#925 by GitHub stars

xiaomi-research/Xiaomi-OCR-0

Xiaomi-OCR-0

History

Review weekly changes to SKILL.md before updating. Dates mark the weeks in which we observed each version; checks apply to the recorded commits.

View weekly changes

`SKILL.md`

Week of Nov 2, 2026 · +0 / −0 lines

Before Not checked yet → After Network or broad shell

67ac29a85150 → b1a9bb4c0706

Diff text is unavailable under the recorded licenses. Line counts and the GitHub link remain available.

Compare on GitHub

Week of Oct 26, 2026 · Line counts unavailable

Before Not checked yet → After Not checked yet

First observation → 67ac29a85150

No earlier version was observed for comparison.

Compare on GitHub

`skills/xiaomi-ocr/SKILL.md`

Week of Nov 2, 2026 · +0 / −0 lines

Before Not checked yet → After Network or broad shell

67ac29a85150 → b1a9bb4c0706

Diff text is unavailable under the recorded licenses. Line counts and the GitHub link remain available.

Compare on GitHub

Week of Oct 26, 2026 · Line counts unavailable

Before Not checked yet → After Not checked yet

First observation → 67ac29a85150

No earlier version was observed for comparison.

Compare on GitHub

View on GitHub Add to my skills

Verified commit b1a9bb4c07060c453785e9151715ccda6b96f0b1

963GitHub stars

587stars in 7 days

2skills in this repository

6commits in 30 days

Heat

8.0

Star growth, last 7 daysAhead of 100% of tracked repositories
Star growth, last 24 hoursAhead of 99% of tracked repositories
Rank movement, last 7 daysNot available
Commits, last 30 daysAhead of 71% of tracked repositories
Accounts asking about it through MCP, last 7 daysNot available
Hot votes this weekAhead of 0% of tracked repositories

Heat never changes the order by stars, and sponsorship never touches it. How heat works

Stars over time

30-day change not yet available: not enough history.

Individual skills in this repository

xiaomi-ocr-setup /
xiaomi-ocr skills/xiaomi-ocr

How it works

Choose a skill

xiaomi-ocr-setup

When your agent uses it

Install and configure the Xiaomi-OCR-0 Skill and local MCP service for document OCR, KIE, and VQA.

What it loads into context

Frontmatter

36 tokens

Always available to the agent

SKILL.md

840 tokens

When the skill is activated

Reference documents

21,638 tokens

Loaded on request

Token counts are estimates. Scripts run separately; they are not counted as reference context.

The playbook

Install this Skill into the current agent

Scripts executed separately

demo/server.py
demo/static/app.js
demo/static/i18n.js
demo/static/katex/katex.min.js
evaluation_pipeline/__init__.py
evaluation_pipeline/cropping.py
evaluation_pipeline/e2e_img2md.py
evaluation_pipeline/layout_detect_paddlex.py
evaluation_pipeline/twostage_img2md.py
postprocess/__init__.py
postprocess/assemble_markdown.py
postprocess/document.py
postprocess/markdown_merge.py
postprocess/otsl.py
postprocess/region.py
postprocess/repetition_guard.py

