feat(思维导图助手):升级mineru解析到2.7.6版本

develop-mineru-2.7.6
panyy 2026-11-09 11:02:55 +08:00
parent 846739cf77
commit 94d2f22b3c
55 changed files with 2602 additions and 725 deletions

View File

@ -15,6 +15,11 @@ tmp
node_modules
web_ui/node_modules
web_ui/dist
mineru/cli/static/web
test_files
tests
build
**/.env
**/.env.*
**/.venv
**/venv*
**/*.egg-info

4
.gitattributes vendored
View File

@ -2,4 +2,6 @@
*.mjs linguist-vendored
*.html linguist-documentation
*.css linguist-vendored
*.scss linguist-vendored
*.scss linguist-vendored
*.sh text eol=lf
docker/** text eol=lf

4
.gitignore vendored
View File

@ -52,5 +52,7 @@ _build/
output/
web_ui/node_modules
build/
*.egg-info/
web_ui/dist
web_ui/dist

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

@ -1,108 +1,27 @@
# ==========================================
# 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)
# ==========================================
# MinerU 2.7.6 / Iluvatar CoreX
FROM crpi-vofi3w62lkohhxsp.cn-shanghai.personal.cr.aliyuncs.com/opendatalab-mineru/corex:4.4.0_torch2.7.1_vllm0.11.2_py3.10
ENV MINERU_VLLM_DEVICE=corex
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 DEBIAN_FRONTEND=noninteractive \
MINERU_MODEL_SOURCE=local \
MINERU_MODEL_DIR=/models \
MINERU_TOOLS_CONFIG_JSON=/root/mineru.json \
MINERU_API_MAX_CONCURRENT_REQUESTS=1
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"]
RUN apt-get update && apt-get install -y --no-install-recommends \
fonts-noto-core fonts-noto-cjk fontconfig libgl1 libglib2.0-0 \
libreoffice-writer libreoffice-core && fc-cache -fv && \
apt-get clean && rm -rf /var/lib/apt/lists/*
COPY . /app
# Resolve against the accelerator wheels already provided by the selected base image.
RUN python3 scripts/accelerator_constraints.py /tmp/accelerator-constraints.txt && \
python3 -m pip install 'setuptools>=61,<81' wheel && \
python3 -m pip install --no-build-isolation \
-c /tmp/accelerator-constraints.txt -c docker/constraints-2.7.6.txt '.[core]' && \
python3 scripts/accelerator_constraints.py /tmp/accelerator-constraints.txt --check && \
python3 -m pip check && \
python3 -c "from mineru.version import __version__; assert __version__ == '2.7.6'" && \
chmod +x docker/entrypoint.mineru-api.sh && python3 -m pip cache purge
EXPOSE 8000
ENTRYPOINT ["/app/docker/entrypoint.mineru-api.sh"]
CMD ["mineru-api", "--host", "0.0.0.0", "--port", "8000"]

View File

@ -1,65 +1,27 @@
# MinerU 2.7.6 / Iluvatar CoreX
FROM crpi-vofi3w62lkohhxsp.cn-shanghai.personal.cr.aliyuncs.com/opendatalab-mineru/corex:4.4.0_torch2.7.1_vllm0.11.2_py3.10
ENV MINERU_VLLM_DEVICE=corex
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 DEBIAN_FRONTEND=noninteractive \
MINERU_MODEL_SOURCE=local \
MINERU_MODEL_DIR=/models \
MINERU_TOOLS_CONFIG_JSON=/root/mineru.json \
MINERU_API_MAX_CONCURRENT_REQUESTS=1
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"]
RUN apt-get update && apt-get install -y --no-install-recommends \
fonts-noto-core fonts-noto-cjk fontconfig libgl1 libglib2.0-0 \
libreoffice-writer libreoffice-core && fc-cache -fv && \
apt-get clean && rm -rf /var/lib/apt/lists/*
COPY . /app
# Resolve against the accelerator wheels already provided by the selected base image.
RUN python3 scripts/accelerator_constraints.py /tmp/accelerator-constraints.txt && \
python3 -m pip install 'setuptools>=61,<81' wheel && \
python3 -m pip install --no-build-isolation \
-c /tmp/accelerator-constraints.txt -c docker/constraints-2.7.6.txt '.[core]' && \
python3 scripts/accelerator_constraints.py /tmp/accelerator-constraints.txt --check && \
python3 -m pip check && \
python3 -c "from mineru.version import __version__; assert __version__ == '2.7.6'" && \
chmod +x docker/entrypoint.mineru-api.sh && python3 -m pip cache purge
EXPOSE 8000
ENTRYPOINT ["/app/docker/entrypoint.mineru-api.sh"]
CMD ["mineru-api", "--host", "0.0.0.0", "--port", "8000"]

View File

@ -1,29 +1,27 @@
# Base image containing the vLLM inference environment, requiring amd64(x86-64) CPU + iluvatar GPU.
# MinerU 2.7.6 / Iluvatar CoreX
FROM crpi-vofi3w62lkohhxsp.cn-shanghai.personal.cr.aliyuncs.com/opendatalab-mineru/corex:4.4.0_torch2.7.1_vllm0.11.2_py3.10
ENV MINERU_VLLM_DEVICE=corex
# Install Noto fonts for Chinese characters
RUN apt-get update && \
apt-get install -y \
fonts-noto-core \
fonts-noto-cjk \
fontconfig && \
fc-cache -fv && \
apt-get clean && \
rm -rf /var/lib/apt/lists/*
# Install mineru latest
RUN python3 -m pip install -U pip -i https://mirrors.aliyun.com/pypi/simple && \
python3 -m pip install 'mineru[core]>=2.7.4' \
numpy==1.26.4 \
opencv-python==4.11.0.86 \
-i https://mirrors.aliyun.com/pypi/simple && \
python3 -m pip install --force-reinstall --no-cache-dir "pydantic>=2.12,<3" -i https://mirrors.aliyun.com/pypi/simple && \
python3 -c "import pydantic; print('pydantic', pydantic.__version__); from pydantic import TypeAdapter; print('pydantic TypeAdapter ok')" && \
python3 -m pip cache purge
# Download models and update the configuration file
RUN /bin/bash -c "mineru-models-download -s modelscope -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 \"$@\"", "--"]
ENV DEBIAN_FRONTEND=noninteractive \
MINERU_MODEL_SOURCE=local \
MINERU_MODEL_DIR=/models \
MINERU_TOOLS_CONFIG_JSON=/root/mineru.json \
MINERU_API_MAX_CONCURRENT_REQUESTS=1
WORKDIR /app
RUN apt-get update && apt-get install -y --no-install-recommends \
fonts-noto-core fonts-noto-cjk fontconfig libgl1 libglib2.0-0 \
libreoffice-writer libreoffice-core && fc-cache -fv && \
apt-get clean && rm -rf /var/lib/apt/lists/*
COPY . /app
# Resolve against the accelerator wheels already provided by the selected base image.
RUN python3 scripts/accelerator_constraints.py /tmp/accelerator-constraints.txt && \
python3 -m pip install 'setuptools>=61,<81' wheel && \
python3 -m pip install --no-build-isolation \
-c /tmp/accelerator-constraints.txt -c docker/constraints-2.7.6.txt '.[core]' && \
python3 scripts/accelerator_constraints.py /tmp/accelerator-constraints.txt --check && \
python3 -m pip check && \
python3 -c "from mineru.version import __version__; assert __version__ == '2.7.6'" && \
chmod +x docker/entrypoint.mineru-api.sh && python3 -m pip cache purge
EXPOSE 8000
ENTRYPOINT ["/app/docker/entrypoint.mineru-api.sh"]
CMD ["mineru-api", "--host", "0.0.0.0", "--port", "8000"]

View File

@ -1,86 +1,27 @@
# ==========================================
# 阶段 1: 构建阶段 (Builder) - 负责前端 UI 编译
# ==========================================
FROM docker.m.daocloud.io/ubuntu:22.04 AS builder
ENV DEBIAN_FRONTEND=noninteractive
# 替换为阿里云镜像源并安装 Node.js [cite: 8, 9]
RUN sed -i 's/archive.ubuntu.com/mirrors.aliyun.com/g' /etc/apt/sources.list && \
apt-get update && apt-get install -y curl && \
curl -fsSL https://deb.nodesource.com/setup_18.x | bash - && \
apt-get install -y nodejs && \
apt-get clean
WORKDIR /app
COPY . .
# 构建前端 Web UI [cite: 12]
WORKDIR /app/web_ui
RUN npm install && npm run build
# ==========================================
# 阶段 2: 运行阶段 (Runtime) - 天数智芯专用环境
# ==========================================
# 使用天数智芯官方适配镜像,该镜像内置了 Python 3.10.18 和 CoreX 驱动 [cite: 15]
# MinerU 2.7.6 / Iluvatar CoreX
FROM crpi-vofi3w62lkohhxsp.cn-shanghai.personal.cr.aliyuncs.com/opendatalab-mineru/corex:4.4.0_torch2.7.1_vllm0.11.2_py3.10
ENV MINERU_VLLM_DEVICE=corex
ENV DEBIAN_FRONTEND=noninteractive
ENV DEBIAN_FRONTEND=noninteractive \
MINERU_MODEL_SOURCE=local \
MINERU_MODEL_DIR=/models \
MINERU_TOOLS_CONFIG_JSON=/root/mineru.json \
MINERU_API_MAX_CONCURRENT_REQUESTS=1
WORKDIR /app
# 1. 安装字体与 LibreOffice 依赖 [cite: 15]
RUN apt-get update && \
apt-get install -y \
fonts-noto-core \
fonts-noto-cjk \
fontconfig \
libgl1 \
libreoffice-writer \
libreoffice-core && \
fc-cache -fv && \
apt-get clean && \
rm -rf /var/lib/apt/lists/*
# 2. 核心环境修复:确保在 CoreX 的 Python 路径下安装依赖
# 注意:使用 python3 -m pip 确保安装到 3.10.18 环境,避免 uvicorn 找不到 [cite: 16]
RUN python3 -m pip install -U pip -i https://mirrors.aliyun.com/pypi/simple && \
python3 -m pip install \
'mineru[core]>=2.7.4' \
"uvicorn" \
"fastapi" \
"python-multipart" \
"modelscope>=1.26.0" \
"huggingface-hub>=0.32.4" \
"mineru-vl-utils>=0.1.19.1" \
"qwen-vl-utils>=0.0.14" \
numpy==1.26.4 \
opencv-python==4.11.0.86 \
-i https://mirrors.aliyun.com/pypi/simple
# 3. 拷贝源码及第一阶段的前端产物 [cite: 10, 13]
COPY . .
RUN mkdir -p /app/mineru/cli/static/web && \
cp -r /app/web_ui/dist/* /app/mineru/cli/static/web/
# 4. 下载模型权重 (离线模式必备) [cite: 16]
RUN /bin/bash -c "mineru-models-download -s modelscope -m all"
# 5. 注入适配天数 GPU 的配置文件 (开启 vLLM 推理) [cite: 13]
RUN mkdir -p /root/ && \
echo '{ \
"models-dir": "/root/.cache/modelscope/hub/models", \
"device-mode": "cuda", \
"vlm-config": { \
"kind": "vllm", \
"precision": "fp16" \
} \
}' > /root/magic-pdf.json
# 6. 设置环境变量 [cite: 17]
ENV MINERU_MODEL_SOURCE=local
ENV PYTHONPATH=/app
RUN apt-get update && apt-get install -y --no-install-recommends \
fonts-noto-core fonts-noto-cjk fontconfig libgl1 libglib2.0-0 \
libreoffice-writer libreoffice-core && fc-cache -fv && \
apt-get clean && rm -rf /var/lib/apt/lists/*
COPY . /app
# Resolve against the accelerator wheels already provided by the selected base image.
RUN python3 scripts/accelerator_constraints.py /tmp/accelerator-constraints.txt && \
python3 -m pip install 'setuptools>=61,<81' wheel && \
python3 -m pip install --no-build-isolation \
-c /tmp/accelerator-constraints.txt -c docker/constraints-2.7.6.txt '.[core]' && \
python3 scripts/accelerator_constraints.py /tmp/accelerator-constraints.txt --check && \
python3 -m pip check && \
python3 -c "from mineru.version import __version__; assert __version__ == '2.7.6'" && \
chmod +x docker/entrypoint.mineru-api.sh && python3 -m pip cache purge
EXPOSE 8000
# 7. 启动服务:使用 /bin/bash 包装以加载 CoreX 环境路径,解决二进制执行错误
ENTRYPOINT ["/bin/bash", "-c", "PYTHONPATH=/app exec python3 -m mineru.cli.fast_api --host 0.0.0.0 --port 8000"]
ENTRYPOINT ["/app/docker/entrypoint.mineru-api.sh"]
CMD ["mineru-api", "--host", "0.0.0.0", "--port", "8000"]

