files) {
const droppedFiles = Array.from(event.dataTransfer.files)
droppedFiles.forEach(file => {
if (file.type.match(/(pdf|png|jpe?g)/i)) {
files.value.push(file)
}
})
}
}
const beforeUpload = (file: File): boolean => {
// 阻止自动上传
return false
}
const removeFile = (index: number) => {
files.value.splice(index, 1)
}
const formatFileSize = (bytes: number): string => {
if (bytes === 0) return '0 Bytes'
const k = 1024
const sizes = ['Bytes', 'KB', 'MB', 'GB']
const i = Math.floor(Math.log(bytes) / Math.log(k))
return parseFloat((bytes / Math.pow(k, i)).toFixed(2)) + ' ' + sizes[i]
}
</script>
<style scoped>
.file-uploader {
margin-bottom: 24px;
}
.upload-container {
border: 2px dashed #C9CDD4;
border-radius: 8px;
background-color: #FFFFFF;
padding: 60px 24px;
text-align: center;
transition: all 0.3s ease;
margin-bottom: 24px;
cursor: pointer;
}
.upload-container:hover {
border-color: #165DFF;
background-color: #E8F3FF;
}
.upload-container.drag-over {
border-color: #165DFF;
background-color: #E8F3FF;
}
.upload-container.drag-over .upload-icon {
transform: scale(1.1);
}
.upload-content {
display: flex;
flex-direction: column;
align-items: center;
gap: 20px;
}
.upload-icon {
font-size: 24px;
color: #165DFF;
transition: transform 0.3s ease;
}
.upload-text {
display: flex;
flex-direction: column;
gap: 8px;
}
.main-text {
font-size: 16px;
font-weight: 400;
color: #1D2129;
}
.sub-text {
font-size: 12px;
color: #4E5969;
}
.hidden-upload {
display: none;
}
.file-preview {
margin-top: 20px;
}
.preview-title {
font-size: 16px;
font-weight: 500;
color: #1D2129;
margin: 0 0 16px 0;
}
.file-card {
background-color: #F2F3F5;
border-radius: 6px;
padding: 16px;
margin-bottom: 12px;
transition: all 0.3s ease;
}
.file-card:hover {
box-shadow: 0 2px 8px rgba(0, 0, 0, 0.08);
}
.file-info {
display: flex;
align-items: center;
gap: 12px;
}
.file-icon {
color: #165DFF;
font-size: 18px;
}
.file-name {
flex: 1;
font-weight: 400;
color: #4E5969;
font-size: 14px;
}
.file-size {
color: #4E5969;
font-size: 12px;
}
.remove-button {
opacity: 0.7;
transition: opacity 0.3s ease;
}
.remove-button:hover {
opacity: 1;
}
</style>