Files

.gitignoreasset281 bytes
INSTALL.mddoc7,563 bytes
LICENSEasset11,349 bytes
OLLAMA.mddoc4,332 bytes
README.mddoc15,332 bytes
README_zh.mddoc13,473 bytes
SKILL.mdcore3,821 bytes
SKILL.mdcore3,821 bytes
SKILL.mdcore3,821 bytes
SKILL.mdcore3,821 bytes
THIRD_PARTY_NOTICES.mddoc773 bytes
demo/README.mddoc4,520 bytes
demo/samples/sample_document.pngasset103,311 bytes
demo/samples/sample_kie.pngasset48,028 bytes
demo/samples/sample_vqa.pngasset43,328 bytes
demo/samples/showcase/showcase_calligraphy.jpgasset169,004 bytes
demo/samples/showcase/showcase_calligraphy.pngasset191,662 bytes
demo/samples/showcase/showcase_dense.pngasset596,294 bytes
demo/samples/showcase/showcase_formula.pngasset614,828 bytes
demo/samples/showcase/showcase_handwriting.pngasset576,680 bytes
demo/samples/showcase/showcase_invoice.pngasset355,436 bytes
demo/samples/showcase/showcase_whiteboard.pngasset262,501 bytes
demo/server.pyruns50,182 bytes
demo/static/app.jsruns49,318 bytes
demo/static/favicon.icoasset4,286 bytes
demo/static/favicon.pngasset299 bytes
demo/static/i18n.jsruns11,678 bytes
demo/static/img/xiaomi-logo.svgasset1,852 bytes
demo/static/index.htmlasset10,472 bytes
demo/static/katex/LICENSEasset1,107 bytes
demo/static/katex/fonts/KaTeX_AMS-Regular.ttfasset63,632 bytes
demo/static/katex/fonts/KaTeX_AMS-Regular.woffasset33,516 bytes
demo/static/katex/fonts/KaTeX_AMS-Regular.woff2asset28,076 bytes
demo/static/katex/fonts/KaTeX_Caligraphic-Bold.ttfasset12,368 bytes
demo/static/katex/fonts/KaTeX_Caligraphic-Bold.woffasset7,716 bytes
demo/static/katex/fonts/KaTeX_Caligraphic-Bold.woff2asset6,912 bytes
demo/static/katex/fonts/KaTeX_Caligraphic-Regular.ttfasset12,344 bytes
demo/static/katex/fonts/KaTeX_Caligraphic-Regular.woffasset7,656 bytes
demo/static/katex/fonts/KaTeX_Caligraphic-Regular.woff2asset6,908 bytes
demo/static/katex/fonts/KaTeX_Fraktur-Bold.ttfasset19,584 bytes
demo/static/katex/fonts/KaTeX_Fraktur-Bold.woffasset13,296 bytes
demo/static/katex/fonts/KaTeX_Fraktur-Bold.woff2asset11,348 bytes
demo/static/katex/fonts/KaTeX_Fraktur-Regular.ttfasset19,572 bytes
demo/static/katex/fonts/KaTeX_Fraktur-Regular.woffasset13,208 bytes
demo/static/katex/fonts/KaTeX_Fraktur-Regular.woff2asset11,316 bytes
demo/static/katex/fonts/KaTeX_Main-Bold.ttfasset51,336 bytes
demo/static/katex/fonts/KaTeX_Main-Bold.woffasset29,912 bytes
demo/static/katex/fonts/KaTeX_Main-Bold.woff2asset25,324 bytes
demo/static/katex/fonts/KaTeX_Main-BoldItalic.ttfasset32,968 bytes
demo/static/katex/fonts/KaTeX_Main-BoldItalic.woffasset19,412 bytes
demo/static/katex/fonts/KaTeX_Main-BoldItalic.woff2asset16,780 bytes
demo/static/katex/fonts/KaTeX_Main-Italic.ttfasset33,580 bytes
demo/static/katex/fonts/KaTeX_Main-Italic.woffasset19,676 bytes
demo/static/katex/fonts/KaTeX_Main-Italic.woff2asset16,988 bytes
demo/static/katex/fonts/KaTeX_Main-Regular.ttfasset53,580 bytes
demo/static/katex/fonts/KaTeX_Main-Regular.woffasset30,772 bytes
demo/static/katex/fonts/KaTeX_Main-Regular.woff2asset26,272 bytes
demo/static/katex/fonts/KaTeX_Math-BoldItalic.ttfasset31,196 bytes
demo/static/katex/fonts/KaTeX_Math-BoldItalic.woffasset18,668 bytes
demo/static/katex/fonts/KaTeX_Math-BoldItalic.woff2asset16,400 bytes
demo/static/katex/fonts/KaTeX_Math-Italic.ttfasset31,308 bytes
demo/static/katex/fonts/KaTeX_Math-Italic.woffasset18,748 bytes