View File

@ -1,107 +1,27 @@
# ==========================================
# 阶段 1: 构建阶段 (Builder) - 适配天数智芯 (Iluvatar CoreX)
# ==========================================
FROM crpi-vofi3w62lkohhxsp.cn-shanghai.personal.cr.aliyuncs.com/opendatalab-mineru/corex:4.4.0_torch2.7.1_vllm0.11.2_py3.10 AS builder
ENV DEBIAN_FRONTEND=noninteractive
# 替换为阿里云镜像源(corex base 已预优化,若无匹配则不影响)
RUN sed -i 's/archive.ubuntu.com/mirrors.aliyun.com/g' /etc/apt/sources.list && \
sed -i 's/security.ubuntu.com/mirrors.aliyun.com/g' /etc/apt/sources.list
# 安装构建环境、Node.js(Web UI)、libreoffice、字体等依赖
# corex base 已包含部分字体与 Python,但仍需补充构建工具与 Node.js
RUN apt-get update && \
apt-get install -y \
build-essential curl wget git fontconfig libgl1 \
libreoffice-writer libreoffice-core \
fonts-noto-core fonts-noto-cjk \
python3-pip && \
curl -fsSL https://deb.nodesource.com/setup_18.x | bash - && \
apt-get install -y nodejs && \
fc-cache -fv && \
apt-get clean && \
rm -rf /var/lib/apt/lists/*
# MinerU 2.7.6 / Iluvatar CoreX
FROM crpi-vofi3w62lkohhxsp.cn-shanghai.personal.cr.aliyuncs.com/opendatalab-mineru/corex:4.4.0_torch2.7.1_vllm0.11.2_py3.10
ENV MINERU_VLLM_DEVICE=corex
ENV DEBIAN_FRONTEND=noninteractive \
MINERU_MODEL_SOURCE=local \
MINERU_MODEL_DIR=/models \
MINERU_TOOLS_CONFIG_JSON=/root/mineru.json \
MINERU_API_MAX_CONCURRENT_REQUESTS=1
WORKDIR /app
COPY . .
# 1. 升级基础 Python 构建工具
RUN python3 -m pip install --upgrade pip setuptools wheel -i https://mirrors.aliyun.com/pypi/simple/
RUN python3 -m pip install --no-cache-dir \
"uvicorn[standard]>=0.30" \
"fastapi>=0.115" \
"python-multipart>=0.0.9" \
-i https://mirrors.aliyun.com/pypi/simple/
# 2. 引入 corex.Dockerfile 的 pinned 依赖(解决版本冲突)
RUN python3 -m pip install \
numpy==1.26.4 \
opencv-python==4.11.0.86 \
-i https://mirrors.aliyun.com/pypi/simple/
# 3. 预装项目所需核心依赖(跳过 torch,因为 corex base 已提供 GPU 版)
RUN python3 -m pip install \
"modelscope>=1.26.0" \
"huggingface-hub>=0.32.4" \
"mineru-vl-utils>=0.1.19.1" \
"qwen-vl-utils>=0.0.14" \
"transformers>=4.51.1" \
"accelerate>=1.5.1" \
-i https://mirrors.aliyun.com/pypi/simple/
# 4. 安装项目及所有可选依赖 [all](自动涵盖 doclayout_yolo、layout/vlm 等)
RUN python3 -m pip install -e ".[all]" -i https://mirrors.aliyun.com/pypi/simple/
# 5. 构建阶段预下载所有权重文件(结合 corex 的下载命令 + 配置)
RUN mkdir -p /root/.cache/modelscope/hub/models && \
echo '{"models-dir": "/root/.cache/modelscope/hub/models", "device-mode":"gpu"}' > /root/magic-pdf.json && \
export MINERU_CONFIG_PATH=/root/magic-pdf.json && \
/bin/bash -c "mineru-models-download -s modelscope -m all"
# 6. 构建前端 Web UI
WORKDIR /app/web_ui
RUN npm install && npm run build
WORKDIR /app
RUN mkdir -p mineru/cli/static/web && cp -r web_ui/dist/* mineru/cli/static/web/
# ==========================================
# 阶段 2: 运行阶段 (Runtime)
# ==========================================
FROM crpi-vofi3w62lkohhxsp.cn-shanghai.personal.cr.aliyuncs.com/opendatalab-mineru/corex:4.4.0_torch2.7.1_vllm0.11.2_py3.10 AS runtime
ENV DEBIAN_FRONTEND=noninteractive
# 替换为阿里云镜像源(安全起见)
RUN sed -i 's/archive.ubuntu.com/mirrors.aliyun.com/g' /etc/apt/sources.list && \
sed -i 's/security.ubuntu.com/mirrors.aliyun.com/g' /etc/apt/sources.list || true
RUN apt-get update && \
apt-get install -y libgl1 libreoffice-writer libreoffice-core \
fonts-noto-core fonts-noto-cjk fontconfig python3 python3-pip && \
fc-cache -fv && apt-get clean && rm -rf /var/lib/apt/lists/*
WORKDIR /app
# 从构建阶段拷贝依赖、预下载模型、源码和 Web UI 静态文件
COPY --from=builder /usr/local/lib/python3.10/dist-packages /usr/local/lib/python3.10/dist-packages
COPY --from=builder /usr/local/bin /usr/local/bin
COPY --from=builder /root/.cache/modelscope/hub/models /root/.cache/modelscope/hub/models
COPY --from=builder /app /app
# 核心修复:注入标准的运行时配置文件(适配 GPU)
RUN mkdir -p /root/ && \
echo '{ \
"models-dir": "/root/.cache/modelscope/hub/models", \
"device-mode": "gpu", \
"vlm-config": { \
"kind": "transformers", \
"precision": "fp16" \
} \
}' > /root/magic-pdf.json
RUN apt-get update && apt-get install -y --no-install-recommends \
fonts-noto-core fonts-noto-cjk fontconfig libgl1 libglib2.0-0 \
libreoffice-writer libreoffice-core && fc-cache -fv && \
apt-get clean && rm -rf /var/lib/apt/lists/*
COPY . /app
# Resolve against the accelerator wheels already provided by the selected base image.
RUN python3 scripts/accelerator_constraints.py /tmp/accelerator-constraints.txt && \
python3 -m pip install 'setuptools>=61,<81' wheel && \
python3 -m pip install --no-build-isolation \
-c /tmp/accelerator-constraints.txt -c docker/constraints-2.7.6.txt '.[core]' && \
python3 scripts/accelerator_constraints.py /tmp/accelerator-constraints.txt --check && \
python3 -m pip check && \
python3 -c "from mineru.version import __version__; assert __version__ == '2.7.6'" && \
chmod +x docker/entrypoint.mineru-api.sh && python3 -m pip cache purge
EXPOSE 8000
# 启动服务(结合 corex 的 MINERU_MODEL_SOURCE=local + 原 fast_api 入口)
ENTRYPOINT ["/bin/sh", "-c", "export MINERU_MODEL_SOURCE=local && PYTHONPATH=/app python3 -m mineru.cli.fast_api --host 0.0.0.0 --port 8000"]
ENTRYPOINT ["/app/docker/entrypoint.mineru-api.sh"]
CMD ["mineru-api", "--host", "0.0.0.0", "--port", "8000"]

View File

@ -0,0 +1,8 @@
# Apply together with accelerator constraints generated from the vendor image.
mineru==2.7.6
numpy==1.26.4
opencv-python==4.11.0.86
transformers==4.57.3
mineru-vl-utils==0.1.19.1
qwen-vl-utils==0.0.14
pydantic>=2.12,<3

View File

@ -1,47 +1,29 @@
#!/usr/bin/env bash
set -euo pipefail
export MINERU_MODEL_SOURCE=${MINERU_MODEL_SOURCE:-local}
export MINERU_MODEL_DIR=${MINERU_MODEL_DIR:-/models}
export MINERU_TOOLS_CONFIG_JSON=${MINERU_TOOLS_CONFIG_JSON:-/root/mineru.json}
export MINERU_API_MAX_CONCURRENT_REQUESTS=${MINERU_API_MAX_CONCURRENT_REQUESTS:-1}
export PYTHONPATH="/app${PYTHONPATH:+:$PYTHONPATH}"
MODEL_DIR=${MINERU_MODEL_DIR:-/models}
PIPELINE_MODEL_DIR=${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")"
export CONFIG_PATH PIPELINE_MODEL_DIR VLM_MODEL_DIR
# Preserve mounted settings, including optional title-aided configuration.
python3 - <<'PY'
import json
import os
from pathlib import Path
config_path = Path(os.environ["CONFIG_PATH"])
config = {
"bucket_info": {"[default]": [None, None, None]},
"latex-delimiter-config": {
"display": {"left": "$$", "right": "$$"},
"inline": {"left": "$", "right": "$"},
},
"llm-aided-config": {
"title_aided": {
"api_key": "",
"base_url": "",
"model": "",
"enable_thinking": False,
"enable": False,
}
},
"models-dir": {
"pipeline": os.environ["PIPELINE_MODEL_DIR"],
"vlm": os.environ["VLM_MODEL_DIR"],
},
"config_version": "1.3.1",
}
config_path.write_text(json.dumps(config, ensure_ascii=False, indent=2), encoding="utf-8")
path = Path(os.environ['MINERU_TOOLS_CONFIG_JSON'])
config = json.loads(path.read_text()) if path.exists() else {}
models = config.setdefault('models-dir', {})
if not isinstance(models, dict):
raise ValueError('models-dir must contain pipeline and vlm paths')
root = Path(os.environ['MINERU_MODEL_DIR'])
for backend in ('pipeline', 'vlm'):
override = os.getenv(f'MINERU_{backend.upper()}_MODEL_DIR')
models[backend] = override or models.get(backend) or str(root / backend)
config.setdefault('config_version', '1.3.1')
path.parent.mkdir(parents=True, exist_ok=True)
if not path.exists() or json.loads(path.read_text()) != config:
path.write_text(json.dumps(config, ensure_ascii=False, indent=2))
PY
export MINERU_TOOLS_CONFIG_JSON="$CONFIG_PATH"
export MINERU_MODEL_SOURCE=${MINERU_MODEL_SOURCE:-local}
export PYTHONPATH=${PYTHONPATH:-/app}
exec "$@"

View File

@ -1,39 +1,26 @@
# 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
# MinerU 2.7.6 / NVIDIA CUDA
FROM vllm/vllm-openai:v0.11.2
# 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 \"$@\"", "--"]
ENV DEBIAN_FRONTEND=noninteractive \
MINERU_MODEL_SOURCE=local \
MINERU_MODEL_DIR=/models \
MINERU_TOOLS_CONFIG_JSON=/root/mineru.json \
MINERU_API_MAX_CONCURRENT_REQUESTS=1
WORKDIR /app
RUN apt-get update && apt-get install -y --no-install-recommends \
fonts-noto-core fonts-noto-cjk fontconfig libgl1 libglib2.0-0 \
libreoffice-writer libreoffice-core && fc-cache -fv && \
apt-get clean && rm -rf /var/lib/apt/lists/*
COPY . /app
# Resolve against the accelerator wheels already provided by the selected base image.
RUN python3 scripts/accelerator_constraints.py /tmp/accelerator-constraints.txt && \
python3 -m pip install 'setuptools>=61,<81' wheel && \
python3 -m pip install --no-build-isolation \
-c /tmp/accelerator-constraints.txt -c docker/constraints-2.7.6.txt '.[core]' && \
python3 scripts/accelerator_constraints.py /tmp/accelerator-constraints.txt --check && \
python3 -m pip check && \
python3 -c "from mineru.version import __version__; assert __version__ == '2.7.6'" && \
chmod +x docker/entrypoint.mineru-api.sh && python3 -m pip cache purge
EXPOSE 8000
ENTRYPOINT ["/app/docker/entrypoint.mineru-api.sh"]
CMD ["mineru-api", "--host", "0.0.0.0", "--port", "8000"]