View File

@ -1,700 +0,0 @@
<template>
<div class="markdown-renderer" @click="openImage">
<el-image-viewer v-if="imageViewer" :url-list="[imageViewer]" @close="imageViewer = ''" />
<template v-if="content">
<div class="content-toolbar render-actions">
<div class="render-actions-left">
<el-segmented
v-if="mode === 'markdown'"
v-model="activeFlavor"
:options="flavorOptions"
size="small"
/>
<el-segmented
v-else-if="mode === 'richtext'"
v-model="activeRichtextFormat"
:options="richtextOptions"
size="small"
/>
</div>
<div class="render-actions-right">
<el-button type="primary" size="small" class="action-button primary" @click="handleDownload()">
<template #icon>
<el-icon><Download /></el-icon>
</template>
下载
</el-button>
</div>
</div>
<div class="render-content">
<el-button v-if="imageError" link type="warning" @click="retryImages">{{ imageError }}</el-button>
<div v-if="mode === 'markdown'" class="render-shell">
<div ref="markdownRef" class="markdown-content rendered-html" v-html="renderedContent"></div>
</div>
<div v-else-if="mode === 'html'" class="render-shell">
<div ref="htmlRef" class="markdown-content rendered-html" v-html="renderedContent"></div>
</div>
<div v-else-if="mode === 'pdf'" class="render-shell">
<div class="pdf-stage">
<div ref="pdfRef" class="pdf-preview-pages">
<div
v-for="(page, index) in pdfPages"
:key="`pdf-page-${index}`"
class="pdf-page"
>
<div class="markdown-content pdf-content" v-html="page"></div>
</div>
</div>
</div>
<div
ref="pdfMeasureRef"
class="pdf-measure markdown-content pdf-content"
v-html="renderedContent"
></div>
</div>
<div v-else class="render-shell">
<div ref="richtextRef" class="markdown-content rendered-html richtext-content" v-html="renderedContent"></div>
</div>
</div>
</template>
<div v-else class="empty-state">
<el-empty :description="$t('results.noResults')" />
</div>
</div>
</template>
<script setup lang="ts">
import { computed, nextTick, onMounted, ref, watch } from 'vue'
import MarkdownIt from 'markdown-it'
import type { PropType } from 'vue'
import { Download } from '@element-plus/icons-vue'
import { ElMessage } from 'element-plus'
import { loadHtml2Pdf } from '@/utils/wordToPdf'
import { useImageAssets, hasImageAssets, exportMarkdownBundle, embedImages } from '@/utils/imageAssets'
type RenderMode = 'markdown' | 'html' | 'pdf' | 'richtext'
type FlavorMode = 'commonmark' | 'gfm'
type RichtextFormat = 'word' | 'rtf'
const props = defineProps({
content: {
type: String as PropType<string>,
default: ''
},
mode: {
type: String as PropType<RenderMode>,
default: 'markdown'
},
flavor: {
type: String as PropType<FlavorMode>,
default: 'gfm'
}
})
const activeFlavor = ref<FlavorMode>(props.flavor)
const activeRichtextFormat = ref<RichtextFormat>('word')
const markdownRef = ref<HTMLElement | null>(null)
const htmlRef = ref<HTMLElement | null>(null)
const pdfRef = ref<HTMLElement | null>(null)
const pdfMeasureRef = ref<HTMLElement | null>(null)
const richtextRef = ref<HTMLElement | null>(null)
const pdfPages = ref<string[]>([])
const imageViewer = ref('')
function openImage(event: MouseEvent) {
if (event.target instanceof HTMLImageElement && event.target.src) imageViewer.value = event.target.src
}
watch(
() => props.flavor,
(value) => {
activeFlavor.value = value
}
)
const flavorOptions = [
{ label: 'CommonMark', value: 'commonmark' },
{ label: 'GFM', value: 'gfm' }
]
const richtextOptions = [
{ label: 'Word', value: 'word' },
{ label: 'RTF', value: 'rtf' }
]
function applyMathSupport(md: MarkdownIt) {
md.use(function (md) {
md.inline.ruler.after('escape', 'math_inline', function (state, silent) {
const start = state.pos
if (state.src.charCodeAt(start) !== 0x24) return false
if (state.src.charCodeAt(start + 1) !== 0x24) return false
let pos = start + 2
let found = false
while (pos < state.posMax) {
if (state.src.charCodeAt(pos) === 0x24 && state.src.charCodeAt(pos + 1) === 0x24) {
found = true
break
}
pos++
}
if (!found) return false
if (!silent) {
const token = state.push('math_inline', 'span', 0)
token.content = state.src.slice(start + 2, pos)
token.markup = '$$'
}
state.pos = pos + 2
return true
})
md.inline.ruler.after('escape', 'math_inline_single', function (state, silent) {
const start = state.pos
if (state.src.charCodeAt(start) !== 0x24) return false
if (state.src.charCodeAt(start + 1) === 0x24) return false
let pos = start + 1
let found = false
while (pos < state.posMax) {
if (state.src.charCodeAt(pos) === 0x24) {
found = true
break
}
pos++
}
if (!found) return false
if (!silent) {
const token = state.push('math_inline_single', 'span', 0)
token.content = state.src.slice(start + 1, pos)
token.markup = '$'
}
state.pos = pos + 1
return true
})
})
md.renderer.rules.math_inline = function(tokens, idx) {
return `<span class="math-inline">$$${tokens[idx].content}$$</span>`
}
md.renderer.rules.math_inline_single = function(tokens, idx) {
return `<span class="math-inline">$${tokens[idx].content}$</span>`
}
}
function createMarkdownRenderer(flavor: FlavorMode) {
const md = flavor === 'commonmark'
? new MarkdownIt('commonmark', {
html: true,
linkify: true,
typographer: true,
breaks: false
})
: new MarkdownIt({
html: true,
linkify: true,
typographer: true,
breaks: true
})
applyMathSupport(md)
return md
}
const { content: previewContent, error: imageError, retry: retryImages } = useImageAssets(() => props.content)
const renderedContent = computed(() => {
if (!previewContent.value) return ''
const flavor = props.mode === 'markdown' ? activeFlavor.value : 'gfm'
const renderer = createMarkdownRenderer(flavor)
return renderer.render(previewContent.value)
})
async function recalculatePdfPages() {
if (props.mode !== 'pdf') {
pdfPages.value = []
return
}
await nextTick()
const measureEl = pdfMeasureRef.value
if (!measureEl) {
pdfPages.value = renderedContent.value ? [renderedContent.value] : []
return
}
await Promise.all(Array.from(measureEl.querySelectorAll('img')).map(image => image.decode().catch(() => undefined)))
const children = Array.from(measureEl.children) as HTMLElement[]
if (children.length === 0) {
pdfPages.value = renderedContent.value ? [renderedContent.value] : []
return
}
const maxPageHeight = 980
const pages: string[] = []
let currentPageHtml = ''
let currentHeight = 0
children.forEach((child) => {
const childHeight = child.offsetHeight || child.scrollHeight || 0
const childHtml = child.outerHTML
if (currentPageHtml && currentHeight + childHeight > maxPageHeight) {
pages.push(currentPageHtml)
currentPageHtml = childHtml
currentHeight = childHeight
return
}
currentPageHtml += childHtml
currentHeight += childHeight
})
if (currentPageHtml) {
pages.push(currentPageHtml)
}
pdfPages.value = pages.length > 0 ? pages : [renderedContent.value]
}
watch(
() => [props.mode, renderedContent.value],
() => {
void recalculatePdfPages()
},
{ immediate: true }
)
onMounted(() => {
void recalculatePdfPages()
})
const baseFileName = computed(() => `markdown_render_${new Date().getTime()}`)
function downloadBlob(blob: Blob, fileName: string) {
const url = URL.createObjectURL(blob)
const a = document.createElement('a')
a.href = url
a.download = fileName
document.body.appendChild(a)
a.click()
document.body.removeChild(a)
URL.revokeObjectURL(url)
}
function createFullHtmlDocument(innerHtml: string, title: string) {
return `<!DOCTYPE html>
<html lang="zh-CN">
<head>
<meta charset="UTF-8" />
<meta name="viewport" content="width=device-width, initial-scale=1.0" />
<title>${title}</title>
<style>
body { font-family: "Microsoft YaHei", Arial, sans-serif; color: #1f2937; padding: 32px; line-height: 1.75; }
h1, h2, h3, h4, h5, h6 { color: #111827; }
pre { background: #f5f7fa; padding: 16px; border-radius: 8px; overflow: auto; }
code { background: #f5f7fa; padding: 2px 4px; border-radius: 4px; }
table { border-collapse: collapse; width: 100%; }
th, td { border: 1px solid #dcdfe6; padding: 8px 12px; text-align: left; }
blockquote { border-left: 4px solid #dcdfe6; margin: 1em 0; padding-left: 1em; color: #606266; }
img { max-width: 100%; }
</style>
</head>
<body>${innerHtml}</body>
</html>`
}
function htmlToPlainText(html: string) {
const temp = document.createElement('div')
temp.innerHTML = html
return temp.innerText || temp.textContent || ''
}
function escapeRtf(text: string) {
return text
.replace(/\\/g, '\\\\')
.replace(/\{/g, '\\{')
.replace(/\}/g, '\\}')
.replace(/\n/g, '\\par\n')
}
async function exportHtml(innerHtml: string, fileName: string) {
const html = createFullHtmlDocument(await embedImages(innerHtml), fileName)
downloadBlob(new Blob([html], { type: 'text/html;charset=utf-8' }), fileName)
}
async function exportMarkdown(markdown: string, fileName: string) {
if (hasImageAssets(markdown)) {
downloadBlob(await exportMarkdownBundle(markdown), fileName.replace(/\.md$/, '.zip'))
return
}
downloadBlob(new Blob([markdown], { type: 'text/markdown;charset=utf-8' }), fileName)
}
async function exportPdf(element: HTMLElement | null, fileName: string) {
if (!element) {
ElMessage.error('当前没有可导出的 PDF 内容')
return
}
await nextTick()
await loadHtml2Pdf()
const pdfBlob = await window.html2pdf()
.set({
margin: 10,
filename: fileName,
image: { type: 'jpeg', quality: 0.98 },
html2canvas: { scale: 2, useCORS: true },
jsPDF: { unit: 'mm', format: 'a4', orientation: 'portrait' }
})
.from(element)
.outputPdf('blob')
downloadBlob(pdfBlob, fileName)
}
async function exportDoc(innerHtml: string, fileName: string) {
const docHtml = createFullHtmlDocument(await embedImages(innerHtml), fileName)
downloadBlob(new Blob(['\ufeff', docHtml], { type: 'application/msword' }), fileName)
}
async function exportRtf(innerHtml: string, fileName: string) {
const plainText = htmlToPlainText(innerHtml)
const rtf = `{\\rtf1\\ansi\\deff0 {\\fonttbl{\\f0 Calibri;}}\n\\fs24 ${escapeRtf(plainText)}}`
downloadBlob(new Blob([rtf], { type: 'application/rtf' }), fileName)
}
async function handleDownload() {
try {
if (props.mode === 'markdown') {
const suffix = activeFlavor.value === 'commonmark' ? '_commonmark' : '_gfm'
await exportMarkdown(props.content, `${baseFileName.value}${suffix}.md`)
return
}
if (props.mode === 'html') {
await exportHtml(renderedContent.value, `${baseFileName.value}.html`)
return
}
if (props.mode === 'pdf') {
await exportPdf(pdfRef.value, `${baseFileName.value}.pdf`)
return
}
if (activeRichtextFormat.value === 'word') {
await exportDoc(renderedContent.value, `${baseFileName.value}.doc`)
return
}
await exportRtf(renderedContent.value, `${baseFileName.value}.rtf`)
} catch { ElMessage.error('下载失败,请重试') }
}
</script>
<style scoped>
.markdown-renderer {
width: 100%;
height: 100%;
display: flex;
flex-direction: column;
background-color: #FFFFFF;
border-radius: 4px;
box-shadow: none;
overflow: hidden;
min-height: 0;
}
.content-toolbar {
display: flex;
align-items: center;
justify-content: space-between;
gap: 12px;
flex-shrink: 0;
}
.content-toolbar-title {
font-size: 14px;
font-weight: 600;
color: #343A40;
white-space: nowrap;
}
.content-toolbar-meta {
font-size: 12px;
color: #86909C;
white-space: nowrap;
}
.render-actions {
padding: 0 0 12px;
background-color: #FFFFFF;
}
.render-actions-left {
display: flex;
align-items: center;
gap: 12px;
min-height: 32px;
flex: 1;
min-width: 0;
}
.render-actions-right {
display: flex;
align-items: center;
gap: 8px;
flex-shrink: 0;
}
.action-button {
border-radius: 6px;
transition: all 0.3s ease;
height: 32px;
padding: 0 12px;
}
.action-button.primary {
background-color: #165dff;
border-color: #165dff;
color: #ffffff;
}
.action-button.primary:hover {
background-color: #0e42d2;
border-color: #0e42d2;
transform: translateY(-1px);
box-shadow: 0 2px 8px rgba(22, 93, 255, 0.3);
}
.render-shell {
min-height: 100%;
display: flex;
flex-direction: column;
gap: 16px;
}
.render-content {
flex: 1;
min-height: 0;
overflow-y: auto;
padding: 0;
}
.render-header {
padding: 14px 16px;
border-radius: 10px;
background: linear-gradient(135deg, #f7faff 0%, #eef4ff 100%);
border: 1px solid #d7e4ff;
}
.render-title {
font-size: 16px;
font-weight: 600;
color: #1f2937;
}
.render-desc {
margin-top: 4px;
font-size: 13px;
line-height: 1.6;
color: #5b6472;
}
.markdown-content {
padding: 20px;
background: white;
border-radius: 4px;
min-height: 400px;
}
.rendered-html {
border: 1px solid #e9ecef;
}
.pdf-stage {
padding: 20px;
border: 1px solid #E9ECEF;
border-radius: 4px;
background: #F8F9FA;
}
.pdf-preview-pages {
display: flex;
flex-direction: column;
gap: 20px;
}
.pdf-page {
width: min(100%, 840px);
min-height: 1160px;
margin: 0 auto;
background: white;
border: 1px solid #d9dee8;
box-shadow: 0 18px 45px rgba(15, 23, 42, 0.12);
padding: 28px 36px 36px;
}
.pdf-content {
padding: 0;
min-height: auto;
}
.pdf-measure {
position: absolute;
left: -99999px;
top: 0;
width: 768px;
visibility: hidden;
pointer-events: none;
z-index: -1;
}
.richtext-stage {
padding: 12px;
border: 1px solid #E9ECEF;
border-radius: 4px;
background: #F8F9FA;
}
.richtext-document {
min-height: 100%;
background: white;
border: 1px solid #dde3ea;
border-radius: 4px;
overflow: hidden;
box-shadow: none;
}
.richtext-content {
padding: 32px 40px 48px;
}
.markdown-content :deep(h1) {
font-size: 2em;
margin: 0.67em 0;
font-weight: bold;
}
.markdown-content :deep(h2) {
font-size: 1.5em;
margin: 0.83em 0;
font-weight: bold;
}
.markdown-content :deep(h3) {
font-size: 1.17em;
margin: 1em 0;
font-weight: bold;
}
.markdown-content :deep(p) {
margin: 1em 0;
line-height: 1.75;
}
.markdown-content :deep(code) {
background-color: #f5f7fa;
padding: 2px 4px;
border-radius: 3px;
font-family: 'Monaco', 'Menlo', 'Ubuntu Mono', monospace;
}
.markdown-content :deep(pre) {
background-color: #f5f7fa;
padding: 16px;
border-radius: 8px;
overflow-x: auto;
}
.markdown-content :deep(pre code) {
background: none;
padding: 0;
}
.markdown-content :deep(blockquote) {
margin: 1em 0;
padding: 0 1em;
border-left: 4px solid #dcdfe6;
color: #606266;
}
.markdown-content :deep(ul), .markdown-content :deep(ol) {
margin: 1em 0;
padding-left: 2em;
}
.markdown-content :deep(li) {
margin: 0.5em 0;
}
.markdown-content :deep(table) {
border-collapse: collapse;
width: 100%;
margin: 1em 0;
}
.markdown-content :deep(th), .markdown-content :deep(td) {
border: 1px solid #dcdfe6;
padding: 8px 12px;
text-align: left;
}
.markdown-content :deep(th) {
background-color: #f5f7fa;
font-weight: bold;
}
.markdown-content :deep(img) {
max-width: 100%;
height: auto;
}
.empty-state {
display: flex;
align-items: center;
justify-content: center;
height: 100%;
min-height: 400px;
}
.math-inline {
font-family: 'Cambria Math', 'STIX Two Math', serif;
font-size: 1.1em;
}
@media (max-width: 768px) {
.render-actions {
flex-direction: column;
align-items: stretch;
}
.render-actions-left {
flex: 1;
min-width: 0;
flex-wrap: wrap;
}
.render-actions-right {
justify-content: flex-end;
}
.pdf-page {
padding: 24px 20px 32px;
min-height: auto;
}
.richtext-content {
padding: 24px 18px 32px;
}
.action-button {
flex-shrink: 0;
}
}
</style>