demo/static/katex/fonts/KaTeX_Math-Italic.woff2asset16,440 bytes
demo/static/katex/fonts/KaTeX_SansSerif-Bold.ttfasset24,504 bytes
demo/static/katex/fonts/KaTeX_SansSerif-Bold.woffasset14,408 bytes
demo/static/katex/fonts/KaTeX_SansSerif-Bold.woff2asset12,216 bytes
demo/static/katex/fonts/KaTeX_SansSerif-Italic.ttfasset22,364 bytes
demo/static/katex/fonts/KaTeX_SansSerif-Italic.woffasset14,112 bytes
demo/static/katex/fonts/KaTeX_SansSerif-Italic.woff2asset12,028 bytes
demo/static/katex/fonts/KaTeX_SansSerif-Regular.ttfasset19,436 bytes
demo/static/katex/fonts/KaTeX_SansSerif-Regular.woffasset12,316 bytes
demo/static/katex/fonts/KaTeX_SansSerif-Regular.woff2asset10,344 bytes
demo/static/katex/fonts/KaTeX_Script-Regular.ttfasset16,648 bytes
demo/static/katex/fonts/KaTeX_Script-Regular.woffasset10,588 bytes
demo/static/katex/fonts/KaTeX_Script-Regular.woff2asset9,644 bytes
demo/static/katex/fonts/KaTeX_Size1-Regular.ttfasset12,228 bytes
demo/static/katex/fonts/KaTeX_Size1-Regular.woffasset6,496 bytes
demo/static/katex/fonts/KaTeX_Size1-Regular.woff2asset5,468 bytes
demo/static/katex/fonts/KaTeX_Size2-Regular.ttfasset11,508 bytes
demo/static/katex/fonts/KaTeX_Size2-Regular.woffasset6,188 bytes
demo/static/katex/fonts/KaTeX_Size2-Regular.woff2asset5,208 bytes
demo/static/katex/fonts/KaTeX_Size3-Regular.ttfasset7,588 bytes
demo/static/katex/fonts/KaTeX_Size3-Regular.woffasset4,420 bytes
demo/static/katex/fonts/KaTeX_Size3-Regular.woff2asset3,624 bytes
demo/static/katex/fonts/KaTeX_Size4-Regular.ttfasset10,364 bytes
demo/static/katex/fonts/KaTeX_Size4-Regular.woffasset5,980 bytes
demo/static/katex/fonts/KaTeX_Size4-Regular.woff2asset4,928 bytes
demo/static/katex/fonts/KaTeX_Typewriter-Regular.ttfasset27,556 bytes
demo/static/katex/fonts/KaTeX_Typewriter-Regular.woffasset16,028 bytes
demo/static/katex/fonts/KaTeX_Typewriter-Regular.woff2asset13,568 bytes
demo/static/katex/katex.min.cssasset23,335 bytes
demo/static/katex/katex.min.jsruns275,414 bytes
demo/static/showcase.jsonasset605 bytes
demo/static/style.cssasset17,943 bytes
evaluation_pipeline/README.mddoc8,614 bytes
evaluation_pipeline/README_zh.mddoc8,063 bytes
evaluation_pipeline/__init__.pyruns0 bytes
evaluation_pipeline/configs/end2end.yamlasset1,153 bytes
evaluation_pipeline/cropping.pyruns2,453 bytes
evaluation_pipeline/e2e_img2md.pyruns29,344 bytes
evaluation_pipeline/layout_detect_paddlex.pyruns14,761 bytes
evaluation_pipeline/twostage_img2md.pyruns46,122 bytes
example_pics/dense-equations.gifasset2,452,638 bytes
example_pics/dense-equations.jpgasset401,147 bytes
example_pics/dense-equations.mddoc4,606 bytes
example_pics/handwrite-formula.gifasset807,861 bytes
example_pics/handwrite-formula.jpgasset247,100 bytes
example_pics/handwrite-formula.mddoc2,152 bytes
example_pics/qianziwen.jpegasset295,671 bytes
example_pics/qianziwen.mddoc83 bytes
example_pics/receipt.jpgasset99,714 bytes
example_pics/receipt.mddoc530 bytes
examples/mcp-config.jsonasset372 bytes
postprocess/README.mddoc1,726 bytes
postprocess/__init__.pyruns0 bytes
postprocess/assemble_markdown.pyruns28,533 bytes
postprocess/document.pyruns3,666 bytes
postprocess/markdown_merge.pyruns54,539 bytes
postprocess/otsl.pyruns8,094 bytes
postprocess/region.pyruns14,352 bytes
postprocess/repetition_guard.pyruns4,514 bytes
requirements-paddlex.txtdoc386 bytes
requirements.txtdoc59 bytes