View File

@ -1,67 +1,26 @@
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
# MinerU 2.7.6 / NVIDIA CUDA
FROM vllm/vllm-openai:v0.11.2
ENV DEBIAN_FRONTEND=noninteractive \
MINERU_MODEL_SOURCE=local \
MINERU_MODEL_DIR=/models \
MINERU_TOOLS_CONFIG_JSON=/root/mineru.json \
MINERU_API_MAX_CONCURRENT_REQUESTS=1
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"]
RUN apt-get update && apt-get install -y --no-install-recommends \
fonts-noto-core fonts-noto-cjk fontconfig libgl1 libglib2.0-0 \
libreoffice-writer libreoffice-core && fc-cache -fv && \
apt-get clean && rm -rf /var/lib/apt/lists/*
COPY . /app
# Resolve against the accelerator wheels already provided by the selected base image.
RUN python3 scripts/accelerator_constraints.py /tmp/accelerator-constraints.txt && \
python3 -m pip install 'setuptools>=61,<81' wheel && \
python3 -m pip install --no-build-isolation \
-c /tmp/accelerator-constraints.txt -c docker/constraints-2.7.6.txt '.[core]' && \
python3 scripts/accelerator_constraints.py /tmp/accelerator-constraints.txt --check && \
python3 -m pip check && \
python3 -c "from mineru.version import __version__; assert __version__ == '2.7.6'" && \
chmod +x docker/entrypoint.mineru-api.sh && python3 -m pip cache purge
EXPOSE 8000
ENTRYPOINT ["/app/docker/entrypoint.mineru-api.sh"]
CMD ["mineru-api", "--host", "0.0.0.0", "--port", "8000"]