View File

@ -1,747 +0,0 @@
<template>
<div class="mindmap-container">
<el-image-viewer v-if="imageViewer" :url-list="[imageViewer]" @close="imageViewer = ''" />
<div class="content-toolbar mindmap-actions">
<div class="mindmap-actions-left">
<span class="content-toolbar-meta">{{ nodeSummary }}</span>
</div>
<div class="mindmap-actions-right">
<el-dropdown @command="handleDownload">
<el-button
type="primary"
size="small"
class="action-button primary"
:disabled="!content"
>
<template #icon>
<el-icon><Download /></el-icon>
</template>
下载
<el-icon class="el-icon--right"><ArrowDown /></el-icon>
</el-button>
<template #dropdown>
<el-dropdown-menu>
<el-dropdown-item command="svg">SVG 格式</el-dropdown-item>
<el-dropdown-item command="png">PNG 格式</el-dropdown-item>
<el-dropdown-item command="jpeg">JPEG 格式</el-dropdown-item>
</el-dropdown-menu>
</template>
</el-dropdown>
<el-button
type="default"
size="small"
@click="resetView"
class="action-button secondary"
>
<template #icon>
<el-icon><Refresh /></el-icon>
</template>
重置视图
</el-button>
</div>
</div>
<div class="mindmap-content" @click="handleNodeClick">
<el-button v-if="imageError" link type="warning" @click.stop="retryImages">{{ imageError }}</el-button>
<div class="empty-state" v-if="!content">
<el-icon class="empty-icon"><Document /></el-icon>
<p class="empty-text">暂无思维导图内容</p>
<p class="empty-subtext">请先上传并转换文档</p>
</div>
<div class="mindmap-loading" v-if="content && isRendering">
<el-icon class="loading-icon"><Loading /></el-icon>
<span>正在生成导图...</span>
</div>
<svg ref="svgRef" class="markmap-svg" v-show="content"></svg>
</div>
</div>
</template>
<script setup lang="ts">
import { ref, computed, onMounted, watch, onUnmounted, nextTick } from 'vue'
import { Download, Refresh, Document, ArrowDown, Loading } from '@element-plus/icons-vue'
import { ElMessage, ElMessageBox } from 'element-plus'
// 导入 markmap 相关包
import { Transformer } from 'markmap-lib'
import { Markmap, loadCSS, loadJS } from 'markmap-view'
import * as markmapView from 'markmap-view'
import { useImageAssets, embedImages } from '@/utils/imageAssets'
import { captureFolds, restoreFolds, type FoldState } from '@/utils/mindmapState'
const props = defineProps({
content: {
type: String,
default: ''
},
exportResolution: {
type: Number,
default: 0
},
documentKey: {
type: String,
default: 'document'
}
})
const emit = defineEmits<{
(e: 'node-edit', payload: { oldText: string; newText: string }): void
}>()
const svgRef = ref<SVGElement | null>(null)
let mmInstance: any = null
let renderTimer: number | undefined
let activeDocumentKey = ''
let renderQueue = Promise.resolve()
let needsInitialFit = true
let viewportObserver: ResizeObserver | undefined
const viewStates = new Map<string, { folds: FoldState; camera: any }>()
function rememberView() {
if (!activeDocumentKey || !mmInstance?.state?.data) return
viewStates.set(activeDocumentKey, { folds: captureFolds(mmInstance.state.data), camera: needsInitialFit ? undefined : mmInstance.svg.property('__zoom') })
if (viewStates.size > 10) viewStates.delete(viewStates.keys().next().value!)
}
const transformer = new Transformer()
const scale = ref(1)
const isRendering = ref(false)
const totalNodeCount = ref(0)
const hiddenNodeCount = ref(0)
type MarkmapNode = {
content?: string
children?: MarkmapNode[]
payload?: {
fold?: number
}
}
const nodeSummary = computed(() => {
if (!props.content) return '暂无节点'
return hiddenNodeCount.value > 0
? `已生成 ${totalNodeCount.value} 个节点,隐藏 ${hiddenNodeCount.value} 个节点`
: `已生成 ${totalNodeCount.value} 个节点`
})
const updateNodeStats = () => {
const root = mmInstance?.state?.data as MarkmapNode | undefined
if (!root || !props.content) {
totalNodeCount.value = 0
hiddenNodeCount.value = 0
return
}
const countSubtree = (node: MarkmapNode): number => {
const children = node.children || []
return 1 + children.reduce((sum, child) => sum + countSubtree(child), 0)
}
const walkVisible = (node: MarkmapNode): { total: number; hidden: number } => {
const children = node.children || []
const total = 1 + children.reduce((sum, child) => sum + countSubtree(child), 0)
if (node.payload?.fold && children.length > 0) {
return {
total,
hidden: children.reduce((sum, child) => sum + countSubtree(child), 0)
}
}
const childStats = children.reduce(
(stats, child) => {
const childStat = walkVisible(child)
return {
total: stats.total + childStat.total,
hidden: stats.hidden + childStat.hidden
}
},
{ total: 1, hidden: 0 }
)
return childStats
}
const stats = walkVisible(root)
totalNodeCount.value = Math.max(0, stats.total - 1)
hiddenNodeCount.value = Math.max(0, stats.hidden)
}
const { content: imageContent, error: imageError, retry: retryImages } = useImageAssets(() => props.content)
const imageViewer = ref('')
const initMarkmap = async () => {
if (!svgRef.value) return
if (props.content && !imageContent.value) return
if (!props.content) {
svgRef.value.innerHTML = ''
if (mmInstance && typeof mmInstance.destroy === 'function') {
mmInstance.destroy()
}
mmInstance = null
return
}
isRendering.value = true
try {
rememberView()
const documentKey = props.documentKey
const saved = viewStates.get(documentKey)
const { root, features } = transformer.transform(imageContent.value)
restoreFolds(root, saved?.folds)
const { styles, scripts } = transformer.getUsedAssets(features)
if (styles) loadCSS(styles)
if (scripts) await loadJS(scripts, { getMarkmap: () => markmapView })
await nextTick()
const initialExpandLevel = -1
if (mmInstance) {
if (typeof mmInstance.setData === 'function') {
await mmInstance.setData(root, { initialExpandLevel, autoFit: false })
} else {
mmInstance = Markmap.create(svgRef.value, {
autoFit: false,
fitRatio: 0.9,
initialExpandLevel
}, root)
}
} else {
mmInstance = Markmap.create(svgRef.value, {
autoFit: false,
fitRatio: 0.9,
initialExpandLevel
}, root)
}
await nextTick()
svgRef.value?.querySelectorAll('text').forEach((text) => {
text.setAttribute('data-editable-node', 'true')
})
updateNodeStats()
await Promise.all(Array.from(svgRef.value?.querySelectorAll('img') || []).map(image => image.decode().catch(() => undefined)))
activeDocumentKey = documentKey
needsInitialFit = !saved?.camera
if (saved?.camera) mmInstance.svg.call(mmInstance.zoom.transform, saved.camera)
else if (svgRef.value.getBoundingClientRect().width > 0) { needsInitialFit = false; await mmInstance.fit() }
} catch (error) {
console.error('Error initializing markmap:', error)
mmInstance = null
} finally {
isRendering.value = false
}
}
// 监听内容变化自动重绘
watch([imageContent, () => props.documentKey], () => {
window.clearTimeout(renderTimer)
renderTimer = window.setTimeout(() => {
renderQueue = renderQueue.catch(() => undefined).then(initMarkmap)
}, 160)
})
onMounted(() => {
renderQueue = renderQueue.then(initMarkmap)
viewportObserver = new ResizeObserver(() => {
if (needsInitialFit && mmInstance?.state?.data && svgRef.value && svgRef.value.getBoundingClientRect().width > 0) {
needsInitialFit = false
void mmInstance.fit()
}
})
if (svgRef.value) viewportObserver.observe(svgRef.value)
// 窗口缩放时自动调整导图大小
window.addEventListener('resize', handleResize)
})
onUnmounted(() => {
viewportObserver?.disconnect()
window.clearTimeout(renderTimer)
window.removeEventListener('resize', handleResize)
if (mmInstance && typeof mmInstance.destroy === 'function') {
mmInstance.destroy()
} else if (mmInstance) {
console.error('mmInstance does not have destroy method:', mmInstance)
}
})
const handleResize = () => {
mmInstance?.fit()
}
function downloadBlob(blob: Blob, fileName: string) {
const url = URL.createObjectURL(blob)
const a = document.createElement('a')
a.href = url
a.download = fileName
document.body.appendChild(a)
a.click()
document.body.removeChild(a)
URL.revokeObjectURL(url)
}
type ExportFormat = 'svg' | 'png' | 'jpeg'
interface ExportBox {
x: number
y: number
width: number
height: number
}
function getViewportSize(svg: SVGElement) {
const rect = svg.getBoundingClientRect()
return {
width: Math.ceil(Math.max(rect.width, 320)),
height: Math.ceil(Math.max(rect.height, 240))
}
}
function cleanExportSvg(svg: SVGElement) {
if (!svg.getAttribute('xmlns')) {
svg.setAttribute('xmlns', 'http://www.w3.org/2000/svg')
}
svg.querySelectorAll('[style*="cursor"]').forEach((node) => node.removeAttribute('style'))
}
function getBoxFromSvgRect(rect: DOMRect): ExportBox {
return {
x: rect.x,
y: rect.y,
width: rect.width,
height: rect.height
}
}
function transformBox(box: ExportBox, matrix: DOMMatrix): ExportBox {
const points = [
new DOMPoint(box.x, box.y).matrixTransform(matrix),
new DOMPoint(box.x + box.width, box.y).matrixTransform(matrix),
new DOMPoint(box.x, box.y + box.height).matrixTransform(matrix),
new DOMPoint(box.x + box.width, box.y + box.height).matrixTransform(matrix)
]
const xs = points.map((point) => point.x)
const ys = points.map((point) => point.y)
const minX = Math.min(...xs)
const minY = Math.min(...ys)
const maxX = Math.max(...xs)
const maxY = Math.max(...ys)
return {
x: minX,
y: minY,
width: maxX - minX,
height: maxY - minY
}
}
function getIntersection(a: ExportBox, b: ExportBox): ExportBox | null {
const x = Math.max(a.x, b.x)
const y = Math.max(a.y, b.y)
const right = Math.min(a.x + a.width, b.x + b.width)
const bottom = Math.min(a.y + a.height, b.y + b.height)
if (right <= x || bottom <= y) return null
return {
x,
y,
width: right - x,
height: bottom - y
}
}
function padBox(box: ExportBox, padding: number, boundary?: ExportBox): ExportBox {
const padded = {
x: box.x - padding,
y: box.y - padding,
width: box.width + padding * 2,
height: box.height + padding * 2
}
if (!boundary) return padded
const clipped = getIntersection(padded, boundary)
return clipped || box
}
function normalizeBox(box: ExportBox): ExportBox {
const x = Math.min(box.x, box.x + box.width)
const y = Math.min(box.y, box.y + box.height)
return {
x,
y,
width: Math.abs(box.width),
height: Math.abs(box.height)
}
}
function isCompleteInViewport(contentBox: ExportBox, viewportBox: ExportBox) {
const tolerance = 4
return (
contentBox.x >= viewportBox.x - tolerance &&
contentBox.y >= viewportBox.y - tolerance &&
contentBox.x + contentBox.width <= viewportBox.x + viewportBox.width + tolerance &&
contentBox.y + contentBox.height <= viewportBox.y + viewportBox.height + tolerance
)
}
function createFullExportSvg() {
const svg = svgRef.value
if (!svg) return null
const svgCopy = svg.cloneNode(true) as SVGElement
cleanExportSvg(svgCopy)
const contentGroup = svg.querySelector('g')
const bbox = contentGroup
? (contentGroup as SVGGElement).getBBox()
: svg.getBBox()
const padding = 48
const width = Math.ceil(Math.max(bbox.width + padding * 2, 320))
const height = Math.ceil(Math.max(bbox.height + padding * 2, 240))
const viewBox = `${Math.floor(bbox.x - padding)} ${Math.floor(bbox.y - padding)} ${width} ${height}`
const viewportGroupCopy = svgCopy.querySelector('g')
viewportGroupCopy?.removeAttribute('transform')
svgCopy.setAttribute('viewBox', viewBox)
svgCopy.setAttribute('width', String(width))
svgCopy.setAttribute('height', String(height))
return {
data: new XMLSerializer().serializeToString(svgCopy),
width,
height
}
}
function createVisibleExportSvg() {
const svg = svgRef.value
if (!svg) return null
const contentGroup = svg.querySelector('g') as SVGGElement | null
if (!contentGroup) return createFullExportSvg()
const viewport = getViewportSize(svg)
const viewportBox = { x: 0, y: 0, width: viewport.width, height: viewport.height }
const ctm = contentGroup.getCTM()
if (!ctm) return createFullExportSvg()
const contentBox = transformBox(getBoxFromSvgRect(contentGroup.getBBox()), ctm)
if (isCompleteInViewport(contentBox, viewportBox)) {
return createFullExportSvg()
}
const svgCopy = svg.cloneNode(true) as SVGElement
cleanExportSvg(svgCopy)
const screenVisibleBox = padBox(getIntersection(contentBox, viewportBox) || viewportBox, 24, viewportBox)
const localVisibleBox = normalizeBox(transformBox(screenVisibleBox, ctm.inverse()))
const localContentBox = getBoxFromSvgRect(contentGroup.getBBox())
const clippedLocalBox = getIntersection(localVisibleBox, localContentBox) || localVisibleBox
const visibleBox = padBox(clippedLocalBox, 24)
const width = Math.ceil(Math.max(visibleBox.width, 320))
const height = Math.ceil(Math.max(visibleBox.height, 240))
const viewBox = `${Math.floor(visibleBox.x)} ${Math.floor(visibleBox.y)} ${width} ${height}`
const viewportGroupCopy = svgCopy.querySelector('g')
viewportGroupCopy?.removeAttribute('transform')
svgCopy.setAttribute('width', String(width))
svgCopy.setAttribute('height', String(height))
svgCopy.setAttribute('viewBox', viewBox)
return {
data: new XMLSerializer().serializeToString(svgCopy),
width,
height
}
}
// 下载思维导图
const downloadMindMap = async (format: ExportFormat = 'svg') => {
const exportSvg = createVisibleExportSvg()
if (!exportSvg) return
exportSvg.data = await embedImages(exportSvg.data)
const time = new Date().getTime()
if (format === 'svg') {
downloadBlob(new Blob([exportSvg.data], { type: 'image/svg+xml;charset=utf-8' }), `mindmap_${time}.svg`)
return
}
const baseScale = 3
const maxResolution = Number(props.exportResolution) || 0
const longEdge = Math.max(exportSvg.width, exportSvg.height)
const scaleFactor = maxResolution > 0
? Math.min(baseScale, Math.max(0.1, maxResolution / longEdge))
: baseScale
const canvas = document.createElement('canvas')
const ctx = canvas.getContext('2d')
if (!ctx) return
canvas.width = Math.ceil(exportSvg.width * scaleFactor)
canvas.height = Math.ceil(exportSvg.height * scaleFactor)
ctx.fillStyle = '#ffffff'
ctx.fillRect(0, 0, canvas.width, canvas.height)
ctx.scale(scaleFactor, scaleFactor)
const img = new Image()
img.onload = () => {
ctx.drawImage(img, 0, 0)
const mimeType = format === 'jpeg' ? 'image/jpeg' : 'image/png'
canvas.toBlob((blob) => {
if (!blob) return
downloadBlob(blob, `mindmap_${time}.${format === 'jpeg' ? 'jpg' : 'png'}`)
}, mimeType, 0.96)
}
img.onerror = () => ElMessage.error('导出图片失败')
img.src = 'data:image/svg+xml;charset=utf-8,' + encodeURIComponent(exportSvg.data)
}
// 处理下载命令
const handleDownload = (command: string) => {
if (!props.content) {
ElMessage.warning('暂无思维导图内容可下载')
return
}
if (command === 'svg' || command === 'png' || command === 'jpeg') {
downloadMindMap(command)
}
}
// 重置视图
const resetView = () => {
mmInstance?.fit()
scale.value = 1
window.setTimeout(updateNodeStats, 0)
}
// 放大
const zoomIn = () => {
if (scale.value < 2) {
scale.value += 0.1
applyZoom()
}
}
// 缩小
const zoomOut = () => {
if (scale.value > 0.5) {
scale.value -= 0.1
applyZoom()
}
}
// 应用缩放
const applyZoom = () => {
if (mmInstance && typeof mmInstance.setScale === 'function') {
mmInstance.setScale(scale.value)
}
}
const handleNodeClick = async (event: MouseEvent) => {
if (event.target instanceof HTMLImageElement && event.target.src) { imageViewer.value = event.target.src; return }
window.setTimeout(updateNodeStats, 0)
const textNode = (event.target as Element | null)?.closest?.('text')
const oldText = textNode?.textContent?.trim()
const rootText = (props.content.split('\n')[0] || '').replace(/^#\s+/, '').replace(/\\(.)/g, '$1')
if (!oldText || !props.content || oldText === rootText) return
try {
const { value } = await ElMessageBox.prompt('编辑节点文本', '编辑思维导图节点', {
inputValue: oldText,
confirmButtonText: '确认',
cancelButtonText: '取消',
inputValidator: (value) => Boolean(value.trim()) || '节点文本不能为空'
})
const newText = String(value).trim()
if (newText && newText !== oldText) {
emit('node-edit', { oldText, newText })
}
} catch {
// 用户取消编辑
}
}
</script>
<style scoped>
.mindmap-container {
width: 100%;
height: 100%;
display: flex;
flex-direction: column;
background-color: transparent;
border-radius: 0;
box-shadow: none;
overflow: hidden;
min-height: 0;
}
.content-toolbar {
display: flex;
align-items: center;
justify-content: space-between;
gap: 12px;
flex-shrink: 0;
}
.mindmap-actions {
padding: 0 0 12px;
}
.mindmap-actions-left,
.mindmap-actions-right {
display: flex;
align-items: center;
gap: 10px;
min-width: 0;
}
.mindmap-actions-left {
flex: 1;
}
.mindmap-actions-right {
flex-shrink: 0;
}
.content-toolbar-title {
font-size: 14px;
font-weight: 600;
color: #343A40;
white-space: nowrap;
}
.content-toolbar-meta {
font-size: 12px;
color: #86909C;
white-space: nowrap;
}
.action-button {
border-radius: 6px;
transition: all 0.3s ease;
height: 32px;
padding: 0 12px;
}
.action-button.primary {
background-color: #165DFF;
border-color: #165DFF;
color: #FFFFFF;
}
.action-button.primary:hover {
background-color: #0E42D2;
border-color: #0E42D2;
transform: translateY(-1px);
box-shadow: 0 2px 8px rgba(22, 93, 255, 0.3);
}
.action-button.secondary {
background-color: #FFFFFF;
border-color: #C9CDD4;
color: #4E5969;
}
.action-button.secondary:hover {
border-color: #165DFF;
color: #165DFF;
transform: translateY(-1px);
}
.mindmap-content {
flex: 1;
overflow: hidden;
position: relative;
background-color: #F9FAFC;
cursor: grab;
min-height: 0;
border: 1px solid #E9ECEF;
border-radius: 4px;
}
.mindmap-content:active {
cursor: grabbing;
}
.empty-state {
height: 100%;
display: flex;
flex-direction: column;
justify-content: center;
align-items: center;
gap: 16px;
}
.mindmap-loading {
position: absolute;
inset: 0;
display: flex;
align-items: center;
justify-content: center;
gap: 8px;
color: #4E5969;
background: rgba(249, 250, 252, 0.82);
z-index: 2;
}
.loading-icon {
color: #165DFF;
animation: spin 1s linear infinite;
}
@keyframes spin {
from { transform: rotate(0deg); }
to { transform: rotate(360deg); }
}
.empty-icon {
font-size: 48px;
color: #C9CDD4;
}
.empty-text {
font-size: 16px;
font-weight: 500;
color: #4E5969;
margin: 0;
}
.empty-subtext {
font-size: 14px;
color: #86909C;
margin: 0;
}
.markmap-svg {
width: 100%;
height: 100%;
transition: transform 0.2s ease;
}
.markmap-svg :deep(text[data-editable-node="true"]) {
cursor: text;
}
/* 响应式设计 */
@media (max-width: 768px) {
.mindmap-actions {
flex-direction: column;
align-items: stretch;
}
.mindmap-actions-left,
.mindmap-actions-right {
justify-content: space-between;
}
.content-toolbar-meta {
font-size: 11px;
}
.action-button {
padding: 0 8px;
font-size: 12px;
}
.mindmap-actions {
gap: 4px;
}
}
</style>