Read SKILL.md


# Xiaomi-OCR-0 Skill

The only user-facing installation prompt is:

```text
Read and execute https://raw.githubusercontent.com/SeerRay-Lab/Xiaomi-OCR-0/main/SKILL.md
```

## Install this Skill into the current agent

If this file was read from the Xiaomi-OCR-0 repository and the user asked to install it,
install only the `xiaomi-ocr` Skill globally for the current agent:

```bash

View the complete source on GitHub

xiaomi-ocr

When your agent uses it

Run Xiaomi-OCR-0 locally through MCP to parse document images and PDFs, extract key fields (KIE), or answer questions about document images (VQA).

What it loads into context

Frontmatter

42 tokens

Always available to the agent

SKILL.md

2,315 tokens

When the skill is activated

Reference documents

1,090 tokens

Loaded on request

Token counts are estimates. Scripts run separately; they are not counted as reference context.

The playbook

Set up local inference
Choose the right task and document mode
Output handling

Scripts executed separately

install.sh
mcp_ocr_server.py
scripts/otsl_convert.py
scripts/register_codex_mcp.py
scripts/selftest.py

Files

../../OLLAMA.mddoc4,332 bytes
SKILL.mdcore9,900 bytes
SKILL.mdcore9,900 bytes
SKILL.mdcore9,900 bytes
install.shruns2,909 bytes
mcp_ocr_server.pyruns19,441 bytes
requirements.txtdoc61 bytes
scripts/otsl_convert.pyruns5,732 bytes
scripts/register_codex_mcp.pyruns3,000 bytes
scripts/selftest.pyruns1,211 bytes

Read SKILL.md


# Xiaomi-OCR-0

Use this Skill when a task needs OCR, document parsing, KIE, or OCR-VQA. Inference runs
on the user's machine with the Hugging Face checkpoint `SeerRay-Lab/Xiaomi-OCR-0` and
SGLang, vLLM, or the experimental Ollama backend. The MCP service owns prompts, image/PDF preparation, optional layout
segmentation, crop processing, concurrency, and output formatting. Call the task tool;
do not recreate these steps with separate tools or direct model requests.

## Set up local inference

### Choose the inference backend

Respect an explicit backend/model preference and the user's accuracy requirements. Otherwise:

View the complete source on GitHub

Skills included

xiaomi-ocr-setup
xiaomi-ocr

Repository facts

License: Apache-2.0
Created: Sep 28, 2026
Last pushed: Oct 29, 2026
Forks: 58
Open issues: 4
Project activity: Active

Rank over the last 12 weeks

Now #925 · best #925

#999 on Nov 3
#962 on Nov 4
#935 on Nov 5
#928 on Nov 6
#925 since Nov 7

Before you install

Network or broad shell

Automated signals, not a guarantee

Checked Nov 4, 2026

Not fully analyzed: some files were too long, unreadable or missing. This never lowers the level.

Show 11 notes

SKILL.md:1 Analysis incomplete: Referenced executable target unavailable; not analyzed
SKILL.md:1 Analysis incomplete: Referenced executable target unavailable; not analyzed
SKILL.md:1 Analysis incomplete: Referenced executable target unavailable; not analyzed
demo/static/katex/katex.min.css:1 Analysis incomplete: physical or joined line exceeds character limit
demo/static/katex/katex.min.js:1 [Credential-related source omitted]
demo/static/style.css:5 Analysis incomplete: physical or joined line exceeds character limit
demo/static/style.css:6 Analysis incomplete: physical or joined line exceeds character limit
demo/static/style.css:7 Analysis incomplete: physical or joined line exceeds character limit