View File

@ -0,0 +1,202 @@
{
"version": "2.7.6",
"url": "https://files.pythonhosted.org/packages/44/99/a4cf2751c4055d25d16e915215c031ae362bf4bf198efed4991161c6159e/mineru-2.7.6-py3-none-any.whl",
"sha256": "b85ea4cef26397013e92dfdf3a32e68b422e233d27ba103a152843559fd7eb51",
"customizations": [
"cli/common.py: Word conversion and file progress",
"cli/fast_api.py: application API and compatibility routes",
"cli/gradio_app.py: Markmap view",
"backend/pipeline/pipeline_middle_json_mkcontent.py: six heading levels",
"backend/vlm/vlm_middle_json_mkcontent.py: six heading levels",
"utils/llm_aided.py: six-level title prompt"
],
"official_source_hashes": {
"version.py": "eb11b65df72d35957f88c01b0dfdcfb1c7b05b08cf7f07e6002db369e311d8a2",
"__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97",
"backend/utils.py": "18b254dc8ce70e6844d7fa8d9a453550eb6cbac9081c17b382c1153ba51aafe6",
"backend/__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97",
"backend/hybrid/hybrid_analyze.py": "49c930e93fa9bccbf757f9ea66405e5bc918e333480655b1a9e4b6bd7a8daad6",
"backend/hybrid/hybrid_magic_model.py": "fc3bc1ab958ffd466f9877e1665a31aaff8c2a85967bfc9ed6434bbfa440e6b5",
"backend/hybrid/hybrid_model_output_to_middle_json.py": "c84f9cd5e19ae4bccfa8b7ca7ef04e37f4894658eacb3d81ecb8a0945704ec54",
"backend/hybrid/__init__.py": "20582bd82f0d7d2023f2aec917b40ea8cbc2889e8573c4c8b94dd48760da1595",
"backend/pipeline/batch_analyze.py": "dd406cd963b0708fa67c6025c59b78dfb3aa48e95e5cf2e7a586eb0fd8790f9e",
"backend/pipeline/model_init.py": "d831d8c2aad65ed2b7bfabb911fa051d9a9f50d134d0c2dfcee10687e3a1a018",
"backend/pipeline/model_json_to_middle_json.py": "ade5e4511ff02a6243eb877bbd135733116f8b09330e51a346c869770b15aae2",
"backend/pipeline/model_list.py": "edc5cc05f66b3f42baa96b9e8350fff96a1400d79220dce49ff89cf44ed840bb",
"backend/pipeline/para_split.py": "2aaf79326be43e6eeb2b1942486893bd52b217b084ac7236786180b39b0b9744",
"backend/pipeline/pipeline_analyze.py": "f365c7ee15729ee0ff9ee93ebfb6bfce1c7fdd9fcc1d2df5b08bab4349479974",
"backend/pipeline/pipeline_magic_model.py": "c3c8c6c7c7fac9937459fdb23cfdcbf6b60a73dae8831ade642ad427326f3cef",
"backend/pipeline/pipeline_middle_json_mkcontent.py": "34908b18a13b06a336e1b45da5709f4da9728aaa33a301598e9773a48532e5a0",
"backend/pipeline/__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97",
"backend/vlm/model_output_to_middle_json.py": "02a617e38812f5cad43bfb7b4ae51ab43ef51158e8c3aa48eb2036224de0434b",
"backend/vlm/utils.py": "8a0c6073e6578def93290bd9da9fd82594cc9074bdc9713bbb5f8571a184559d",
"backend/vlm/vlm_analyze.py": "55cf2b473bdc1397a08d6fc9ecbd1bb9ea3674b2b76f728a233bc22b60320519",
"backend/vlm/vlm_magic_model.py": "4687685703498738f2b912e7e48a398053085f53710aeb8bcc26d4c4668a57cd",
"backend/vlm/vlm_middle_json_mkcontent.py": "c3e4b36e6e078ad47b318e298a7482659757b4f4aa9ad954f5c8cd87820e4328",
"backend/vlm/__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97",
"cli/client.py": "98f35f3044a1546fa2b619a5250e6719121a7768026605230461cdef30262e14",
"cli/common.py": "7cc3dcdb7e43b67ba942487db8520c1d4a71392bc2a79c9cdc0e7ab0069b0166",
"cli/fast_api.py": "6606e9f208a291cb8e9e065a966cebb1f50cadd2a739df5300066b31f190597b",
"cli/gradio_app.py": "d882163a6d9b1071eae4164c95f98ddf2009c3535ff125002d343de68f9b632d",
"cli/models_download.py": "2cd7e822952524cee6eea6f64a20b1b63303c388c82c142d67034fd89a18f355",
"cli/vlm_server.py": "dbb1daa8edf0a4c5d203f9c0dc20ba24e0531cadeaeba48fd34703ea6f47b901",
"cli/__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97",
"data/__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97",
"data/data_reader_writer/base.py": "9ea98071d1ce5cc3893ba4404f720b96280315ac3c3a9d124f20b0e723b301b2",
"data/data_reader_writer/dummy.py": "312c5069938af22fa59643d70e7d3c92fbee22d7b9a0907fe25463af59a75ed0",
"data/data_reader_writer/filebase.py": "8255d548927eb80fffa83d09feb7611d4d956043866a20b6824f52270ca73e3e",
"data/data_reader_writer/multi_bucket_s3.py": "6f05488a65556a3d0c35e549a4d000ad6e35746caaaaf0b3abd259e8e85db298",
"data/data_reader_writer/s3.py": "3707fc89c1d542a9bc448da7e00cf356ec0aef9775ab925956e0e97bbe028478",
"data/data_reader_writer/__init__.py": "f6a9c636bb2e181330c1fb10cbaa0285d933fbe6bf2cf758584d15659af4cabe",
"data/io/base.py": "4aa350a9edf466fa15be0ec656ffa12cc0a337ac81803c9041678b806b137e14",
"data/io/http.py": "776f826411a330f3af7409240bdcf150429862264f80befa66a706e21203793d",
"data/io/s3.py": "ddf0e2b4deeb1069f50ca0e48cab5fdb20baf260eb27eb55cb28bc55890d61e0",
"data/io/__init__.py": "9a343f2c049a5fee3fa60fb5bb368c272b0812582552cde8f9a92aac03fc3024",
"data/utils/exceptions.py": "a3116951056e9bc2d1140820d5c66f328378c604947bde6b80d0d4da6cd39707",
"data/utils/path_utils.py": "ca47a8f965b5eb723618a016a34bc8a4fd4caedc08f7d3cfaad082d39ba1bd53",
"data/utils/schemas.py": "30afe99d690aebd3119d56b29278b6b42472eacc7b71d09e3ebcbf5bb5148217",
"data/utils/__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97",
"model/__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97",
"model/layout/doclayoutyolo.py": "0edb4835dba5cd38987150e5ffbd280ed51d7d598673daa429699b3ce180795d",
"model/layout/__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97",
"model/mfd/yolo_v8.py": "388e40c5582ddc5bd7a783589340435d7be90e5a3d62333a6f20f20bf4d9f09b",
"model/mfd/__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97",
"model/mfr/utils.py": "a408b51e7913b8ed11eb6e751dd97ea39d26d301f409ef3d3c07fbe160ac5cf5",
"model/mfr/__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97",
"model/mfr/pp_formulanet_plus_m/predict_formula.py": "b586f1746fe834b6f5f02b1092eb1903eaf77f11d0775b839dfcc8150e88214e",
"model/mfr/pp_formulanet_plus_m/processors.py": "3122b26a7c620c38e00d01c1a2ff868cab4f9ccc7db529b1042f462243377edf",
"model/mfr/pp_formulanet_plus_m/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"model/mfr/unimernet/Unimernet.py": "64ad0cf5f3e667388ae03df71f760d0fb4671e2424455092fa5bcd7d8f8dedda",
"model/mfr/unimernet/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"model/mfr/unimernet/unimernet_hf/modeling_unimernet.py": "fe5377cc32b17aab16fa1f6d5f1efd0d8893e6e4f83fd8b11b8f56ad260a1711",
"model/mfr/unimernet/unimernet_hf/__init__.py": "907708486f064b84d6256df84822427a2d63c68e87c6f3b4d1a0b476ac8d1602",
"model/mfr/unimernet/unimernet_hf/unimer_mbart/configuration_unimer_mbart.py": "87045aaf7a6a37f715b374d14cd4ae841e3069c06771f27eaaf6936a345bd317",
"model/mfr/unimernet/unimernet_hf/unimer_mbart/modeling_unimer_mbart.py": "5e49cbe940f6b217dc12b003f242e4e754f2dcbb5bbfbb838bf6379311b58def",
"model/mfr/unimernet/unimernet_hf/unimer_mbart/tokenization_unimer_mbart.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"model/mfr/unimernet/unimernet_hf/unimer_mbart/__init__.py": "f53dab069c865f96054188fcf7e996ba346890ebb3e3180dade06e7a0720d7f7",
"model/mfr/unimernet/unimernet_hf/unimer_swin/configuration_unimer_swin.py": "397dde4542819ca097b7124e1b7b1d368075215f99ede7e0594fa069c9583860",
"model/mfr/unimernet/unimernet_hf/unimer_swin/image_processing_unimer_swin.py": "6bd902bf0cc925246b29036d5b6a0ea53c2b6a9cdea7c06318558b84b1754f49",
"model/mfr/unimernet/unimernet_hf/unimer_swin/modeling_unimer_swin.py": "43f7dd9851d406d12801f5a9f5aa307704c213630814c38f6d88e84b25cada25",
"model/mfr/unimernet/unimernet_hf/unimer_swin/__init__.py": "f3fd432b00c20d4064787622249e8c667a1d641b1ab476ea872821d75b3d7840",
"model/ocr/pytorch_paddle.py": "707313979b0aca7e01636db4efeec643879997d05051cb39b9cc70fcd15ecc82",
"model/ocr/__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97",
"model/ori_cls/paddle_ori_cls.py": "5484b6d887ab1d24fb83ad000bdaf63c44081ba35059e41a1f53ab5c8c4d4ec8",
"model/ori_cls/__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97",
"model/reading_order/layout_reader.py": "21550570d303177fa48a8f812318e9a479d64b77873eabef362848c367db2053",
"model/reading_order/xycut.py": "7b3352abf6385178b3b41e7c85b5c9b234c994e06a5888ee5b9036b8b49a652a",
"model/reading_order/__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97",
"model/table/__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97",
"model/table/cls/paddle_table_cls.py": "e4fb6278a42703380c3514d91609ea42c196293110df2c855909c14488df4380",
"model/table/cls/__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97",
"model/table/rec/RapidTable.py": "d9d35d189b1576c19f466eabddd78445432cb79448c47d18ba21802e441b353c",
"model/table/rec/__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97",
"model/table/rec/slanet_plus/main.py": "bdfadcbd0f4905fdf661953d78da1eb68a9da5cac536c03558da9006c1bc8b6a",
"model/table/rec/slanet_plus/matcher.py": "bb017ec022da625690dc943ff98cb01959755d0627c7b1bf453b962d6f099419",
"model/table/rec/slanet_plus/matcher_utils.py": "f70b7fc9d79e5622dde7b6d4ea0de59d75e78b8f6a2d2adad82eb0485419922c",
"model/table/rec/slanet_plus/table_structure.py": "aadf873d8210ca9d1a586fcc9a733fb0c40257c64b6f8ac02d2b9ec82a213e03",
"model/table/rec/slanet_plus/table_structure_utils.py": "6184a4c0dd9674bc7baa458c48690f63bc97387e44355860e42b111a1b59e569",
"model/table/rec/slanet_plus/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"model/table/rec/unet_table/main.py": "275dd0effeacb58c9e76655e75ff42643ed1d2dba0b867d3837677a1be060ee3",
"model/table/rec/unet_table/table_recover.py": "ad2c9e5b2b8fd7433c74b280e5ed27e0fd835cc61b55b9a02f111c75903cff41",
"model/table/rec/unet_table/table_structure_unet.py": "8679982f36454652b463882bf3be06f4668b6a170a9cd65836e9fcebd15d987f",
"model/table/rec/unet_table/utils.py": "09802a256d3078f264e0d01e99bf16db4dcdec4df6bf4ba38966f16a70e177c2",
"model/table/rec/unet_table/utils_table_line_rec.py": "eb3d2360ee92f300267c77b9b40c847f3599210bfcc2b9ff75153d182ee82aba",
"model/table/rec/unet_table/utils_table_recover.py": "5e4b09b18f36652d24a949cd4fe8ef698cf1277577b2f312ce3d29bb021abb59",
"model/table/rec/unet_table/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"model/utils/__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97",
"model/utils/pytorchocr/base_ocr_v20.py": "e5b23b300bbae6bfaf9f6f24af0749ea98d932412f5a3b2913b110693b11111c",
"model/utils/pytorchocr/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"model/utils/pytorchocr/data/__init__.py": "618bb773e5b87e0104af1c4333df2e437a16c0f121fbac3be4b638cdc8f856d3",
"model/utils/pytorchocr/data/imaug/operators.py": "79d05a0de7b39910241a476e3c5e8859c529136597461ee6011a929f0cb29c40",
"model/utils/pytorchocr/data/imaug/__init__.py": "7381f48173d1c1e434c0c16792b08b4d1e8caed1be7b89148a7c70ab61b557d5",
"model/utils/pytorchocr/modeling/common.py": "debda34ef3d8412e08813bc8a91e25e9b055c11ee39fef3bad29a9bf7b65ab12",
"model/utils/pytorchocr/modeling/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"model/utils/pytorchocr/modeling/architectures/base_model.py": "4615760e6fa8b34f24085ca54f9ef346eef61eafd125d172df141ed4c3da0ae5",
"model/utils/pytorchocr/modeling/architectures/__init__.py": "0c203d152e26139a021c3941854ae460b77115e4086e13e3e0ff2827f811643f",
"model/utils/pytorchocr/modeling/backbones/det_mobilenet_v3.py": "af48169c0d579add19c381502f1ee47fe596c1977fdba3df37384cd3976b72e1",
"model/utils/pytorchocr/modeling/backbones/rec_donut_swin.py": "9530a37bb98fb84f9f1ded3901325ee42efbc319340914e8940b43fe07d04e68",
"model/utils/pytorchocr/modeling/backbones/rec_hgnet.py": "52c21bcea37f928c86a128754c0f6bdbb4a08291db78a4b71e6992f80d80c34e",
"model/utils/pytorchocr/modeling/backbones/rec_lcnetv3.py": "f954734cfb6becb4e6c09e1d0adb08fbe803f35600761de59556aef4ae159dcf",
"model/utils/pytorchocr/modeling/backbones/rec_mobilenet_v3.py": "989984eb11a98c7647d956b0d7a2e5225a911459bd47dc91b122446b7627de7c",
"model/utils/pytorchocr/modeling/backbones/rec_mv1_enhance.py": "2b8a7d28560d9a5b55df2dd0b311c8012371a2a946b71800a02c5e1687f20a6c",
"model/utils/pytorchocr/modeling/backbones/rec_pphgnetv2.py": "61fa75c51e5286ce5bd6ec468e1152ad880fabd065dcda20612a3e28e26ed3c9",
"model/utils/pytorchocr/modeling/backbones/rec_svtrnet.py": "008694677216064473da95a66a70634b441d25c6278a6312578316a1f36941c8",
"model/utils/pytorchocr/modeling/backbones/__init__.py": "74e9aa86f2eb04cfb68a6730b7bde44ba00f98c6e1a15613a33966a8e9117ea0",
"model/utils/pytorchocr/modeling/heads/cls_head.py": "a6e232e46954b4029eafa792e0758abb4ecfcd177e1e50c0a88cf95aa8c11da0",
"model/utils/pytorchocr/modeling/heads/det_db_head.py": "fa4f1ba6e190c3fc48543b2e9ab7e298ec60d0efa83f630e0320c94e353bd11a",
"model/utils/pytorchocr/modeling/heads/rec_ctc_head.py": "cb0ca4e512605084dd5efad4664db205258ab1a648aa74e87dd15bb9052dc235",
"model/utils/pytorchocr/modeling/heads/rec_multi_head.py": "2b8d12300f2d0155aefb77f081f8778c659e54501d5673321e3799788f4e3bb4",
"model/utils/pytorchocr/modeling/heads/rec_ppformulanet_head.py": "e5c2c951ab032ab6020b8ef34f1d83e4eb25e829cf87a24099dc1beac69e0d68",
"model/utils/pytorchocr/modeling/heads/rec_unimernet_head.py": "c2c1f16445f654eea43442d1e37794325796257ac352070fe672d67cd22b33e1",
"model/utils/pytorchocr/modeling/heads/__init__.py": "7650d6008083ff73eb622862a421c33f91822551fff9fc128fb59fa237c808ca",
"model/utils/pytorchocr/modeling/necks/db_fpn.py": "4cb176a52caf442d283f32f4795c8d960dd6e99bdfaf827c7c3d67ce2541eee2",
"model/utils/pytorchocr/modeling/necks/intracl.py": "c3641dc1d23d069896f76552e26a8bdf5b1511121b639dd37db0f943aa2489a6",
"model/utils/pytorchocr/modeling/necks/rnn.py": "4c052ae267b883fc97c5ab79c053a0193aa7a82d9450ad456762defb61162aa0",
"model/utils/pytorchocr/modeling/necks/__init__.py": "eb7e0bd72f905afe4ff28a4d89298abd0131dd4b24736d111bc0d888275b97c5",
"model/utils/pytorchocr/postprocess/cls_postprocess.py": "d555565d3fdbd6f8466fb3c6beac9f510dc8a7b2eea47eb6bcfbdaf7c1ad8d30",
"model/utils/pytorchocr/postprocess/db_postprocess.py": "01d64f17b7eb850dbb555769d0616631c5ed8afc036597d7621cc93a53f8cd4b",
"model/utils/pytorchocr/postprocess/rec_postprocess.py": "a86077a271441611e3aa4b0847520a3b1d8463df1ec1fb268c00e3ad15e0df46",
"model/utils/pytorchocr/postprocess/__init__.py": "882d4e97a093c51599054c90ff9215311ea422eaeff5624f3965a8ecd02e6410",
"model/utils/pytorchocr/utils/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"model/utils/pytorchocr/utils/resources/arch_config.yaml": "06f697b075d7572c5838562392dec7b72656884c8baef1ec1ec2587d06528436",
"model/utils/pytorchocr/utils/resources/models_config.yml": "408baa7aef75a5fbaac9ae096f199941290e6247be876073db53b6a4e2d03d87",
"model/utils/pytorchocr/utils/resources/pp_formulanet_arch_config.yaml": "6bbcae78e4d4adfa59a3c0ac2babd0a246cb341d89f8fefb8a168287f2e8cef4",
"model/utils/pytorchocr/utils/resources/dict/ka_dict.txt": "fad3f765943275eec2134a6fbc9b52785426641104d4e7db3a159dc73f341dde",
"model/utils/pytorchocr/utils/resources/dict/ppocrv4_doc_dict.txt": "a5bc3887c43c901e5a3f97b13ffadf1c5754ede7cc8c9f5abe22e875a7c48372",
"model/utils/pytorchocr/utils/resources/dict/ppocrv5_arabic_dict.txt": "7f92f7dbb9b75a4787a83bfb4f6d14a8ab515525130c9d40a9036f61cf6999e9",
"model/utils/pytorchocr/utils/resources/dict/ppocrv5_cyrillic_dict.txt": "db40aa52ceb112055be80c694afdf655d5d2c4f7873704524cc16a447ca913ba",
"model/utils/pytorchocr/utils/resources/dict/ppocrv5_devanagari_dict.txt": "09c7440bfc5477e5c41052304b6b185aff8c4a5e8b2b4c23c1c706f6fe1ee9fc",
"model/utils/pytorchocr/utils/resources/dict/ppocrv5_dict.txt": "d1979e9f794c464c0d2e0b70a7fe14dd978e9dc644c0e71f14158cdf8342af1b",
"model/utils/pytorchocr/utils/resources/dict/ppocrv5_el_dict.txt": "31defc62c0c3ad3674a82da6192226a2ba98ef4ff014a7045cb88d59f9c3de31",
"model/utils/pytorchocr/utils/resources/dict/ppocrv5_en_dict.txt": "e025a66d31f327ba0c232e03f407ae8d105e1e709e7ccb3f408aa778c24e70d6",
"model/utils/pytorchocr/utils/resources/dict/ppocrv5_eslav_dict.txt": "3e95f1581557162870cacdba5af91a4c6be2890710d395b0c3c7578e7ee5e6eb",
"model/utils/pytorchocr/utils/resources/dict/ppocrv5_korean_dict.txt": "a88071c68c01707489baa79ebe0405b7beb5cca229f4fc94cc3ef992328802d7",
"model/utils/pytorchocr/utils/resources/dict/ppocrv5_latin_dict.txt": "ccbcc45730b3fbbd9050c5bc74db6a99067141ef1035e3d14889a84a6b9b1aff",
"model/utils/pytorchocr/utils/resources/dict/ppocrv5_ta_dict.txt": "85b541352ae18dc6ba6d47152d8bf8adff6b0266e605d2eef2990c1bf466117b",
"model/utils/pytorchocr/utils/resources/dict/ppocrv5_te_dict.txt": "42f83f5d3fdb50778e4fa5b66c58d99a59ab7792151c5e74f34b8ffd7b61c9d6",
"model/utils/pytorchocr/utils/resources/dict/ppocrv5_th_dict.txt": "57f5406f94bb6688fb7077f7be65f08bbd71cecf48c01ea26c522cb5c4836b7a",
"model/utils/tools/__init__.py": "c44a91eb967c60ecce2e8acb8cad0b087a2cdb35feb42bb14abc67763534d211",
"model/utils/tools/infer/predict_cls.py": "f1198a975bde8e7665eb971a1d9355dad6ba84cb20e41fcb13e16ea823bc4fc0",
"model/utils/tools/infer/predict_det.py": "bd8411127eef10b5f106cafbd8208256f9b583057cd8e35a0b019fc542239def",
"model/utils/tools/infer/predict_rec.py": "f811fddc90e2b2ed244fa0b21c0e2995438a71bd8bf94bc393b4de8a7e6ec2de",
"model/utils/tools/infer/predict_system.py": "8647a0927ea1ab6bf6cea1d500feb5e7e93e7e44d2f2cc116126d97a8aa6488f",
"model/utils/tools/infer/pytorchocr_utility.py": "8b53c537efe479f254519e1593b8a0b354d4f207c4ad30e50d763afbc51ababc",
"model/utils/tools/infer/__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97",
"model/vlm/lmdeploy_server.py": "3efc4935c508281f15cd63035de575b744874a0cff50b3b7e9903325ba69cfdd",
"model/vlm/vllm_server.py": "802e1b93005b9d05e99b16a2ab59cf7fb460585fa95181128cbb2cbde26ae83a",
"model/vlm/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"resources/header.html": "ef1adfe9b1a5a11fb765e4c3c80f89bdabc4fcd791b940dede9d3b7042945d22",
"resources/fasttext-langdetect/lid.176.ftz": "e33db98de5d095f003b0645244841d33b0763864daf8d0516ae3cc1cfa759a56",
"utils/block_pre_proc.py": "b86066c5fd8c47d6e99539d023cc478c223e923de9951849af485c58a89d6ce4",
"utils/block_sort.py": "e52d5576946023bd83d9d45bdd0a795d08aa4a688fa442f0148a9ba6c847d630",
"utils/boxbase.py": "c67180d64ee156d4d0aeb7aa94998af92037cbd79d4c780b9868aa1ab4977241",
"utils/char_utils.py": "ef84f9625af99a2d6e75d008b8992ef59eac1fbbe147bb79f79fc7eea99bbb87",
"utils/check_sys_env.py": "4d18f3838c56ca8486ae0bf829a3f6db903ef7dc4180b01f67570f706aab0400",
"utils/cli_parser.py": "e2c78502ed64ba5b189f0e96336abf7318043addad12b655908f8b34417f906c",
"utils/config_reader.py": "d37012a89509225e8291755cb1ec299cf0c0a3d23b598763fe1c7ea2c50aae51",
"utils/cut_image.py": "8379b89df7093569718be3f49295d3b659aca5790c70b09f1b0998b90f99da11",
"utils/draw_bbox.py": "164829a6353346137ebb1bc285d9211d77036af244680a4c0faa82eaaa11c1f4",
"utils/engine_utils.py": "2666a8f7e3beb190f31fbbc034a103698e8d27cb6eb612ac4ebdb72c521e04b5",
"utils/enum_class.py": "59e8edef35fe11b9af40cfda8883c24b870537f2b4fbc398ce53b93f2cacba07",
"utils/format_utils.py": "dacf3dbc77122128eea2593c1ef8372bee7ead16e24f752cede14bcd42ab34ab",
"utils/guess_suffix_or_lang.py": "6940b6c002706b92c7d121f1c136ce109a95560be9754096145ea87fc7835647",
"utils/hash_utils.py": "50f4bff0d441995ba67723afd7a2e6bfa2f2db12bc3950c911ee601ba80a2059",
"utils/language.py": "ed14f79b149aee3769a02e7249ded975d1c0ed33bb52c9e60ce5a2619031bb28",
"utils/llm_aided.py": "f56532b6fc5e9d202e696478b1342154f43987ca58d3054e1f53b6b23ffa74bb",
"utils/magic_model_utils.py": "f07bfe04393d133e13531ea1ad527febbe726598203e3e9489bf35952a66d228",
"utils/models_download_utils.py": "51f8efc21c4ee819091dae494a9115359ef51a82cc3cc989a729b74c7113d93e",
"utils/model_utils.py": "61a77136e46fb96679c96d8d912a43d1961325d8f46554b65fc285fe19465820",
"utils/ocr_utils.py": "94f22bc0d5226f99abcd4929a274581ee502763576aaf11334b4b338bc9f96b5",
"utils/os_env_config.py": "5472bd952dd014986bc166bd14e154d52c26ee85e76f2e1268d3634f2a274d38",
"utils/pdf_classify.py": "e83179a47ffd52af377ec16da7b9f88be39d6101b3a0d395f4bd1505486004c4",
"utils/pdf_image_tools.py": "b534a4dfd7e024a2c4b21c0f02e2462e5fe9552ae610a8d6039e45e9d19caf88",
"utils/pdf_page_id.py": "7a6f7aea4d760915b8463e3d44688b07ff082f09173c19d645eb40a5ac77793b",
"utils/pdf_reader.py": "59e20d9b94b25815174f4c0ff65cc86de1ecf0fe9650890dea754be57e0bcc6e",
"utils/pdf_text_tool.py": "284ced8df76ab081c7ba24c400c00becbaf5392dd1ed4afeb93a8688246345e4",
"utils/run_async.py": "acf78fe010997ab47c541c910e1898cdf6626a12d5aa8644055012e787408cd8",
"utils/span_block_fix.py": "d1e5508c90ab4f4df0a11b7d86887a52ee364e9d5d69c7c64efdb1dc1f6aabde",
"utils/span_pre_proc.py": "9eee818794d63ca1731ee15f6c4b3402caf4e0ce3138be4838dcfff062479f8e",
"utils/table_merge.py": "2ce471cf44c77a60aa1fbe3a14c562a84b8bcccd3cf0ce07808904bb00c87cd5",
"utils/__init__.py": "f0246b09056e1316b406db51161dd9e34945cb62b98cdd2ca7aeca5a33a58f97"
}
}