View File

@ -1,198 +0,0 @@
<template>
<div class="result-panel">
<div class="panel-header" v-if="result">
<el-button
type="primary"
:icon="Download"
@click="downloadResult"
size="small"
class="download-button"
>
{{ $t('results.download') }}
</el-button>
</div>
<!-- 加载状态 -->
<div v-if="isProcessing" class="loading-container">
<el-spinner :size="64" />
<p class="loading-text">处理中...</p>
</div>
<!-- 结果内容 -->
<el-tabs v-else v-model="activeTab" class="result-tabs">
<el-tab-pane :label="$t('results.tabs.markdown')" name="markdown">
<MarkdownRenderer :content="result?.markdown || ''" />
</el-tab-pane>
<el-tab-pane :label="$t('results.tabs.source')" name="source">
<el-input
:model-value="result?.source"
type="textarea"
:rows="20"
readonly
class="source-textarea"
/>
</el-tab-pane>
<el-tab-pane :label="$t('results.tabs.mindmap')" name="mindmap">
<MindMapRenderer :content="result?.mindmap || ''" />
</el-tab-pane>
</el-tabs>
</div>
</template>
<script setup lang="ts">
import { ref } from 'vue'
import { Download } from '@element-plus/icons-vue'
import type { ProcessResult } from '@/composables/useDocumentProcessor'
import MarkdownRenderer from './MarkdownRenderer.vue'
import MindMapRenderer from './MindMapRenderer.vue'
interface Props {
result: ProcessResult | null
isProcessing: boolean
}
const props = defineProps<Props>()
const activeTab = ref('markdown')
const downloadResult = () => {
if (!props.result?.markdown) return
const blob = new Blob([props.result.markdown], { type: 'text/markdown' })
const url = URL.createObjectURL(blob)
const a = document.createElement('a')
a.href = url
a.download = `result_${new Date().getTime()}.md`
document.body.appendChild(a)
a.click()
document.body.removeChild(a)
URL.revokeObjectURL(url)
}
</script>
<style scoped>
.result-panel {
height: 100%;
display: flex;
flex-direction: column;
}
.panel-header {
display: flex;
justify-content: flex-end;
align-items: center;
margin-bottom: 16px;
}
.download-button {
border-radius: 6px;
transition: all 0.3s ease;
background-color: #165DFF;
border-color: #165DFF;
}
.download-button:hover {
background-color: #0E42D2;
border-color: #0E42D2;
transform: scale(1.05);
}
.result-tabs {
flex: 1;
min-height: 0;
}
.result-tabs :deep(.el-tabs__content) {
height: calc(100% - 48px);
padding-top: 16px;
}
.result-tabs :deep(.el-tab-pane) {
height: 100%;
}
/* 极简Tab样式 */
.result-tabs :deep(.el-tabs__nav) {
border: none;
padding: 0;
}
.result-tabs :deep(.el-tabs__item) {
height: 48px;
line-height: 48px;
padding: 0 24px;
margin-right: 24px;
color: #4E5969;
font-size: 14px;
font-weight: 400;
position: relative;
transition: all 0.3s ease;
}
.result-tabs :deep(.el-tabs__item:hover) {
color: #165DFF;
}
.result-tabs :deep(.el-tabs__item.is-active) {
color: #165DFF;
font-weight: 500;
}
.result-tabs :deep(.el-tabs__item.is-active::after) {
content: '';
position: absolute;
bottom: 0;
left: 0;
right: 0;
height: 2px;
background-color: #165DFF;
border-radius: 1px;
transition: all 0.3s ease;
}
.result-tabs :deep(.el-tabs__active-bar) {
display: none;
}
.source-textarea {
height: 100%;
}
.source-textarea :deep(.el-textarea__wrapper) {
border: 1px solid #C9CDD4;
border-radius: 8px;
box-shadow: none;
transition: all 0.3s ease;
}
.source-textarea :deep(.el-textarea__wrapper:hover) {
border-color: #165DFF;
}
.source-textarea :deep(.el-textarea__inner) {
height: 100%;
font-family: 'Monaco', 'Menlo', 'Ubuntu Mono', monospace;
font-size: 14px;
line-height: 1.5;
color: #4E5969;
padding: 16px;
}
/* 加载状态 */
.loading-container {
flex: 1;
display: flex;
flex-direction: column;
justify-content: center;
align-items: center;
min-height: 400px;
}
.loading-text {
margin-top: 16px;
font-size: 16px;
color: #4E5969;
}
</style>