demo/static/style.css:8 Analysis incomplete: physical or joined line exceeds character limit
skills/xiaomi-ocr/SKILL.md:1 Analysis incomplete: Referenced executable target unavailable; not analyzed
skills/xiaomi-ocr/SKILL.md:1 Analysis incomplete: Referenced executable target unavailable; not analyzed

No broad tool permissions
2 dangerous shell commandsin 1 file: example_pics/receipt.md:25
223 network referencesin 21 files: README.md:5 README_zh.md:5 demo/static/app.js:934 and 18 more
7 obfuscation signalsin 3 files: demo/server.py:62 demo/static/katex/katex.min.js:1 xiaomi-ocr/mcp_ocr_server.py:121
No instruction overrides
21 executable scriptsin 21 files: demo/server.py:1 demo/static/app.js:1 demo/static/i18n.js:1 and 18 more
24 references to secretsin 6 files: xiaomi-ocr/mcp_ocr_server.py:26 xiaomi-ocr/scripts/register_codex_mcp.py:40 demo/server.py:85 and 3 more
No requests to vote in Hot

Show all 277 signals by file

`INSTALL.md`

Network accessline 29

- [SGLang Qwen3.5 deployment guide](https://github.com/sgl-project/sglang/blob/main/docs_new/cookbook/autoregressive/Qwen/Qwen3.5.mdx)

Network accessline 30

- [vLLM Qwen3.5 implementation](https://docs.vllm.ai/en/latest/api/vllm/model_executor/models/qwen3_5/)

Network accessline 31

- [vLLM supported models](https://docs.vllm.ai/en/latest/models/supported_models/)

Network accessline 53

Install [Ollama](https://ollama.com/download) and keep its local service running

Network accessline 62

Use `http://127.0.0.1:11434/v1` as the MCP inference URL and

Network accessline 73

XIAOMI_OCR_LOCAL_URL=http://127.0.0.1:8000/v1 \

Network accessline 81

XIAOMI_OCR_LOCAL_URL=http://127.0.0.1:11434/v1 \

Network accessline 92

own MCP configuration, then restart the agent. If the local model runs on another port, set `XIAOMI_OCR_LOCAL_URL` before running the installer, for example `http://127.0.0.1:8001/v1`.

Network accessline 94
```
`http://127.0.0.1:8001/v1`.
```

Network accessline 103

For manual installation, use the official [PaddlePaddle installation guide](https://paddlepaddle.github.io/PaddleX/latest/en/installation/paddlepaddle_install.html)

Network accessline 104

[PaddlePaddle installation guide](https://paddlepaddle.github.io/PaddleX/latest/en/installation/paddlepaddle_install.html)

Network accessline 105

and [PaddleX installation guide](https://paddlepaddle.github.io/PaddleX/latest/en/installation/installation.html).

`LICENSE`

Network accessline 1

Apache License                            Version 2.0, January 2004                         http://www.apache.org/licenses/

Network accessline 3
```
http://www.apache.org/licenses/
```

Network accessline 195

http://www.apache.org/licenses/LICENSE-2.0

`OLLAMA.md`

Network accessline 41

export XIAOMI_OCR_BACKEND=ollama export XIAOMI_OCR_MODEL=longwayxu/xiaomi-ocr-0:bf16 export XIAOMI_OCR_LOCAL_URL=http://127.0.0.1:11434/v1

Network accessline 41

export XIAOMI_OCR_BACKEND=ollama export XIAOMI_OCR_MODEL=longwayxu/xiaomi-ocr-0:bf16 export XIAOMI_OCR_LOCAL_URL=http://127.0.0.1:11434/v1

Network accessline 42

export XIAOMI_OCR_MODEL=longwayxu/xiaomi-ocr-0:bf16 export XIAOMI_OCR_LOCAL_URL=http://127.0.0.1:11434/v1 python demo/server.py --port 8787

Network accessline 42

export XIAOMI_OCR_MODEL=longwayxu/xiaomi-ocr-0:bf16 export XIAOMI_OCR_LOCAL_URL=http://127.0.0.1:11434/v1 python demo/server.py --port 8787