View File

@ -0,0 +1,67 @@
# MinerU 2.7.6
本分支基于官方 PyPI 2.7.6,原始包校验值及文件清单见 `2.7.6-baseline.json`。
保留 Word 转 PDF、解析进度、Gradio/独立导图页面及导图整理接口。
解析内核使用 2.7.6 的跨页表格、PDF 超时和国产卡适配修复。
## 六级标题和 MaxKB
Pipeline、VLM 的 Markdown 与 content-list 输出保留一级至六级标题。
可选的 LLM 辅助标题分级提示也允许六级。实际识别结果取决于文档和启用的分级方式。
MaxKB 新建解析资源选择 V2,默认版本为 2.7.6。
服务地址填写 API 基址,例如 `http://parser-host:8100`。
`POST /file_parse` 和独立页面使用的 `POST /api/file_parse` 指向同一处理函数。
响应按请求返回 Markdown、middle_json、model_output、content_list、Base64 图片。
MaxKB 保留原有 2.7.1、2.7.4、3.4.5、4.x 资源的版本定义。
## 构建
在仓库根目录执行。Docker 命令需有相应权限。
```bash
API_FLAVOR=cuda bash scripts/build-images.sh
API_FLAVOR=corex bash scripts/build-images.sh
```
NVIDIA API 镜像为 `unis-mindmap-api:2.7.6-cuda`,使用 vLLM 0.11.2 配套基础镜像。
天数 API 镜像为 `unis-mindmap-api:2.7.6-corex`,使用厂商 CoreX 4.4.0、Torch 2.7.1、vLLM 0.11.2、Python 3.10 基础镜像。
独立页面镜像为 `unis-mindmap-web:2.7.6`。
API 镜像从本仓库安装 `.[core]`,并约束 `mineru==2.7.6`。
安装前记录基础镜像的 Torch、torchvision、vLLM 等版本,作为 pip 约束,安装后再次核对并运行 `pip check`。
新增依赖包括 Word 备用转换所需的 python-docx 和 Pydantic 2。
CoreX 与 CUDA 的依赖解析冲突会使构建失败,日志会保留 pip 的冲突原因。
## 准备模型和启动
先准备与 MinerU 2.7.6 对应的 Pipeline、VLM 权重。
`MINERU_MODEL_DIR` 下使用 `pipeline`、`vlm` 两个目录,分别放对应模型仓库的根内容。
构建和启动过程使用挂载的本地模型。
```bash
API_FLAVOR=cuda GPU_IDS=0 MINERU_MODEL_DIR=/data/unis/mineru-models bash scripts/start-mindmap.sh
API_FLAVOR=corex GPU_IDS=0 MINERU_MODEL_DIR=/data/unis/mineru-models bash scripts/start-mindmap.sh
```
脚本默认使用容器名 `mineru-276-api`、`mindmap-276-web`。
API 映射至宿主机 `127.0.0.1:8100`,页面端口为 `9898`。
可通过 `API_NAME`、`WEB_NAME`、`API_PORT`、`WEB_PORT` 调整。
已有同名容器时脚本退出,由运维显式管理旧容器。
天数启动传递 `MINERU_VLLM_DEVICE=corex`,并挂载厂商运行所需的设备和内核目录。
可挂载 `MINERU_TOOLS_CONFIG_JSON` 指定的 JSON 配置。
入口保留已有配置,并补齐缺失的模型目录。
`MINERU_PIPELINE_MODEL_DIR`、`MINERU_VLM_MODEL_DIR` 可分别覆盖模型路径。
`MINERU_GPU_MEMORY_UTILIZATION` 默认 `0.5`,用于 vLLM 显存预算。
API 默认单请求并发,`MINERU_API_MAX_CONCURRENT_REQUESTS` 可配置上限。
## 检查
```bash
python -m unittest discover -s tests/unittest -p test_parser_276.py -v
python -m build --wheel
```
协议测试模拟推理输出,检查 HTTP 协议和标题保留。
目标卡验收需另行检查实际驱动、模型加载、PDF/OCR、显存和并发行为。

View File

@ -1,6 +1,6 @@
Metadata-Version: 2.4
Name: mineru
Version: 2.7.1
Version: 2.7.6
Summary: A practical tool for converting PDF to Markdown
License: AGPL-3.0
Project-URL: homepage, https://mineru.net/
@ -14,6 +14,7 @@ Classifier: Programming Language :: Python :: 3.12
Classifier: Programming Language :: Python :: 3.13
Requires-Python: <3.14,>=3.10
Description-Content-Type: text/markdown
License-File: LICENSE.md
Requires-Dist: boto3>=1.28.43
Requires-Dist: click>=8.1.7
Requires-Dist: loguru>=0.7.2
@ -26,6 +27,8 @@ Requires-Dist: pillow>=11.0.0
Requires-Dist: pypdfium2>=4.30.0
Requires-Dist: pypdf>=5.6.0
Requires-Dist: reportlab
Requires-Dist: python-docx<2,>=1.2
Requires-Dist: pydantic<3,>=2.12
Requires-Dist: pdftext>=0.6.3
Requires-Dist: modelscope>=1.26.0
Requires-Dist: huggingface-hub>=0.32.4
@ -59,7 +62,7 @@ Requires-Dist: matplotlib<4,>=3.10; extra == "pipeline"
Requires-Dist: ultralytics<9,>=8.3.48; extra == "pipeline"
Requires-Dist: doclayout_yolo==0.0.4; extra == "pipeline"
Requires-Dist: dill<1,>=0.3.8; extra == "pipeline"
Requires-Dist: PyYAML<7,>=6.0.2; extra == "pipeline"
Requires-Dist: PyYAML<7,>=6.0.1; extra == "pipeline"
Requires-Dist: ftfy<7,>=6.3.1; extra == "pipeline"
Requires-Dist: shapely<3,>=2.0.7; extra == "pipeline"
Requires-Dist: pyclipper<2,>=1.3.0; extra == "pipeline"
@ -75,9 +78,6 @@ Requires-Dist: uvicorn; extra == "api"
Provides-Extra: gradio
Requires-Dist: gradio==5.49.1; extra == "gradio"
Requires-Dist: gradio-pdf==0.0.22; extra == "gradio"
Provides-Extra: vuex
Requires-Dist: nodejs>=16.0.0; extra == "vuex"
Requires-Dist: npm>=8.0.0; extra == "vuex"
Provides-Extra: core
Requires-Dist: mineru[vlm]; extra == "core"
Requires-Dist: mineru[pipeline]; extra == "core"
@ -87,4 +87,8 @@ Provides-Extra: all
Requires-Dist: mineru[core]; extra == "all"
Requires-Dist: mineru[mlx]; sys_platform == "darwin" and extra == "all"
Requires-Dist: mineru[vllm]; sys_platform == "linux" and extra == "all"
Requires-Dist: mineru[lmdeploy]; sys_platform == "windows" and extra == "all"
Requires-Dist: mineru[lmdeploy]; sys_platform == "win32" and extra == "all"
Dynamic: license-file
markitdown

View File

@ -1,3 +1,5 @@
LICENSE.md
README.md
pyproject.toml
mineru/__init__.py
mineru/version.py
@ -36,6 +38,22 @@ mineru/cli/fast_api.py
mineru/cli/gradio_app.py
mineru/cli/models_download.py
mineru/cli/vlm_server.py
mineru/cli/static/markmap/d3.min.js
mineru/cli/static/markmap/markmap-lib.js
mineru/cli/static/markmap/markmap-toolbar.js
mineru/cli/static/markmap/markmap-view.js
mineru/cli/static/web/index.html
mineru/cli/static/web/assets/index-B_viQAxH.css
mineru/cli/static/web/assets/index-CxJ0NebI.css
mineru/cli/static/web/assets/index-D7YNJWSR.js
mineru/cli/static/web/assets/index-zcV97yfy.js
mineru/cli/static/web/assets/markdown-_Dy2zPzC.js
mineru/cli/static/web/assets/vendor-DDOvA1Mb.js
mineru/cli/static/web/assets/vendor-DSv9k51x.js
mineru/cli/static/web/markmap/d3.min.js
mineru/cli/static/web/markmap/markmap-lib.js
mineru/cli/static/web/markmap/markmap-toolbar.js
mineru/cli/static/web/markmap/markmap-view.js
mineru/data/__init__.py
mineru/data/data_reader_writer/__init__.py
mineru/data/data_reader_writer/base.py

View File

@ -1,9 +1,8 @@
[console_scripts]
mineru = mineru.cli:client.main
mineru = mineru.cli.client:main
mineru-api = mineru.cli.fast_api:main
mineru-gradio = mineru.cli.gradio_app:main
mineru-lmdeploy-server = mineru.cli.vlm_server:lmdeploy_server
mineru-models-download = mineru.cli.models_download:download_models
mineru-openai-server = mineru.cli.vlm_server:openai_server
mineru-vllm-server = mineru.cli.vlm_server:vllm_server
mineru-vue-dev = web_ui:start.sh