View File

@ -1,411 +0,0 @@
import { ref, reactive, watch } from 'vue'
import type { Ref } from 'vue'
import { documentApi, type ParseParams } from '@/api/document'
import { convertWordToPdf, isWordFile } from '@/utils/wordToPdf'
import { buildMindmapMarkdown, documentTitleFromName } from '@/utils/mindmapMarkdown'
// 根据上一级标题自动补全下一级标题
export function autoPromoteParagraphsToSubheading(text: string): string {
return buildMindmapMarkdown(text)
}
export interface DocumentConfig {
maxPages: number
backend: string
serverUrl: string
exportResolution: number
mindmapPrompt: string
parserResourceId: string
mindmapModelId: string
mindmapLlmTimeout: number
mindmapLlmTemperature: number
mindmapLlmMaxContextTokens: number
mindmapLlmMaxOutputTokens: number
mindmapLlmSafetyTokens: number
tableEnable: boolean
formulaEnable: boolean
language: string
forceOcr: boolean
}
export interface ProcessResult {
markdown: string
source: string
mindmap: string
downloadUrl?: string
}
export const DOCUMENT_CONFIG_STORAGE_KEY = 'mineru.documentProcessor.config.v4'
export const DEFAULT_MINDMAP_PROMPT = `你是文档结构整理助手。请基于用户提供的 Markdown 生成适合思维导图展示的 Markdown。
要求:
1. 保留原文标题结构,不要重写或打乱主要标题层级。
2. 将标题下的段落内容总结为要点,合并相近段落,避免逐段照抄。
3. 保留原文语言:英文内容输出英文,中文内容输出中文,多语言内容按原文语言分别保留。
4. 不要编造原文没有的信息。
5. 保留关键数字、公式、专有名词、步骤和结论。
6. 最大层级不超过 4 层。
7. 每个父节点下最多 8 个子节点。
8. 节点标题尽量简短,正文说明使用短句列表。
9. 只输出 Markdown,不要输出解释、代码块围栏或额外说明。`
export const DEFAULT_DOCUMENT_CONFIG: DocumentConfig = {
maxPages: 1000,
backend: 'hybrid-auto-engine',
serverUrl: 'http://localhost:30000',
exportResolution: 7680,
mindmapPrompt: DEFAULT_MINDMAP_PROMPT,
parserResourceId: '',
mindmapModelId: '',
mindmapLlmTimeout: 180,
mindmapLlmTemperature: 0.2,
mindmapLlmMaxContextTokens: 32768,
mindmapLlmMaxOutputTokens: 4096,
mindmapLlmSafetyTokens: 1024,
tableEnable: true,
formulaEnable: true,
language: 'ch',
forceOcr: false
}
function loadCachedConfig(): Partial<DocumentConfig> {
try {
const raw = window.localStorage.getItem(DOCUMENT_CONFIG_STORAGE_KEY)
const cached = raw ? JSON.parse(raw) : {}
const safe = Object.fromEntries(Object.keys(DEFAULT_DOCUMENT_CONFIG).filter(key => key in cached).map(key => [key, cached[key]]))
window.localStorage.setItem(DOCUMENT_CONFIG_STORAGE_KEY, JSON.stringify(safe))
return safe
} catch {
try { window.localStorage.removeItem(DOCUMENT_CONFIG_STORAGE_KEY) } catch { /* Storage unavailable. */ }
return {}
}
}
export function useDocumentProcessor() {
// 文件相关
const uploadedFiles: Ref<File[]> = ref([])
const uploadedDisplayFiles: Ref<File[]> = ref([])
const isUploading = ref(false)
// 配置相关
const config = reactive<DocumentConfig>({
...DEFAULT_DOCUMENT_CONFIG,
...loadCachedConfig()
})
watch(
config,
(value) => {
try {
window.localStorage.setItem(DOCUMENT_CONFIG_STORAGE_KEY, JSON.stringify(value))
} catch {
// 浏览器禁用本地存储时不影响当前页面使用。
}
},
{ deep: true }
)
// 结果相关
const results = ref<ProcessResult | null>(null)
const isProcessing = ref(false)
const processingStage = ref('')
const error = ref<string | null>(null)
// 后端选项
const backendOptions = [
{ value: 'pipeline', label: '传统管道解析' },
{ value: 'vlm-auto-engine', label: 'VLM本地引擎' },
{ value: 'hybrid-auto-engine', label: '混合本地引擎' },
{ value: 'vlm-http-client', label: 'VLM远程客户端' },
{ value: 'hybrid-http-client', label: '混合远程客户端' }
]
// 语言选项
const languageOptions = [
{ value: 'ch', label: '中文(简体)' },
{ value: 'en', label: '英语' },
{ value: 'korean', label: '韩语' },
{ value: 'japan', label: '日语' },
{ value: 'chinese_cht', label: '中文(繁体)' },
{ value: 'ta', label: '泰米尔语' },
{ value: 'te', label: '泰卢固语' },
{ value: 'ka', label: '卡纳达语' },
{ value: 'th', label: '泰语' },
{ value: 'el', label: '希腊语' },
{ value: 'latin', label: '拉丁语系' },
{ value: 'arabic', label: '阿拉伯语系' },
{ value: 'east_slavic', label: '东斯拉夫语系' },
{ value: 'cyrillic', label: '西里尔语系' },
{ value: 'devanagari', label: '梵文字母语系' }
]
// 文件上传处理
const handleFileUpload = async (files: FileList | null) => {
console.log('handleFileUpload called with files:', files)
if (!files || files.length === 0) return
const validFiles: File[] = []
const displayFiles: File[] = []
const file = files[0]
uploadedFiles.value = []
uploadedDisplayFiles.value = []
results.value = null
error.value = null
if (files.length > 1) {
error.value = '一次只能选择一个文件,已使用第一个文件'
}
const fileType = file.type
const fileName = file.name.toLowerCase()
console.log('Processing file:', fileName, 'type:', fileType)
// 验证文件类型 - 支持 Markdown、PDF、Word 和后端可转换为 PDF 的图片格式
const supportedImageExtensions = ['.png', '.jpg', '.jpeg', '.webp', '.bmp', '.tiff', '.gif', '.jp2']
const isImage = fileType.startsWith('image/') && supportedImageExtensions.some((ext) => fileName.endsWith(ext))
const isPdf = fileType === 'application/pdf'
const isWord = fileName.endsWith('.docx') || fileName.endsWith('.doc') || fileType === 'application/vnd.openxmlformats-officedocument.wordprocessingml.document'
const isMarkdown = fileName.endsWith('.md') || fileType === 'text/markdown'
console.log('File type checks - isImage:', isImage, 'isPdf:', isPdf, 'isWord:', isWord, 'isMarkdown:', isMarkdown)
if (!isImage && !isPdf && !isWord && !isMarkdown) {
error.value = '不支持的文件类型,请选择 MD、PDF、Word、JPG、PNG、WEBP、BMP、TIFF、GIF 或 JP2 文件'
console.log('Unsupported file type:', fileName, fileType)
return
}
// 验证文件大小 (100MB)
if (file.size > 100 * 1024 * 1024) {
error.value = '文件大小超出限制'
console.log('File too large:', fileName, file.size)
return
}
if (isMarkdown) {
try {
isUploading.value = true
const markdown = await file.text()
validFiles.push(file)
displayFiles.push(file)
results.value = {
markdown,
source: markdown,
mindmap: buildMindmapMarkdown(markdown, documentTitleFromName(file.name))
}
error.value = null
console.log('Markdown file loaded directly:', fileName, markdown.length)
} catch (err) {
error.value = 'Markdown 文件读取失败: ' + (err as Error).message
console.error('Markdown read failed:', (err as Error).message)
return
} finally {
isUploading.value = false
}
} else if (isWord) {
try {
isUploading.value = true
error.value = '正在将 Word 文档转换为 PDF...'
console.log('Converting Word to PDF:', fileName)
const pdfFile = await convertWordToPdf(file)
console.log('Conversion successful, PDF file:', pdfFile.name, pdfFile.size)
validFiles.push(pdfFile)
displayFiles.push(file)
error.value = null
} catch (err) {
error.value = 'Word 转换为 PDF 失败: ' + (err as Error).message
console.error('Conversion failed:', (err as Error).message)
return
} finally {
isUploading.value = false
}
} else {
validFiles.push(file)
displayFiles.push(file)
console.log('Adding file directly:', fileName)
}
uploadedFiles.value = validFiles
uploadedDisplayFiles.value = displayFiles
console.log('Final uploaded files:', validFiles.map(f => f.name))
}
// 清除所有数据
const clearAll = () => {
uploadedFiles.value = []
uploadedDisplayFiles.value = []
results.value = null
error.value = null
processingStage.value = ''
}
// 演示模式:上传文件后由用户手动粘贴 Markdown 源码
const initializeManualResult = () => {
results.value = {
markdown: '',
source: '',
mindmap: ''
}
error.value = null
}
// 进度相关
const progressPercent = ref<number | null>(null)
const progressStage = ref('')
const progressElapsedLabel = ref('')
let progressStartedAt = 0
let progressTimer: ReturnType<typeof setInterval> | null = null
// 启动进度轮询
const startProgressPolling = (taskId: string) => {
stopProgressPolling()
progressPercent.value = null
progressStartedAt = Date.now()
progressElapsedLabel.value = '已用时 0 秒'
progressStage.value = '准备中'
progressTimer = setInterval(async () => {
const elapsed = Math.floor((Date.now() - progressStartedAt) / 1000)
progressElapsedLabel.value = elapsed < 60 ? `已用时 ${elapsed} 秒` : `已用时 ${Math.floor(elapsed / 60)} 分 ${elapsed % 60} 秒`
try {
const data = await documentApi.getParseProgress(taskId)
progressPercent.value = data.progress
progressStage.value = data.stage
if (data.status === 'completed' || data.status === 'failed') {
stopProgressPolling()
}
} catch {
// 忽略轮询失败
}
}, 1000)
}
// 停止进度轮询
const stopProgressPolling = () => {
if (progressTimer) {
clearInterval(progressTimer)
progressTimer = null
}
}
// 处理文档转换
const processDocument = async () => {
if (uploadedFiles.value.length === 0) {
error.value = '请先上传文件'
return
}
const firstFileName = uploadedFiles.value[0]?.name.toLowerCase() || ''
if (firstFileName.endsWith('.md')) {
if (!results.value) {
error.value = 'Markdown 文件内容为空,请重新上传'
}
return
}
isProcessing.value = true
error.value = null
processingStage.value = '准备提交解析任务'
// 生成任务ID
const taskId = `task-${Date.now()}-${Math.random().toString(36).slice(2, 8)}`
try {
await documentApi.createParseTask(taskId)
startProgressPolling(taskId)
processingStage.value = '提交文档到解析服务'
const params: ParseParams = {
files: uploadedFiles.value,
parser_resource_id: config.parserResourceId,
output_dir: './output',
lang_list: config.language,
backend: config.backend,
parse_method: config.forceOcr ? 'ocr' : 'auto',
formula_enable: config.formulaEnable,
table_enable: config.tableEnable,
start_page_id: 0,
end_page_id: config.maxPages - 1,
return_md: true,
return_middle_json: false,
response_format_zip: false
}
if (config.backend.includes('http-client') && config.serverUrl) {
params.server_url = config.serverUrl
}
processingStage.value = '服务端正在解析文档'
const response = await documentApi.parseDocument(params, taskId)
if (response.results) {
processingStage.value = '生成 Markdown 和思维导图'
const resultData = Object.values(response.results)[0]
const mdContent = resultData.md_content || ''
const mindmapContent = buildMindmapMarkdown(mdContent, documentTitleFromName(uploadedDisplayFiles.value[0]?.name || ''))
results.value = {
markdown: mdContent,
source: mdContent,
mindmap: mindmapContent
}
}
} catch (err: any) {
error.value = err.message || '转换失败'
} finally {
stopProgressPolling()
processingStage.value = ''
isProcessing.value = false
}
}
// 根据后端类型获取公式标签
const getFormulaLabel = (backend: string) => {
if (backend.startsWith('vlm')) {
return '启用行间公式识别'
} else if (backend === 'pipeline') {
return '启用公式识别'
} else if (backend.startsWith('hybrid')) {
return '启用行内公式识别'
}
return '启用公式识别'
}
// 根据后端类型获取公式说明
const getFormulaInfo = (backend: string) => {
if (backend.startsWith('vlm')) {
return '禁用后,行间公式将显示为图片。'
} else if (backend === 'pipeline') {
return '禁用后,行间公式将显示为图片,行内公式将不会被检测或解析。'
} else if (backend.startsWith('hybrid')) {
return '禁用后,行内公式将不会被检测或解析。'
}
return ''
}
return {
// 数据
uploadedFiles,
uploadedDisplayFiles,
config,
results,
isUploading,
isProcessing,
processingStage,
progressPercent,
progressStage,
progressElapsedLabel,
error,
// 选项
backendOptions,
languageOptions,
// 方法
handleFileUpload,
initializeManualResult,
clearAll,
processDocument,
getFormulaLabel,
getFormulaInfo
}
}