Network accessline 43

export XIAOMI_OCR_LOCAL_URL=http://127.0.0.1:11434/v1

Network accessline 43

export XIAOMI_OCR_LOCAL_URL=http://127.0.0.1:11434/v1

Network accessline 59

XIAOMI_OCR_BACKEND=ollama \ XIAOMI_OCR_LOCAL_URL=http://127.0.0.1:11434/v1 \

Network accessline 59

XIAOMI_OCR_BACKEND=ollama \ XIAOMI_OCR_LOCAL_URL=http://127.0.0.1:11434/v1 \

Network accessline 60

XIAOMI_OCR_LOCAL_URL=http://127.0.0.1:11434/v1 \

Network accessline 60

XIAOMI_OCR_LOCAL_URL=http://127.0.0.1:11434/v1 \

`README.md`

Network accessline 5

<p><b>English</b> · <a href="README_zh.md">简体中文</a></p> <p>   <a href="https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/Hugging%20Face-Model-FFD21E?logo=huggingface&amp;logoColor=FFD21E" alt="Hugging F

Network accessline 6

<p>   <a href="https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/Hugging%20Face-Model-FFD21E?logo=huggingface&amp;logoColor=FFD21E" alt="Hugging Face Model"></a>   <a href="https://arxiv.org/abs/2609.36

Network accessline 7

<a href="https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/Hugging%20Face-Model-FFD21E?logo=huggingface&amp;logoColor=FFD21E" alt="Hugging Face Model"></a>

Network accessline 8

<a href="https://arxiv.org/abs/2609.36136"><img src="https://img.shields.io/badge/arXiv-2609.36136-b31b1b.svg" alt="arXiv:2609.36136"></a>

Network accessline 9

<a href="https://huggingface.co/spaces/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/Project-Page-FF6900?logo=huggingface&amp;logoColor=FFD21E" alt="Project Page"></a>

Network accessline 10

<a href="https://github.com/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/GitHub-SeerRay--Lab%2FXiaomi--OCR--0-181717?logo=github" alt="GitHub"></a>

Network accessline 11

<a href="https://ollama.com/longwayxu/xiaomi-ocr-0:bf16"><img src="https://img.shields.io/badge/Ollama-BF16%20%28experimental%29-000000?logo=ollama&amp;logoColor=white" alt="Ollama BF16 (experimental)"></a>

Network accessline 12

<a href="https://www.modelscope.cn/models/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/ModelScope-Model-624AFF?logo=ModelScope&amp;logoColor=white" alt="ModelScope Model"></a>

Network accessline 32

- **2026-10-08:** Thanks to the community for supporting Xiaomi-OCR-0 and sharing new formats: [prithivMLmods’ GGUF](https://huggingface.co/prithivMLmods/Xiaomi-OCR-0-GGUF), [xzl01’s GGUF](https://github.com/xzl01/Xiaomi-OCR-0-GGUF), and [B

Network accessline 52

| [`evaluation_pipeline/`](evaluation_pipeline/) | Reproducible e2e / two-stage inference and evaluation | | [`postprocess/`](postprocess/) | Shared table conversion, formula/text assembly and repetition handling | | [Hugging Face model](ht

Network accessline 53

| [`postprocess/`](postprocess/) | Shared table conversion, formula/text assembly and repetition handling | | [Hugging Face model](https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0) | Model card, examples, and benchmark assets | | [ModelScope

Network accessline 54

| [Hugging Face model](https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0) | Model card, examples, and benchmark assets |

Network accessline 55

| [ModelScope model](https://www.modelscope.cn/models/SeerRay-Lab/Xiaomi-OCR-0) | Model card, examples, and benchmark assets |

Network accessline 56

| [Hugging Face Space](https://huggingface.co/spaces/SeerRay-Lab/Xiaomi-OCR-0/tree/main) | Space assets and project-page source |