View File

@ -10,6 +10,8 @@ pillow>=11.0.0
pypdfium2>=4.30.0
pypdf>=5.6.0
reportlab
python-docx<2,>=1.2
pydantic<3,>=2.12
pdftext>=0.6.3
modelscope>=1.26.0
huggingface-hub>=0.32.4
@ -32,7 +34,7 @@ mineru[mlx]
[all:sys_platform == "linux"]
mineru[vllm]
[all:sys_platform == "windows"]
[all:sys_platform == "win32"]
mineru[lmdeploy]
[api]
@ -61,7 +63,7 @@ matplotlib<4,>=3.10
ultralytics<9,>=8.3.48
doclayout_yolo==0.0.4
dill<1,>=0.3.8
PyYAML<7,>=6.0.2
PyYAML<7,>=6.0.1
ftfy<7,>=6.3.1
shapely<3,>=2.0.7
pyclipper<2,>=1.3.0
@ -85,7 +87,3 @@ vllm<0.12,>=0.10.1.1
torch<3,>=2.6.0
transformers!=4.57.2,<5.0.0,>=4.51.1
accelerate>=1.5.1
[vuex]
nodejs>=16.0.0
npm>=8.0.0

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

@ -296,8 +296,8 @@ def union_make(pdf_info_dict: list,
def get_title_level(block):
title_level = block.get('level', 1)
if title_level > 4:
title_level = 4
if title_level > 6:
title_level = 6
elif title_level < 1:
title_level = 0
return title_level

View File

@ -18,6 +18,15 @@ 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"
elif hasattr(torch, 'sdaa') and torch.sdaa.is_available():
compute_capability = "8.0"
else:
logger.info("CUDA not available, disabling custom_logits_processors")
return False
@ -71,16 +80,6 @@ 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)
@ -107,3 +106,127 @@ def set_default_batch_size() -> int:
logger.warning(f'Error determining VRAM: {e}, using default batch_ratio: 1')
batch_size = 1
return batch_size
def _get_device_config(device_type: str) -> dict | None:
"""获取不同设备类型的配置参数"""
# 各设备类型的配置定义
DEVICE_CONFIGS = {
# "musa": {
# "compilation_config_dict": {
# "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,
# },
"corex": {
"compilation_config_dict": {
"cudagraph_mode": "FULL_DECODE_ONLY",
"level": 0
},
},
"kxpu": {
"compilation_config_dict": {
"splitting_ops": [
"vllm.unified_attention", "vllm.unified_attention_with_output",
"vllm.unified_attention_with_output_kunlun", "vllm.mamba_mixer2",
"vllm.mamba_mixer", "vllm.short_conv", "vllm.linear_attention",
"vllm.plamo2_mamba_mixer", "vllm.gdn_attention", "vllm.sparse_attn_indexer"
]
},
"block_size": 128,
"dtype": "float16",
"distributed_executor_backend": "mp",
"enable_chunked_prefill": False,
"enable_prefix_caching": False,
},
}
return DEVICE_CONFIGS.get(device_type.lower())
def _check_server_arg_exists(args: list, arg_name: str) -> bool:
"""检查命令行参数列表中是否已存在指定参数"""
return any(arg == f"--{arg_name}" or arg.startswith(f"--{arg_name}=") for arg in args)
def _add_server_arg_if_missing(args: list, arg_name: str, value: str) -> None:
"""如果参数不存在,则添加到命令行参数列表"""
if not _check_server_arg_exists(args, arg_name):
args.extend([f"--{arg_name}", value])
def _add_server_flag_if_missing(args: list, flag_name: str) -> None:
"""如果 flag 不存在,则添加到命令行参数列表"""
if not _check_server_arg_exists(args, flag_name):
args.append(f"--{flag_name}")
def _add_engine_kwarg_if_missing(kwargs: dict, key: str, value) -> None:
"""如果参数不存在,则添加到 kwargs 字典"""
if key not in kwargs:
kwargs[key] = value
def mod_kwargs_by_device_type(kwargs_or_args: dict | list, vllm_mode: str) -> dict | list:
"""根据设备类型修改 vllm 配置参数
Args:
kwargs_or_args: 配置参数,server 模式为 list,engine 模式为 dict
vllm_mode: vllm 运行模式 ("server", "sync_engine", "async_engine")
Returns:
修改后的配置参数
"""
device_type = os.getenv("MINERU_VLLM_DEVICE", "")
config = _get_device_config(device_type)
if config is None:
return kwargs_or_args
if vllm_mode == "server":
_apply_server_config(kwargs_or_args, config)
else:
_apply_engine_config(kwargs_or_args, config, vllm_mode)
return kwargs_or_args
def _apply_server_config(args: list, config: dict) -> None:
"""应用 server 模式的配置"""
import json
for key, value in config.items():
if key == "compilation_config_dict":
_add_server_arg_if_missing(
args, "compilation-config",
json.dumps(value, separators=(',', ':'))
)
else:
# 转换 key 格式: block_size -> block-size
arg_name = key.replace("_", "-")
if arg_name in {"enable-chunked-prefill", "enable-prefix-caching"} and value is False:
_add_server_flag_if_missing(args, f"no-{arg_name}")
continue
_add_server_arg_if_missing(args, arg_name, str(value))
def _apply_engine_config(kwargs: dict, config: dict, vllm_mode: str) -> None:
"""应用 engine 模式的配置"""
try:
from vllm.config import CompilationConfig
except ImportError:
raise ImportError("Please install vllm to use the vllm-async-engine backend.")
for key, value in config.items():
if key == "compilation_config_dict":
if vllm_mode == "sync_engine":
compilation_config = value
elif vllm_mode == "async_engine":
compilation_config = CompilationConfig(**value)
else:
continue
_add_engine_kwarg_if_missing(kwargs, "compilation_config", compilation_config)
else:
_add_engine_kwarg_if_missing(kwargs, key, value)

View File

@ -1,11 +1,12 @@
# Copyright (c) Opendatalab. All rights reserved.
import os
import time
import json
from loguru import logger
from .utils import enable_custom_logits_processors, set_default_gpu_memory_utilization, set_default_batch_size, \
set_lmdeploy_backend
set_lmdeploy_backend, mod_kwargs_by_device_type
from .model_output_to_middle_json import result_to_middle_json
from ...data.data_reader_writer import DataWriter
from mineru.utils.pdf_image_tools import load_images_from_pdf
@ -99,6 +100,17 @@ class ModelSingleton:
import vllm
except ImportError:
raise ImportError("Please install vllm to use the vllm-engine backend.")
kwargs = mod_kwargs_by_device_type(kwargs, vllm_mode="sync_engine")
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 +122,27 @@ 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.")
kwargs = mod_kwargs_by_device_type(kwargs, vllm_mode="async_engine")
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 +151,6 @@ 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

@ -651,8 +651,8 @@ def union_make(pdf_info_dict: list,
def get_title_level(block):
title_level = block.get('level', 1)
if title_level > 4:
title_level = 4
if title_level > 6:
title_level = 6
elif title_level < 1:
title_level = 0
return title_level

View File

@ -14,10 +14,11 @@ import urllib.request
import urllib.error
from pathlib import Path
import glob
import shutil
os.environ.setdefault("PYTORCH_CUDA_ALLOC_CONF", "expandable_segments:True")
from fastapi import Depends, FastAPI, HTTPException, UploadFile, File, Form, APIRouter, Header
from fastapi import BackgroundTasks, Depends, FastAPI, HTTPException, UploadFile, File, Form, APIRouter, Header
from pydantic import BaseModel
from fastapi.middleware.gzip import GZipMiddleware
from fastapi.middleware.cors import CORSMiddleware
@ -338,7 +339,9 @@ def sanitize_filename(filename: str) -> str:
def cleanup_file(file_path: str) -> None:
try:
if os.path.exists(file_path):
if os.path.isdir(file_path):
shutil.rmtree(file_path)
elif os.path.exists(file_path):
os.remove(file_path)
except Exception as e:
logger.warning(f"fail clean file {file_path}: {e}")
@ -359,6 +362,7 @@ def get_infer_result(file_suffix_identifier: str, pdf_name: str, parse_dir: str)
# --- 创建 API 路由组 ---
api_router = APIRouter(prefix="/api")
parser_router = APIRouter()
class MindmapLlmConfig(BaseModel):
@ -418,7 +422,7 @@ DEFAULT_MINDMAP_ORGANIZE_PROMPT = """你是文档结构整理助手。请基于
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。
6. 最大层级不超过 6 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。"""
@ -430,7 +434,7 @@ MINDMAP_MERGE_PROMPT = """你是 Markdown 思维导图结构校对助手。下
2. 只检查和调整标题层级结构、顺序、重复标题和父子关系。
3. 不要重新总结正文,不要扩写内容,不要新增原文没有的信息。
4. 保留各局部大纲中的源语言:英文保持英文,中文保持中文,多语言分别保留。
5. 最大层级不超过 4 层。
5. 最大层级不超过 6 层。
6. 每个父节点下最多 8 个子节点,必要时只合并相近标题。
7. 只输出 Markdown,不要输出解释、代码块围栏或额外说明。"""
@ -792,7 +796,9 @@ async def get_mindmap_progress(task_id: str):
@api_router.post(path="/file_parse", dependencies=[Depends(limit_concurrency)])
@parser_router.post(path="/file_parse", dependencies=[Depends(limit_concurrency)])
async def parse_pdf(
background_tasks: BackgroundTasks,
files: List[UploadFile] = File(..., description="Upload pdf, image, or Word files for parsing"),
output_dir: str = Form("./output", description="Output local directory"),
lang_list: List[str] = Form(["ch"]),
@ -814,7 +820,7 @@ async def parse_pdf(
):
# 从 app 实例状态中获取配置 (FastAPI 实例会在下方创建)
from fastapi import Request
config = getattr(app.state, "config", {})
config = dict(getattr(app.state, "config", {}))
# 初始化进度跟踪
task_id = x_task_id or form_task_id or str(uuid.uuid4())
@ -831,13 +837,14 @@ async def parse_pdf(
try:
unique_dir = os.path.join(output_dir, str(uuid.uuid4()))
os.makedirs(unique_dir, exist_ok=True)
background_tasks.add_task(cleanup_file, unique_dir)
pdf_file_names = []
pdf_bytes_list = []
for file in files:
content = await file.read()
file_path = Path(file.filename)
file_path = Path((file.filename or 'document.pdf').replace('\\', '/').rsplit('/', 1)[-1])
temp_path = Path(unique_dir) / file_path.name
with open(temp_path, "wb") as f:
f.write(content)
@ -847,7 +854,10 @@ async def parse_pdf(
try:
pdf_bytes = read_fn(temp_path)
pdf_bytes_list.append(pdf_bytes)
pdf_file_names.append(file_path.stem)
name = sanitize_filename(file_path.stem)
if name in pdf_file_names:
name = f"{name}_{len(pdf_file_names) + 1}"
pdf_file_names.append(name)
os.remove(temp_path)
except Exception as e:
return JSONResponse(status_code=400, content={"error": f"Failed to load file: {str(e)}"})
@ -922,13 +932,22 @@ async def parse_pdf(
path = os.path.join(p_dir, f"{pdf_name}.md")
if os.path.exists(path): zf.write(path,
arcname=os.path.join(safe_pdf_name, f"{safe_pdf_name}.md"))
for enabled, suffix in (
(return_middle_json, '_middle.json'),
(return_model_output, '_model.json'),
(return_content_list, '_content_list.json'),
):
path = os.path.join(p_dir, pdf_name + suffix)
if enabled and os.path.exists(path):
zf.write(path, arcname=os.path.join(safe_pdf_name, safe_pdf_name + suffix))
if return_images:
images_dir = os.path.join(p_dir, "images")
for img in glob.glob(os.path.join(glob.escape(images_dir), "*.jpg")):
zf.write(img, arcname=os.path.join(safe_pdf_name, "images", os.path.basename(img)))
background_tasks.add_task(cleanup_file, zip_path)
return FileResponse(path=zip_path, media_type="application/zip", filename="results.zip",
background=BackgroundTask(cleanup_file, zip_path))
background=background_tasks)
else:
result_dict = {}
for pdf_name in pdf_file_names:
@ -943,6 +962,13 @@ async def parse_pdf(
if os.path.exists(p_dir):
if return_md: data["md_content"] = get_infer_result(".md", pdf_name, p_dir)
for enabled, key, suffix in (
(return_middle_json, 'middle_json', '_middle.json'),
(return_model_output, 'model_output', '_model.json'),
(return_content_list, 'content_list', '_content_list.json'),
):
if enabled:
data[key] = get_infer_result(suffix, pdf_name, p_dir)
if return_images:
img_dir = os.path.join(p_dir, "images")
data["images"] = {os.path.basename(p): f"data:image/jpeg;base64,{encode_image(p)}"
@ -978,7 +1004,7 @@ def create_app():
global _request_semaphore
try:
mcr = int(os.getenv("MINERU_API_MAX_CONCURRENT_REQUESTS", "0"))
mcr = int(os.getenv("MINERU_API_MAX_CONCURRENT_REQUESTS", "1"))
if mcr > 0:
_request_semaphore = asyncio.Semaphore(mcr)
logger.info(f"Concurrency limited to {mcr}")
@ -991,11 +1017,12 @@ def create_app():
# 1. 首先挂载 API 路由组 (处理所有 /api/* 请求)
app.include_router(api_router)
app.include_router(parser_router)
# 2. 根路径重定向或特定处理 (可选)
@app.get("/health")
async def health():
return {"status": "ok"}
return {"status": "ok", "version": __version__}
# 3. 最后挂载静态文件服务 (处理剩下的所有请求,如 /, /index.html, /assets/*)
static_dir = Path(__file__).parent / "static" / "web"
@ -1018,12 +1045,19 @@ app = create_app()
@click.option('--reload', is_flag=True)
def main(ctx, host, port, reload, **kwargs):
kwargs.update(arg_parse(ctx))
utilization = os.getenv('MINERU_GPU_MEMORY_UTILIZATION')
if utilization is not None:
utilization = float(utilization)
if not 0 < utilization <= 1:
raise click.BadParameter('MINERU_GPU_MEMORY_UTILIZATION must be greater than 0 and at most 1')
kwargs.setdefault('gpu_memory_utilization', utilization)
app.state.config = kwargs
mcr = str(kwargs.get("mineru_api_max_concurrent_requests", "0") or "0")
os.environ["MINERU_API_MAX_CONCURRENT_REQUESTS"] = mcr
uvicorn.run("mineru.cli.fast_api:app", host=host, port=port, reload=reload)
mcr = int(kwargs.pop('mineru_api_max_concurrent_requests', os.getenv('MINERU_API_MAX_CONCURRENT_REQUESTS', '1')))
os.environ['MINERU_API_MAX_CONCURRENT_REQUESTS'] = str(mcr)
global _request_semaphore
_request_semaphore = asyncio.Semaphore(mcr) if mcr > 0 else None
uvicorn.run('mineru.cli.fast_api:app' if reload else app, host=host, port=port, reload=reload)
if __name__ == "__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