View File

@ -1,87 +0,0 @@
export default {
app: {
title: 'Multimodal Mind Map Assistant'
},
common: {
upload: 'Upload',
clear: 'Clear',
convert: 'Convert',
cancel: 'Cancel',
confirm: 'Confirm',
loading: 'Loading...',
success: 'Success',
error: 'Error',
warning: 'Warning'
},
upload: {
title: 'Please upload PDF or image',
placeholder: 'Click to upload or drag files here',
supportedTypes: 'Supported file types: PDF, PNG, JPG, JPEG',
maxSize: 'Maximum file size: 100MB'
},
config: {
title: 'Configuration Options',
maxPages: 'Max convert pages',
backend: 'Backend',
serverUrl: 'Server URL',
serverUrlInfo: 'OpenAI-compatible server URL for http-client backend.',
recognitionOptions: 'Recognition Options',
tableEnable: 'Enable table recognition',
tableInfo: 'If disabled, tables will be shown as images.',
formulaLabelVlm: 'Enable display formula recognition',
formulaLabelPipeline: 'Enable formula recognition',
formulaLabelHybrid: 'Enable inline formula recognition',
formulaInfoVlm: 'If disabled, display formulas will be shown as images.',
formulaInfoPipeline: 'If disabled, display formulas will be shown as images, and inline formulas will not be detected or parsed.',
formulaInfoHybrid: 'If disabled, inline formulas will not be detected or parsed.',
ocrLanguage: 'OCR Language',
ocrLanguageInfo: 'Select the OCR language for image-based PDFs and images.',
forceOcr: 'Force enable OCR',
forceOcrInfo: 'Enable only if the result is extremely poor. Requires correct OCR language.',
backendInfoVlm: 'High-precision parsing via VLM, supports Chinese and English documents only.',
backendInfoPipeline: 'Traditional Multi-model pipeline parsing, supports multiple languages, hallucination-free.',
backendInfoHybrid: 'High-precision hybrid parsing, supports multiple languages.',
backendInfoDefault: 'Select the backend engine for document parsing.'
},
results: {
title: 'Convert Result',
tabs: {
markdown: 'Markdown Rendering',
source: 'Markdown Source',
mindmap: 'Mind Map'
},
download: 'Download Result',
noResults: 'No conversion results yet'
},
languages: {
ch: 'Chinese (Simplified)',
en: 'English',
korean: 'Korean',
japan: 'Japanese',
chinese_cht: 'Chinese (Traditional)',
ta: 'Tamil',
te: 'Telugu',
ka: 'Kannada',
th: 'Thai',
el: 'Greek',
latin: 'Latin Languages',
arabic: 'Arabic Languages',
east_slavic: 'East Slavic Languages',
cyrillic: 'Cyrillic Languages',
devanagari: 'Devanagari Languages'
},
backends: {
pipeline: 'Traditional Pipeline Parsing',
'vlm-auto-engine': 'VLM Local Engine',
'hybrid-auto-engine': 'Hybrid Local Engine',
'vlm-http-client': 'VLM Remote Client',
'hybrid-http-client': 'Hybrid Remote Client'
},
errors: {
uploadFailed: 'File upload failed',
conversionFailed: 'Conversion failed',
invalidFileType: 'Unsupported file type',
fileSizeExceeded: 'File size exceeded limit',
networkError: 'Network connection error'
}
}

View File

@ -1,2 +0,0 @@
export { default as zh } from './zh'
export { default as en } from './en'

View File

@ -1,87 +0,0 @@
export default {
app: {
title: '多模态思维导图助手'
},
common: {
upload: '上传',
clear: '清除',
convert: '转换',
cancel: '取消',
confirm: '确定',
loading: '加载中...',
success: '成功',
error: '错误',
warning: '警告'
},
upload: {
title: '请上传 PDF 或图片',
placeholder: '点击上传或将文件拖拽到此处',
supportedTypes: '支持的文件类型: PDF, PNG, JPG, JPEG',
maxSize: '最大文件大小: 100MB'
},
config: {
title: '配置选项',
maxPages: '最大转换页数',
backend: '解析后端',
serverUrl: '服务器地址',
serverUrlInfo: 'http-client 后端的 OpenAI 兼容服务器地址。',
recognitionOptions: '识别选项',
tableEnable: '启用表格识别',
tableInfo: '禁用后,表格将显示为图片。',
formulaLabelVlm: '启用行间公式识别',
formulaLabelPipeline: '启用公式识别',
formulaLabelHybrid: '启用行内公式识别',
formulaInfoVlm: '禁用后,行间公式将显示为图片。',
formulaInfoPipeline: '禁用后,行间公式将显示为图片,行内公式将不会被检测或解析。',
formulaInfoHybrid: '禁用后,行内公式将不会被检测或解析。',
ocrLanguage: 'OCR 语言',
ocrLanguageInfo: '为扫描版 PDF 和图片选择 OCR 语言。',
forceOcr: '强制启用 OCR',
forceOcrInfo: '仅在识别效果极差时启用,需选择正确的 OCR 语言。',
backendInfoVlm: '多模态大模型高精度解析,仅支持中英文文档。',
backendInfoPipeline: '传统多模型管道解析,支持多语言,无幻觉。',
backendInfoHybrid: '高精度混合解析,支持多语言。',
backendInfoDefault: '选择文档解析的后端引擎。'
},
results: {
title: '转换结果',
tabs: {
markdown: 'Markdown 渲染',
source: 'Markdown 源码',
mindmap: '思维导图'
},
download: '下载结果',
noResults: '暂无转换结果'
},
languages: {
ch: '中文(简体)',
en: '英语',
korean: '韩语',
japan: '日语',
chinese_cht: '中文(繁体)',
ta: '泰米尔语',
te: '泰卢固语',
ka: '卡纳达语',
th: '泰语',
el: '希腊语',
latin: '拉丁语系',
arabic: '阿拉伯语系',
east_slavic: '东斯拉夫语系',
cyrillic: '西里尔语系',
devanagari: '梵文字母语系'
},
backends: {
pipeline: '传统管道解析',
'vlm-auto-engine': 'VLM本地引擎',
'hybrid-auto-engine': '混合本地引擎',
'vlm-http-client': 'VLM远程客户端',
'hybrid-http-client': '混合远程客户端'
},
errors: {
uploadFailed: '文件上传失败',
conversionFailed: '转换失败',
invalidFileType: '不支持的文件类型',
fileSizeExceeded: '文件大小超出限制',
networkError: '网络连接错误'
}
}

View File

@ -1,21 +0,0 @@
import { createApp } from 'vue'
import ElementPlus from 'element-plus'
import 'element-plus/dist/index.css'
import { createI18n } from 'vue-i18n'
import App from './App.vue'
import { zh, en } from './locales'
const i18n = createI18n({
legacy: false,
locale: navigator.language.includes('zh') ? 'zh' : 'en',
fallbackLocale: 'en',
messages: {
zh,
en
}
})
const app = createApp(App)
app.use(ElementPlus)
app.use(i18n)
app.mount('#app')