Network accessline 73

Read and execute https://raw.githubusercontent.com/SeerRay-Lab/Xiaomi-OCR-0/main/SKILL.md

Network accessline 90

git clone https://github.com/SeerRay-Lab/Xiaomi-OCR-0.git

Network accessline 96

You can run the browser demo without installing the Agent Skill or MCP server. The demo sends requests to a local SGLang or vLLM inference server at `http://127.0.0.1:8000/v1`.

Network accessline 102

- [SGLang installation guide](https://docs.sglang.ai/get_started/install.html) · [supported models](https://docs.sglang.io/docs/supported-models/multimodal_language_models.md)

Network accessline 103

- [vLLM GPU installation guide](https://docs.vllm.ai/en/latest/getting_started/installation/gpu/) · [Qwen3.5 model support](https://docs.vllm.ai/en/latest/api/vllm/model_executor/models/qwen3_5/)

Network accessline 138

Open <http://127.0.0.1:8787>. The interface supports English and Chinese; switch in the header or open <http://127.0.0.1:8787/?lang=en> for English. The same browser page supports whole-page/region document parsing, KIE, VQA, and PDF parsin

Network accessline 142

The [community BF16 GGUF package](https://ollama.com/longwayxu/xiaomi-ocr-0:bf16) includes the text model and vision projector (about 1.8 GB). Install [Ollama](https://ollama.com/download) and keep its local service running, then pull the m

Network accessline 156

export XIAOMI_OCR_BACKEND=ollama export XIAOMI_OCR_MODEL=longwayxu/xiaomi-ocr-0:bf16 export XIAOMI_OCR_LOCAL_URL=http://127.0.0.1:11434/v1

Network accessline 157

export XIAOMI_OCR_MODEL=longwayxu/xiaomi-ocr-0:bf16 export XIAOMI_OCR_LOCAL_URL=http://127.0.0.1:11434/v1 python demo/server.py --port 8787

Network accessline 158

export XIAOMI_OCR_LOCAL_URL=http://127.0.0.1:11434/v1

Network accessline 162

Open <http://127.0.0.1:8787>. Whole-page OCR, PDF parsing, KIE and VQA share the existing pipeline. Region mode additionally requires the layout dependencies described below. See [Ollama setup and validation notes](OLLAMA.md) for direct MCP

Network accessline 168

Region mode additionally needs PaddlePaddle, PaddleX and PP-DocLayoutV3. In the Python environment that runs the Demo or MCP server, follow the official [PaddlePaddle installation guide](https://paddlepaddle.github.io/PaddleX/latest/en/inst

Network accessline 169

environment that runs the Demo or MCP server, follow the official [PaddlePaddle installation guide](https://paddlepaddle.github.io/PaddleX/latest/en/installation/paddlepaddle_install.html) to select a compatible CPU/GPU build, then the

Network accessline 170

[PaddlePaddle installation guide](https://paddlepaddle.github.io/PaddleX/latest/en/installation/paddlepaddle_install.html)

Network accessline 171

to select a compatible CPU/GPU build, then the [PaddleX installation guide](https://paddlepaddle.github.io/PaddleX/latest/en/installation/installation.html). From the repository root, install the layout dependencies:

Network accessline 172

[PaddleX installation guide](https://paddlepaddle.github.io/PaddleX/latest/en/installation/installation.html).

Network accessline 188

[prithivMLmods/Xiaomi-OCR-0-GGUF](https://huggingface.co/prithivMLmods/Xiaomi-OCR-0-GGUF) provides `BF16`, `Q3_K_L`, `Q3_K_M`, `Q4_K_M`, `Q4_K_S`, `Q5_K_M`, `Q5_K_S` and `Q6_K` weights. Image inference also requires the repository's `Xiaomi