View File

@ -1,10 +1,10 @@
<!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>
<!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%;
@ -15,14 +15,14 @@
#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>
</style>
<script type="module" crossorigin src="./assets/index-D7YNJWSR.js"></script>
<link rel="modulepreload" crossorigin href="./assets/vendor-DDOvA1Mb.js">
<link rel="modulepreload" crossorigin href="./assets/markdown-_Dy2zPzC.js">
<link rel="stylesheet" crossorigin href="./assets/index-CxJ0NebI.css">
</head>
<body>
<div id="app"></div>
</body>
</html>

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

@ -89,7 +89,11 @@ class FormulaRecognizer(BaseOCRV20):
return rec_formula
def batch_predict(
self, images_mfd_res: list, images: list, batch_size: int = 64
self,
images_mfd_res: list,
images: list,
batch_size: int = 64,
interline_enable: bool = True,
) -> list:
images_formula_list = []
mf_image_list = []
@ -105,6 +109,8 @@ class FormulaRecognizer(BaseOCRV20):
for idx, (xyxy, conf, cla) in enumerate(
zip(mfd_res.boxes.xyxy, mfd_res.boxes.conf, mfd_res.boxes.cls)
):
if not interline_enable and cla.item() == 1:
continue # Skip interline regions if not enabled
xmin, ymin, xmax, ymax = [int(p.item()) for p in xyxy]
new_item = {
"category_id": 13 + int(cla.item()),

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

@ -1,7 +1,8 @@
import os
import sys
from mineru.backend.vlm.utils import set_default_gpu_memory_utilization, enable_custom_logits_processors
from mineru.backend.vlm.utils import set_default_gpu_memory_utilization, enable_custom_logits_processors, \
mod_kwargs_by_device_type
from mineru.utils.models_download_utils import auto_download_and_get_model_root_path
from vllm.entrypoints.cli.main import main as vllm_main
@ -50,6 +51,8 @@ def main():
if (not has_logits_processors_arg) and custom_logits_processors:
args.extend(["--logits-processors", "mineru_vl_utils:MinerULogitsProcessor"])
args = mod_kwargs_by_device_type(args, vllm_mode="server")
# 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 +186,26 @@ 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
elif device_name.startswith("sdaa"):
if hasattr(torch, 'sdaa') and torch.sdaa.is_available():
if torch.sdaa.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 +86,24 @@ 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:
try:
if torch.sdaa.is_available():
return "sdaa"
except Exception as e:
pass
return "cpu"

View File

@ -54,7 +54,7 @@ def llm_aided_title(page_info_list, title_aided_config):
- 根据标题内容的语义为每个标题元素添加适当的层次结构
- 行高较大的标题一般是更高级别的标题
- 标题从前至后的层级必须是连续的,不能跳过层级
- 标题层级最多为4级,不要添加过多的层级
- 标题层级最多为6级,不要添加过多的层级
- 优化后的标题只保留代表该标题的层级的整数,不要保留其他信息
5. 合理性检查与微调:

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 +423,18 @@ 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()
elif str(device).startswith("sdaa"):
if torch.sdaa.is_available():
torch.sdaa.empty_cache()
gc.collect()
@ -458,5 +470,17 @@ 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
elif str(device).startswith("sdaa"):
if torch.sdaa.is_available():
total_memory = round(torch.sdaa.get_device_properties(device).total_memory / (1024 ** 3)) # 转为 GB
return total_memory

View File

@ -11,6 +11,11 @@ def get_load_images_timeout() -> int:
return get_value_from_string(env_value, 300)
def get_load_images_threads() -> int:
env_value = os.getenv('MINERU_PDF_RENDER_THREADS', None)
return get_value_from_string(env_value, 4)
def get_value_from_string(env_value: str, default_value: int) -> int:
if env_value is not None:
try:

View File

@ -1,5 +1,7 @@
# Copyright (c) Opendatalab. All rights reserved.
import os
import signal
import time
from io import BytesIO
import numpy as np
@ -9,13 +11,13 @@ from PIL import Image, ImageOps
from mineru.data.data_reader_writer import FileBasedDataWriter
from mineru.utils.check_sys_env import is_windows_environment
from mineru.utils.os_env_config import get_load_images_timeout
from mineru.utils.os_env_config import get_load_images_timeout, get_load_images_threads
from mineru.utils.pdf_reader import image_to_b64str, image_to_bytes, page_to_image
from mineru.utils.enum_class import ImageType
from mineru.utils.hash_utils import str_sha256
from mineru.utils.pdf_page_id import get_end_page_id
from concurrent.futures import ProcessPoolExecutor, TimeoutError as FuturesTimeoutError
from concurrent.futures import ProcessPoolExecutor, wait, ALL_COMPLETED
def pdf_page_to_image(page: pdfium.PdfPage, dpi=200, image_type=ImageType.PIL) -> dict:
@ -57,7 +59,7 @@ def load_images_from_pdf(
end_page_id=None,
image_type=ImageType.PIL,
timeout=None,
threads=4,
threads=None,
):
"""带超时控制的 PDF 转图片函数,支持多进程加速
@ -67,8 +69,8 @@ def load_images_from_pdf(
start_page_id (int, optional): 起始页码. Defaults to 0.
end_page_id (int | None, optional): 结束页码. Defaults to None.
image_type (ImageType, optional): 图片类型. Defaults to ImageType.PIL.
timeout (int | None, optional): 超时时间(秒)。如果为 None,则从环境变量 MINERU_PDF_LOAD_IMAGES_TIMEOUT 读取,若未设置则默认为 300 秒。
threads (int): 进程数,默认 4
timeout (int | None, optional): 超时时间(秒)。如果为 None,则从环境变量 MINERU_PDF_RENDER_TIMEOUT 读取,若未设置则默认为 300 秒。
threads (int): 进程数, 如果为 None,则从环境变量 MINERU_PDF_RENDER_THREADS 读取,若未设置则默认为 4.
Raises:
TimeoutError: 当转换超时时抛出
@ -86,6 +88,9 @@ def load_images_from_pdf(
else:
if timeout is None:
timeout = get_load_images_timeout()
if threads is None:
threads = get_load_images_threads()
end_page_id = get_end_page_id(end_page_id, len(pdf_doc))
# 计算总页数
@ -108,11 +113,13 @@ def load_images_from_pdf(
page_ranges.append((range_start, range_end))
# logger.debug(f"PDF to images using {actual_threads} processes, page ranges: {page_ranges}")
logger.debug(f"PDF to images using {actual_threads} processes, page ranges: {page_ranges}")
with ProcessPoolExecutor(max_workers=actual_threads) as executor:
executor = ProcessPoolExecutor(max_workers=actual_threads)
try:
# 提交所有任务
futures = []
future_to_range = {}
for range_start, range_end in page_ranges:
future = executor.submit(
_load_images_from_pdf_worker,
@ -122,27 +129,68 @@ def load_images_from_pdf(
range_end,
image_type,
)
futures.append((range_start, future))
futures.append(future)
future_to_range[future] = range_start
try:
# 收集结果并按页码排序
all_results = []
for range_start, future in futures:
images_list = future.result(timeout=timeout)
all_results.append((range_start, images_list))
# 使用 wait() 设置单一全局超时
done, not_done = wait(futures, timeout=timeout, return_when=ALL_COMPLETED)
# 按起始页码排序并合并结果
all_results.sort(key=lambda x: x[0])
images_list = []
for _, imgs in all_results:
images_list.extend(imgs)
return images_list, pdf_doc
except FuturesTimeoutError:
# 检查是否有未完成的任务(超时情况)
if not_done:
# 超时:强制终止所有子进程
_terminate_executor_processes(executor)
pdf_doc.close()
executor.shutdown(wait=False, cancel_futures=True)
raise TimeoutError(f"PDF to images conversion timeout after {timeout}s")
# 所有任务完成,收集结果
all_results = []
for future in futures:
range_start = future_to_range[future]
# 这里不需要 timeout,因为任务已完成
images_list = future.result()
all_results.append((range_start, images_list))
# 按起始页码排序并合并结果
all_results.sort(key=lambda x: x[0])
images_list = []
for _, imgs in all_results:
images_list.extend(imgs)
return images_list, pdf_doc
except Exception as e:
# 发生任何异常时,确保清理子进程
_terminate_executor_processes(executor)
pdf_doc.close()
if isinstance(e, TimeoutError):
raise
raise
finally:
executor.shutdown(wait=False, cancel_futures=True)
def _terminate_executor_processes(executor):
"""强制终止 ProcessPoolExecutor 中的所有子进程"""
if hasattr(executor, '_processes'):
for pid, process in executor._processes.items():
if process.is_alive():
try:
# 先发送 SIGTERM 允许优雅退出
os.kill(pid, signal.SIGTERM)
except (ProcessLookupError, OSError):
pass
# 给子进程一点时间响应 SIGTERM
time.sleep(0.1)
# 对仍然存活的进程发送 SIGKILL 强制终止
for pid, process in executor._processes.items():
if process.is_alive():
try:
os.kill(pid, signal.SIGKILL)
except (ProcessLookupError, OSError):
pass
def load_images_from_pdf_core(
pdf_bytes: bytes,

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__ = "2.7.6"

View File

@ -2,7 +2,7 @@
# 天枢项目依赖
# Core MinerU
mineru[core]>=2.5.0
mineru[core]==2.7.6
# Image Augmentation (Version Pinned for Compatibility)
albumentations>=1.4.4,<2.0.0

View File

@ -29,6 +29,8 @@ dependencies = [
"pypdfium2>=4.30.0",
"pypdf>=5.6.0",
"reportlab",
"python-docx>=1.2,<2",
"pydantic>=2.12,<3",
"pdftext>=0.6.3",
"modelscope>=1.26.0",
"huggingface-hub>=0.32.4",
@ -99,7 +101,7 @@ all = [
"mineru[core]",
"mineru[mlx] ; sys_platform == 'darwin'",
"mineru[vllm] ; sys_platform == 'linux'",
"mineru[lmdeploy] ; sys_platform == 'windows'",
"mineru[lmdeploy] ; sys_platform == 'win32'",
]
[project.urls]
@ -125,7 +127,7 @@ include = ["mineru*"]
namespaces = false
[tool.setuptools.package-data]
"mineru" = ["resources/**"]
"mineru" = ["resources/**", "cli/static/**"]
"mineru.model.utils.pytorchocr.utils" = ["resources/**"]
[tool.setuptools]
@ -162,4 +164,4 @@ exclude_also = [
'if TYPE_CHECKING:',
'class .*\bProtocol\):',
'@(abc\.)?abstractmethod',
]
]