View File

@ -1,86 +0,0 @@
import { onBeforeUnmount, ref, watch, type Ref } from 'vue'
import { hubCall } from '@/api/hub'
const pattern = () => /\/admin\/api\/workspace\/[^/\s"<>]+\/ai_hub\/assets\/([a-f0-9-]{36})\/content/g
const cache = new Map<string, Blob>()
const pending = new Map<string, Promise<Blob>>()
let cachedBytes = 0
async function imageBlob(url: string, id: string) {
if (cache.has(url)) return cache.get(url)!
if (pending.has(url)) return pending.get(url)!
const request = hubCall('assets', 'get', undefined, id).then(result => {
if (!(result.content instanceof ArrayBuffer) || !/^image\/(png|jpeg|gif|webp|bmp)$/.test(result.mime)) throw new Error('图片加载失败')
const blob = new Blob([result.content], { type: result.mime })
while (cache.size && cachedBytes + blob.size > 64 * 1024 * 1024) {
const oldest = cache.keys().next().value!
cachedBytes -= cache.get(oldest)!.size
cache.delete(oldest)
}
cache.set(url, blob); cachedBytes += blob.size
return blob
}).finally(() => pending.delete(url))
pending.set(url, request)
return request
}
export function hasImageAssets(content: string) { return pattern().test(content) }
async function replaceAssets(content: string, convert: (blob: Blob) => Promise<string> | string) {
const replacements = new Map<string, string>()
for (const match of content.matchAll(pattern())) {
if (!replacements.has(match[0])) replacements.set(match[0], await convert(await imageBlob(match[0], match[1])))
}
return content.replace(pattern(), url => replacements.get(url)!)
}
export function useImageAssets(source: () => string): { content: Ref<string>; error: Ref<string>; retry: () => void } {
const content = ref(''), error = ref(''), revision = ref(0)
let generation = 0, urls: string[] = []
watch([source, revision], async ([value]) => {
const current = ++generation, nextUrls: string[] = []
error.value = ''
try {
const resolved = await replaceAssets(String(value), blob => {
const url = URL.createObjectURL(blob); nextUrls.push(url); return url
})
if (current === generation) {
const previous = urls; urls = nextUrls; content.value = resolved
// Give both Vue and markmap time to detach the previous DOM.
window.setTimeout(() => previous.forEach(url => URL.revokeObjectURL(url)), 1000)
} else nextUrls.forEach(url => URL.revokeObjectURL(url))
} catch {
nextUrls.forEach(url => URL.revokeObjectURL(url))
if (current === generation) {
content.value = String(value).replace(pattern(), '')
error.value = '图片加载失败,点击重试'
}
}
}, { immediate: true })
onBeforeUnmount(() => { generation++; urls.forEach(url => URL.revokeObjectURL(url)) })
return { content, error, retry: () => revision.value++ }
}
export function blobDataUrl(blob: Blob): Promise<string> {
return new Promise((resolve, reject) => {
const reader = new FileReader()
reader.onload = () => resolve(String(reader.result))
reader.onerror = () => reject(new Error('图片读取失败'))
reader.readAsDataURL(blob)
})
}
// Standalone HTML and SVG must not contain browser-session Blob URLs.
export async function embedImages(html: string) {
const document = new DOMParser().parseFromString(html, 'text/html')
for (const image of Array.from(document.querySelectorAll('img'))) {
if (image.src.startsWith('blob:')) image.setAttribute('src', await blobDataUrl(await (await fetch(image.src)).blob()))
}
return document.body.innerHTML
}
export async function exportMarkdownBundle(markdown: string) {
const result = await hubCall('exports', 'post', { markdown })
if (!(result.content instanceof ArrayBuffer) || result.mime !== 'application/zip') throw new Error('导出失败')
return new Blob([result.content], { type: result.mime })
}

View File

@ -1,55 +0,0 @@
import MarkdownIt from 'markdown-it'
import type Token from 'markdown-it/lib/token.mjs'
const md = new MarkdownIt({
html: true,
linkify: true,
typographer: true,
breaks: true
})
type AstNode = {
type: string
tag?: string
nesting?: number
level?: number
content?: string
markup?: string
info?: string
block?: boolean
hidden?: boolean
map?: number[] | null
attrs?: Record<string, string>
children?: AstNode[]
}
function tokenToAstNode(token: Token): AstNode {
return {
type: token.type,
tag: token.tag || undefined,
nesting: token.nesting,
level: token.level,
content: token.content || undefined,
markup: token.markup || undefined,
info: token.info || undefined,
block: token.block,
hidden: token.hidden,
map: token.map || null,
attrs: token.attrs ? Object.fromEntries(token.attrs) : undefined,
children: token.children?.map(tokenToAstNode)
}
}
export function markdownToAstString(markdown: string): string {
if (!markdown.trim()) {
return ''
}
const tokens = md.parse(markdown, {})
const ast = {
type: 'root',
children: tokens.map(tokenToAstNode)
}
return JSON.stringify(ast, null, 2)
}

View File

@ -1,45 +0,0 @@
export interface ReplacementRule {
id: string
search: string
replace: string
}
export interface MarkdownTemplateRenderResult {
markdown: string
error: string | null
}
export function createEmptyReplacementRule(): ReplacementRule {
return {
id: `${Date.now()}_${Math.random().toString(16).slice(2)}`,
search: '',
replace: ''
}
}
export function renderMarkdownTemplate(template: string, rules: ReplacementRule[]): MarkdownTemplateRenderResult {
if (!template.trim()) {
return {
markdown: '',
error: null
}
}
const activeRules = rules.filter((rule) => rule.search.trim())
try {
const markdown = activeRules.reduce((current, rule) => {
return current.split(rule.search).join(rule.replace)
}, template)
return {
markdown,
error: null
}
} catch (error) {
return {
markdown: template,
error: `替换规则执行失败:${(error as Error).message}`
}
}
}

View File

@ -1,175 +0,0 @@
import MarkdownIt from 'markdown-it'
import type Token from 'markdown-it/lib/token.mjs'
const ROOT_TITLE = '文档思维导图'
const MAX_HEADING_LEVEL = 6
const md = new MarkdownIt({
html: true,
linkify: true,
typographer: true,
breaks: true
})
function normalizeNodeText(text: string) {
return text
.replace(/\r?\n/g, ' ')
.replace(/\s+/g, ' ')
.trim()
}
function clampHeadingLevel(level: number) {
return Math.min(Math.max(level, 1), MAX_HEADING_LEVEL)
}
function pushHeading(result: string[], level: number, text: string) {
const normalized = normalizeNodeText(text)
if (!normalized) return
result.push(`${'#'.repeat(clampHeadingLevel(level))} ${normalized}`)
}
function getHeadingLevel(token: Token) {
const tagLevel = Number(token.tag.replace(/^h/i, ''))
return Number.isFinite(tagLevel) && tagLevel > 0 ? tagLevel : 1
}
function getNextInlineContent(tokens: Token[], index: number) {
const next = tokens[index + 1]
return next?.type === 'inline' ? next.content : ''
}
function getFenceContent(token: Token) {
const info = normalizeNodeText(token.info || '')
const content = token.content.trim()
if (!content) return info ? `代码块 ${info}` : '代码块'
return info ? `代码块 ${info}: ${content}` : `代码块: ${content}`
}
function collectTableRows(tokens: Token[], startIndex: number) {
const rows: string[][] = []
let currentRow: string[] | null = null
let endIndex = startIndex
for (let i = startIndex + 1; i < tokens.length; i += 1) {
const token = tokens[i]
endIndex = i
if (token.type === 'table_close') break
if (token.type === 'tr_open') {
currentRow = []
continue
}
if (token.type === 'tr_close') {
if (currentRow && currentRow.some((cell) => cell.trim())) {
rows.push(currentRow)
}
currentRow = null
continue
}
if ((token.type === 'th_open' || token.type === 'td_open') && currentRow) {
currentRow.push(getNextInlineContent(tokens, i))
}
}
return { rows, endIndex }
}
export function documentTitleFromName(name: string) {
return name.replace(/\.[^.\\/]+$/, '').trim() || ROOT_TITLE
}
export function buildMindmapMarkdown(markdown: string, title = ROOT_TITLE) {
if (!markdown.trim()) return ''
const tokens = md.parse(markdown, {})
const rootTitle = normalizeNodeText(title).replace(/([\\`*{}\[\]()#+.!_<>])/g, '\\$1')
const result: string[] = [`# ${rootTitle || ROOT_TITLE}`]
let currentHeadingLevel = 1
let listDepth = 0
for (let i = 0; i < tokens.length; i += 1) {
const token = tokens[i]
if (token.type === 'heading_open') {
currentHeadingLevel = clampHeadingLevel(getHeadingLevel(token) + 1)
pushHeading(result, currentHeadingLevel, getNextInlineContent(tokens, i) || '未命名章节')
i += 2
continue
}
if (token.type === 'bullet_list_open' || token.type === 'ordered_list_open') {
listDepth += 1
continue
}
if (token.type === 'bullet_list_close' || token.type === 'ordered_list_close') {
listDepth = Math.max(0, listDepth - 1)
continue
}
if (token.type === 'paragraph_open') {
const level = currentHeadingLevel + Math.max(listDepth, 1)
pushHeading(result, level, getNextInlineContent(tokens, i))
i += 2
continue
}
if (token.type === 'fence' || token.type === 'code_block') {
pushHeading(result, currentHeadingLevel + 1, getFenceContent(token))
continue
}
if (token.type === 'html_block') {
pushHeading(result, currentHeadingLevel + 1, token.content)
continue
}
if (token.type === 'blockquote_open') {
pushHeading(result, currentHeadingLevel + 1, '引用')
currentHeadingLevel = clampHeadingLevel(currentHeadingLevel + 1)
continue
}
if (token.type === 'blockquote_close') {
currentHeadingLevel = Math.max(1, currentHeadingLevel - 1)
continue
}
if (token.type === 'hr') {
pushHeading(result, currentHeadingLevel + 1, '---')
continue
}
if (token.type === 'table_open') {
const tableLevel = currentHeadingLevel + 1
const { rows, endIndex } = collectTableRows(tokens, i)
pushHeading(result, tableLevel, '表格')
rows.forEach((row) => {
pushHeading(result, tableLevel + 1, row.map(normalizeNodeText).join(' | '))
})
i = endIndex
}
}
return result.join('\n')
}
export function replaceFirstMindmapText(markdown: string, oldText: string, newText: string) {
const normalizedOld = oldText.replace(/^(引用|表格|代码块)(\s|:|:)?/, '').trim()
if (!normalizedOld || !newText.trim()) return markdown
const escaped = normalizedOld.replace(/[.*+?^${}()|[\]\\]/g, '\\$&')
const directMatch = new RegExp(escaped)
if (directMatch.test(markdown)) {
return markdown.replace(directMatch, newText.trim())
}
const lines = markdown.split(/\r?\n/)
const targetLineIndex = lines.findIndex((line) => normalizeNodeText(line).includes(normalizedOld))
if (targetLineIndex >= 0) {
lines[targetLineIndex] = lines[targetLineIndex].replace(normalizedOld, newText.trim())
return lines.join('\n')
}
return markdown
}

View File

@ -1,46 +0,0 @@
export interface MindmapNode {
content?: string
children?: MindmapNode[]
payload?: { fold?: number; [key: string]: unknown }
}
export interface FoldState { paths: Map<string, number>; unique: Map<string, number> }
function identity(node: MindmapNode) {
// Browser Blob URLs change on every preview render; image identity must not.
return (node.content || '').replace(/src="blob:[^"]*"/g, 'src=""').replace(/\s+/g, ' ').trim()
}
function visit(root: MindmapNode, fn: (node: MindmapNode, path: string, text: string) => void) {
const walk = (node: MindmapNode, path: string) => {
fn(node, path, identity(node))
const occurrences = new Map<string, number>()
for (const child of node.children || []) {
const text = identity(child), ordinal = occurrences.get(text) || 0
occurrences.set(text, ordinal + 1)
walk(child, `${path}/${JSON.stringify([text, ordinal])}`)
}
}
walk(root, 'root')
}
export function captureFolds(root: MindmapNode): FoldState {
const paths = new Map<string, number>(), unique = new Map<string, number>(), repeated = new Set<string>()
visit(root, (node, path, text) => {
const fold = node.payload?.fold || 0
paths.set(path, fold)
if (unique.has(text)) repeated.add(text)
else unique.set(text, fold)
})
repeated.forEach(text => unique.delete(text))
return { paths, unique }
}
export function restoreFolds(root: MindmapNode, saved?: FoldState) {
const counts = new Map<string, number>()
visit(root, (_node, _path, text) => counts.set(text, (counts.get(text) || 0) + 1))
visit(root, (node, path, text) => {
const fold = saved?.paths.get(path) ?? (counts.get(text) === 1 ? saved?.unique.get(text) : undefined) ?? 0
node.payload = { ...node.payload, fold: node.children?.length ? fold : 0 }
})
}

View File

@ -1,79 +0,0 @@
import axios from 'axios'
const getApiBaseURL = () => {
const pathname = window.location.pathname
const basePath = pathname === '/mindmap' || pathname.startsWith('/mindmap/') ? '/mindmap' : ''
return `${basePath}/api`
}
// ?? axios ??
const request = axios.create({
baseURL: getApiBaseURL(),
timeout: 300000, // 5????
headers: {
'Content-Type': 'application/json'
}
})
// ?????
request.interceptors.request.use(
(config) => {
// ?????????token?
return config
},
(error) => {
return Promise.reject(error)
}
)
// ?????
request.interceptors.response.use(
(response) => {
return response.data
},
(error) => {
// ??????
if (error.response) {
const serverMessage = error.response.data?.error || error.response.data?.detail
if (serverMessage) {
error.message = serverMessage
return Promise.reject(error)
}
switch (error.response.status) {
case 400:
error.message = '??????'
break
case 401:
error.message = '?????????'
break
case 403:
error.message = '????'
break
case 404:
error.message = '???????'
break
case 500:
error.message = '???????'
break
case 502:
error.message = '????'
break
case 503:
error.message = '?????'
break
case 504:
error.message = '????'
break
default:
error.message = `????${error.response.status}`
}
} else {
error.message = '??????'
}
return Promise.reject(error)
}
)
export default request

View File

@ -1,10 +0,0 @@
// crypto.randomUUID 仅在安全上下文(HTTPS/localhost)可用;管理平台可能以 HTTP 访问,需要降级。
export function uuid(): string {
if (typeof crypto !== 'undefined' && typeof crypto.randomUUID === 'function') return crypto.randomUUID()
const buf = new Uint8Array(16)
crypto.getRandomValues(buf)
buf[6] = (buf[6] & 0x0f) | 0x40
buf[8] = (buf[8] & 0x3f) | 0x80
const hex = Array.from(buf, b => b.toString(16).padStart(2, '0')).join('')
return `${hex.slice(0, 8)}-${hex.slice(8, 12)}-${hex.slice(12, 16)}-${hex.slice(16, 20)}-${hex.slice(20)}`
}

View File

@ -1,143 +0,0 @@
/**
* Word 转 PDF 工具
* 使用 mammoth 将 Word 文档转换为 HTML,然后使用 html2pdf.js 转换为 PDF
*/
declare global {
interface Window {
mammoth: {
convertToHtml: (arrayBuffer: ArrayBuffer) => Promise<{ value: string }>;
convertToHtmlOptions: (options: object) => object;
};
html2pdf: {
(): {
set: (options: object) => ReturnType<typeof html2pdf>;
from: (element: HTMLElement | string) => ReturnType<typeof html2pdf>;
save: () => Promise<string>;
outputPdf: (type: string) => Promise<Blob>;
};
};
}
}
/**
* 动态加载 mammoth 库
*/
async function loadMammoth(): Promise<void> {
if (typeof window.mammoth === 'undefined') {
return new Promise((resolve, reject) => {
const script = document.createElement('script');
script.src = 'https://cdn.jsdelivr.net/npm/mammoth@1.6.0/mammoth.browser.min.js';
script.onload = () => resolve();
script.onerror = () => reject(new Error('Failed to load mammoth library'));
document.head.appendChild(script);
});
}
}
/**
* 动态加载 html2pdf 库
*/
export async function loadHtml2Pdf(): Promise<void> {
if (typeof window.html2pdf === 'undefined') {
return new Promise((resolve, reject) => {
const script = document.createElement('script');
script.src = 'https://cdn.jsdelivr.net/npm/html2pdf.js@0.10.1/dist/html2pdf.bundle.min.js';
script.onload = () => resolve();
script.onerror = () => reject(new Error('Failed to load html2pdf library'));
document.head.appendChild(script);
});
}
}
/**
* 将 Word 文档转换为 PDF
* @param file Word 文件
* @param onProgress 进度回调
* @returns PDF 文件
*/
export async function convertWordToPdf(
file: File,
onProgress?: (progress: number) => void
): Promise<File> {
console.log('convertWordToPdf called with file:', file.name, file.size)
onProgress?.(10);
// 加载必要的库
console.log('Loading mammoth library...')
await loadMammoth();
console.log('Mammoth library loaded')
onProgress?.(20);
console.log('Loading html2pdf library...')
await loadHtml2Pdf();
console.log('html2pdf library loaded')
onProgress?.(30);
// 读取 Word 文件
console.log('Reading Word file...')
const arrayBuffer = await file.arrayBuffer();
console.log('Word file read successfully')
onProgress?.(50);
// 转换为 HTML
console.log('Converting Word to HTML...')
const result = await window.mammoth.convertToHtml({ arrayBuffer });
console.log('Word to HTML conversion successful, HTML length:', result.value.length)
const html = result.value;
onProgress?.(70);
// 创建临时容器
console.log('Creating temporary container...')
const container = document.createElement('div');
container.innerHTML = `
<div style="font-family: 'Microsoft YaHei', Arial, sans-serif; padding: 20px;">
${html}
</div>
`;
container.style.width = '595px'; // A4 width in pixels at 72 DPI
document.body.appendChild(container);
console.log('Temporary container created')
onProgress?.(80);
// 生成 PDF
console.log('Generating PDF...')
const pdfOptions = {
margin: 10,
filename: file.name.replace(/\.(docx?|doc)$/i, '.pdf'),
image: { type: 'jpeg', quality: 0.98 },
html2canvas: { scale: 2, useCORS: true },
jsPDF: { unit: 'mm', format: 'a4', orientation: 'portrait' }
};
const pdfBlob = await window.html2pdf()
.set(pdfOptions)
.from(container)
.outputPdf('blob');
console.log('PDF generated successfully, blob size:', pdfBlob.size)
onProgress?.(95);
// 清理
document.body.removeChild(container);
console.log('Temporary container removed')
// 转换为 File 对象
const pdfFile = new File([pdfBlob], pdfOptions.filename, {
type: 'application/pdf'
});
console.log('PDF file created:', pdfFile.name, pdfFile.size)
onProgress?.(100);
return pdfFile;
}
/**
* 检查文件是否为 Word 文档
*/
export function isWordFile(file: File): boolean {
const fileName = file.name.toLowerCase();
return fileName.endsWith('.docx') || fileName.endsWith('.doc');
}

File diff suppressed because it is too large Load Diff

View File

@ -1,15 +0,0 @@
#!/bin/bash
# MinerU Vue.js Web UI 启动脚本
echo "🚀 启动 MinerU Vue.js Web UI..."
# 检查是否已安装依赖
if [ ! -d "node_modules" ]; then
echo "📦 安装依赖..."
npm install
fi
# 启动开发服务器
echo "🌐 启动前端开发服务器..."
npm run dev -- --host

View File

@ -1,25 +0,0 @@
{
"compilerOptions": {
"target": "ES2020",
"useDefineForClassFields": true,
"module": "ESNext",
"lib": ["ES2020", "DOM", "DOM.Iterable"],
"skipLibCheck": true,
"moduleResolution": "bundler",
"allowImportingTsExtensions": true,
"resolveJsonModule": true,
"isolatedModules": true,
"noEmit": true,
"jsx": "preserve",
"strict": true,
"noUnusedLocals": true,
"noUnusedParameters": true,
"noFallthroughCasesInSwitch": true,
"baseUrl": ".",
"paths": {
"@/*": ["src/*"]
}
},
"include": ["src/**/*.ts", "src/**/*.tsx", "src/**/*.vue"],
"references": [{ "path": "./tsconfig.node.json" }]
}

View File

@ -1,9 +0,0 @@
{
"compilerOptions": {
"composite": true,
"skipLibCheck": true,
"module": "ESNext",
"moduleResolution": "bundler"
},
"include": ["vite.config.ts"]
}

View File

@ -1,39 +0,0 @@
import { defineConfig } from 'vite'
import vue from '@vitejs/plugin-vue'
import { resolve } from 'path'
export default defineConfig({
base: './',
plugins: [vue()],
resolve: {
alias: {
'@': resolve(__dirname, 'src')
}
},
server: {
port: 3000,
proxy: {
'/api': {
target: process.env.VITE_API_PROXY || 'http://127.0.0.1:8000',
changeOrigin: true
},
'/mindmap/api': {
target: process.env.VITE_API_PROXY || 'http://127.0.0.1:8000',
changeOrigin: true,
rewrite: path => path.replace(/^\/mindmap\/api/, '/api')
}
}
},
build: {
outDir: 'dist',
assetsDir: 'assets',
rollupOptions: {
output: {
manualChunks: {
vendor: ['vue', 'element-plus'],
markdown: ['markdown-it']
}
}
}
}
})