View File

@ -0,0 +1,33 @@
"""Keep the base image's accelerator wheels while resolving application dependencies."""
import argparse
from importlib.metadata import PackageNotFoundError, version
from pathlib import Path
def accelerator_versions():
packages = {}
for name in ('torch', 'torchvision', 'torchaudio', 'vllm', 'triton', 'xformers'):
try:
packages[name] = version(name)
except PackageNotFoundError:
if name in ('torch', 'torchvision', 'vllm'):
raise RuntimeError(f'The accelerator base image must provide {name}') from None
return packages
def main():
parser = argparse.ArgumentParser()
parser.add_argument('path', type=Path)
parser.add_argument('--check', action='store_true')
args = parser.parse_args()
installed = accelerator_versions()
if args.check:
expected = dict(line.split('==', 1) for line in args.path.read_text().splitlines() if line)
if installed != expected:
raise RuntimeError('Accelerator packages changed during dependency installation')
else:
args.path.write_text(''.join(f'{name}=={value}\n' for name, value in installed.items()))
if __name__ == '__main__':
main()

View File

@ -1,34 +1,14 @@
#!/usr/bin/env bash
set -euo pipefail
VERSION=${VERSION:-$(date +%Y%m%d%H%M%S)}
VERSION=${VERSION:-2.7.6}
API_FLAVOR=${API_FLAVOR:-corex}
ROOT_DIR=$(cd "$(dirname "$0")/.." && pwd)
DOCKER=${DOCKER:-docker}
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: $API_FLAVOR. Use corex or cuda." >&2
exit 1
;;
corex) API_DOCKERFILE=docker/china/Dockerfile.corex ;;
cuda) API_DOCKERFILE=docker/global/Dockerfile.cuda ;;
*) echo 'API_FLAVOR must be corex or cuda.' >&2; exit 1 ;;
esac
echo "Building UnisMindMap API image (${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
"$DOCKER" build -f "$API_DOCKERFILE" -t "unis-mindmap-api:${VERSION}-${API_FLAVOR}" -t "unis-mindmap-api:${API_FLAVOR}" .
"$DOCKER" build -f web_ui/Dockerfile.portable -t "unis-mindmap-web:${VERSION}" web_ui

View File

@ -0,0 +1,13 @@
"""Start the checked-out MinerU 2.7.6 application with its normal CLI options."""
from pathlib import Path
import sys
ROOT = Path(__file__).resolve().parents[1]
sys.path.insert(0, str(ROOT))
if __name__ == '__main__':
from mineru.version import __version__
if __version__ != '2.7.6':
raise RuntimeError('This launcher requires MinerU 2.7.6')
from mineru.cli.fast_api import main
main()

View File

@ -1,33 +1,39 @@
#!/usr/bin/env bash
set -euo pipefail
DOCKER=${DOCKER:-docker}
API_FLAVOR=${API_FLAVOR:-corex}
VERSION=${VERSION:-2.7.6}
NETWORK=${NETWORK:-unis-ai}
API_NAME=${API_NAME:-mineru-api}
WEB_NAME=${WEB_NAME:-mindmap-web}
API_NAME=${API_NAME:-mineru-276-api}
WEB_NAME=${WEB_NAME:-mindmap-276-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}
API_PORT=${API_PORT:-8100}
MINERU_MODEL_DIR=${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="$MINERU_API_URL" \
unis-mindmap-web:latest
case "$API_FLAVOR" in
cuda)
GPU_ARGS=(--gpus "device=${GPU_IDS:-0}")
;;
corex)
GPU_ARGS=(--privileged --pid=host --group-add video
-v /dev:/dev -v /usr/src:/usr/src:ro -v /lib/modules:/lib/modules:ro
-e MINERU_VLLM_DEVICE=corex -e "CUDA_VISIBLE_DEVICES=${GPU_IDS:-0}")
;;
*) echo 'API_FLAVOR must be corex or cuda.' >&2; exit 1 ;;
esac
for name in "$API_NAME" "$WEB_NAME"; do
if "$DOCKER" container inspect "$name" >/dev/null 2>&1; then
echo "Container $name already exists. Set a new name or manage it explicitly." >&2
exit 1
fi
done
if [ ! -d "$MINERU_MODEL_DIR/pipeline" ] || [ ! -d "$MINERU_MODEL_DIR/vlm" ]; then
echo 'Prepare the pipeline and vlm model directories before starting.' >&2
exit 1
fi
"$DOCKER" network inspect "$NETWORK" >/dev/null 2>&1 || "$DOCKER" network create "$NETWORK"
"$DOCKER" run -d --name "$API_NAME" --network "$NETWORK" "${GPU_ARGS[@]}" --ipc=host --shm-size=32g \
-p "127.0.0.1:${API_PORT}:8000" -v "${MINERU_MODEL_DIR}:/models:ro" \
-e MINERU_MODEL_SOURCE=local -e "MINERU_GPU_MEMORY_UTILIZATION=${MINERU_GPU_MEMORY_UTILIZATION:-0.5}" \
"unis-mindmap-api:${VERSION}-${API_FLAVOR}"
"$DOCKER" run -d --name "$WEB_NAME" --network "$NETWORK" -p "${WEB_PORT}:9898" \
-e "MINERU_API_URL=http://${API_NAME}:8000" "unis-mindmap-web:${VERSION}"

View File

@ -0,0 +1,137 @@
"""Protocol regressions without loading models or accelerator runtimes."""
import ast
import importlib.util
import json
import os
from pathlib import Path
import sys
import tempfile
from types import ModuleType
import unittest
from unittest.mock import Mock, patch
import zipfile
from io import BytesIO
ROOT = Path(__file__).resolve().parents[2]
sys.path.insert(0, str(ROOT))
class HeadingTests(unittest.TestCase):
def test_markdown_and_content_lists_keep_levels_five_and_six(self):
from mineru.utils.enum_class import BlockType, ContentType, MakeMode
for backend, render in (
('pipeline', 'make_blocks_to_markdown'), ('vlm', 'mk_blocks_to_markdown'),
):
path = ROOT / f'mineru/backend/{backend}/{backend}_middle_json_mkcontent.py'
source = ast.parse(path.read_text(encoding='utf-8'))
names = {'get_title_level', render, 'make_blocks_to_content_list'}
functions = ast.Module(body=[node for node in source.body if isinstance(node, ast.FunctionDef)
and node.name in names], type_ignores=[])
scope = {'BlockType': BlockType, 'ContentType': ContentType, 'MakeMode': MakeMode,
'merge_para_with_text': lambda block: block['text']}
exec(compile(functions, str(path), 'exec'), scope)
for level in (1, 4, 5, 6, 9):
with self.subTest(backend=backend, level=level):
block = {'type': BlockType.TITLE, 'level': level, 'text': 'Heading', 'bbox': [0, 0, 10, 10]}
args = ([block], MakeMode.MM_MD)
if backend == 'vlm':
args += (True, True)
self.assertEqual(scope[render](*args), ['#' * min(level, 6) + ' Heading'])
content = scope['make_blocks_to_content_list'](block, '', 0, [100, 100])
self.assertEqual(content['text_level'], min(level, 6))
class ProtocolTests(unittest.TestCase):
@classmethod
def setUpClass(cls):
common = ModuleType('mineru.cli.common')
common.aio_do_parse = Mock()
common.read_fn = lambda path: Path(path).read_bytes()
common.pdf_suffixes, common.image_suffixes, common.word_suffixes = ['pdf'], ['png'], ['docx', 'doc']
config = ModuleType('mineru.utils.config_reader')
config.get_device = lambda: 'cpu'
models = ModuleType('mineru.utils.model_utils')
models.clean_memory = Mock()
suffixes = ModuleType('mineru.utils.guess_suffix_or_lang')
suffixes.guess_suffix_by_path = lambda path: Path(path).suffix.lstrip('.')
cls.stubs = patch.dict(sys.modules, {module.__name__: module for module in (common, config, models, suffixes)})
cls.stubs.start()
spec = importlib.util.spec_from_file_location('parser276_api_test', ROOT / 'mineru/cli/fast_api.py')
cls.api = importlib.util.module_from_spec(spec)
spec.loader.exec_module(cls.api)
@classmethod
def tearDownClass(cls):
cls.stubs.stop()
def setUp(self):
from fastapi.testclient import TestClient
self.temp = tempfile.TemporaryDirectory(dir=ROOT / 'tests')
self.addCleanup(self.temp.cleanup)
self.directory = Path(self.temp.name)
temp_patch = patch.object(tempfile, 'tempdir', self.temp.name)
temp_patch.start()
self.addCleanup(temp_patch.stop)
self.api._task_progress_dir = self.directory / 'progress'
self.api._task_progress.clear()
self.api.app = self.api.create_app()
self.client = TestClient(self.api.app)
self.addCleanup(self.client.close)
async def fake_parse(**kwargs):
self.engine_options = kwargs
for name in kwargs['pdf_file_names']:
backend = kwargs['backend']
subdir = kwargs['parse_method'] if backend == 'pipeline' else (
'vlm' if backend.startswith('vlm') else 'hybrid_' + kwargs['parse_method'])
folder = Path(kwargs['output_dir']) / name / subdir
(folder / 'images').mkdir(parents=True)
(folder / f'{name}.md').write_text('# Chapter\n##### Detail\n###### Leaf\n![Figure](images/a.jpg)')
for suffix, data in (('_middle.json', {'pdf_info': [{'level': 6}]}),
('_model.json', [{'level': 6}]), ('_content_list.json', [{'text_level': 6}])):
(folder / (name + suffix)).write_text(json.dumps(data))
(folder / 'images/a.jpg').write_bytes(b'test-image')
self.parse_patch = patch.object(self.api, 'aio_do_parse', fake_parse)
self.parse_patch.start()
self.addCleanup(self.parse_patch.stop)
def request(self, route='/file_parse', **options):
fields = {'backend': 'pipeline', 'output_dir': str(self.directory / 'output'),
'return_md': 'true', 'return_middle_json': 'true', 'return_images': 'true',
'return_model_output': 'true', 'return_content_list': 'true', **options}
return self.client.post(route, files={'files': ('test276.pdf', b'%PDF-test', 'application/pdf')}, data=fields)
def test_maxkb_and_application_routes_return_complete_results_and_clean_outputs(self):
for route in ('/file_parse', '/api/file_parse'):
with self.subTest(route=route):
response = self.request(route)
self.assertEqual(response.status_code, 200, response.text)
self.assertEqual(response.json()['version'], '2.7.6')
data = response.json()['results']['test276']
self.assertIn('###### Leaf', data['md_content'])
self.assertEqual(json.loads(data['middle_json'])['pdf_info'][0]['level'], 6)
self.assertEqual(json.loads(data['content_list'])[0]['text_level'], 6)
self.assertTrue(data['images']['a.jpg'].startswith('data:image/jpeg;base64,'))
self.assertEqual(list((self.directory / 'output').iterdir()), [])
def test_zip_contains_requested_json_and_images(self):
response = self.request(response_format_zip='true')
self.assertEqual(response.status_code, 200, response.text)
with zipfile.ZipFile(BytesIO(response.content)) as archive:
self.assertEqual(set(archive.namelist()), {
'test276/test276.md', 'test276/test276_middle.json', 'test276/test276_model.json',
'test276/test276_content_list.json', 'test276/images/a.jpg'})
def test_gpu_option_reaches_the_served_app_and_invalid_value_fails(self):
from click.testing import CliRunner
with patch.object(self.api.uvicorn, 'run') as serve, patch.dict(os.environ, {'MINERU_GPU_MEMORY_UTILIZATION': '0.4'}):
result = CliRunner().invoke(self.api.main, ['--host', '127.0.0.1'])
self.assertEqual(result.exit_code, 0, result.output)
self.assertIs(serve.call_args.args[0], self.api.app)
self.assertEqual(self.api.app.state.config['gpu_memory_utilization'], 0.4)
with patch.dict(os.environ, {'MINERU_GPU_MEMORY_UTILIZATION': '1.1'}):
self.assertNotEqual(CliRunner().invoke(self.api.main, []).exit_code, 0)
if __name__ == '__main__':
unittest.main()

View File

@ -0,0 +1,8 @@
node_modules
dist
.git
.idea
*.log
*.pid
.env
.env.*

View File

@ -5,7 +5,7 @@ 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
RUN npm config set registry https://registry.npmmirror.com && npm ci
COPY . .
RUN npm run build