Network accessline 190

[xzl01/Xiaomi-OCR-0-GGUF](https://github.com/xzl01/Xiaomi-OCR-0-GGUF) also provides Q4_K_M and Q8_0 llama.cpp weights with an F16 vision projector. See the repository for its conversion scripts and measured results.

Network accessline 194

[ByronLeeee/Xiaomi-OCR-0-Ninfer](https://huggingface.co/ByronLeeee/Xiaomi-OCR-0-Ninfer) provides a BF16 NInfer package with language and vision weights, tokenizer, and image-processing configuration. See its model page for the build guide a

Network accessline 198

![Parameter efficiency on OmniDocBench v1.6](https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0/resolve/main/assets/omni-parameters.png)

Network accessline 200

![Parameter efficiency on OCR-oriented VQA](https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0/resolve/main/assets/vqa-parameters.png)

Network accessline 202

![Parameter efficiency on Real5-OmniDocBench](https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0/resolve/main/assets/real5-parameters.png)

Network accessline 204

![Parameter efficiency on Wild-OmniDocBench](https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0/resolve/main/assets/wild-parameters.png)

Network accessline 248

archivePrefix={arXiv},       primaryClass={cs.CV},       url={https://arxiv.org/abs/2609.36136},

Network accessline 249

primaryClass={cs.CV},       url={https://arxiv.org/abs/2609.36136}, }

Network accessline 250
```
url={https://arxiv.org/abs/2609.36136},
```

`README_zh.md`

Network accessline 5

<p><a href="README.md">English</a> · <b>简体中文</b></p> <p>   <a href="https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/Hugging%20Face-Model-FFD21E?logo=huggingface&amp;logoColor=FFD21E" alt="Hugging Face

Network accessline 6

<p>   <a href="https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/Hugging%20Face-Model-FFD21E?logo=huggingface&amp;logoColor=FFD21E" alt="Hugging Face Model"></a>   <a href="https://arxiv.org/abs/2609.36

Network accessline 7

<a href="https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/Hugging%20Face-Model-FFD21E?logo=huggingface&amp;logoColor=FFD21E" alt="Hugging Face Model"></a>

Network accessline 8

<a href="https://arxiv.org/abs/2609.36136"><img src="https://img.shields.io/badge/arXiv-2609.36136-b31b1b.svg" alt="arXiv:2609.36136"></a>

Network accessline 9

<a href="https://huggingface.co/spaces/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/Project-Page-FF6900?logo=huggingface&amp;logoColor=FFD21E" alt="Project Page"></a>

Network accessline 10

<a href="https://github.com/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/GitHub-SeerRay--Lab%2FXiaomi--OCR--0-181717?logo=github" alt="GitHub"></a>

Network accessline 11

<a href="https://ollama.com/longwayxu/xiaomi-ocr-0:bf16"><img src="https://img.shields.io/badge/Ollama-BF16%20%28experimental%29-000000?logo=ollama&amp;logoColor=white" alt="Ollama BF16 (experimental)"></a>

Network accessline 12

<a href="https://www.modelscope.cn/models/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/ModelScope-Model-624AFF?logo=ModelScope&amp;logoColor=white" alt="ModelScope Model"></a>

Network accessline 29

- **2026-10-08：** 感谢社区对 Xiaomi-OCR-0 的支持与贡献：[prithivMLmods 的 GGUF 版本](https://huggingface.co/prithivMLmods/Xiaomi-OCR-0-GGUF)、[xzl01 的 GGUF 版本](https://github.com/xzl01/Xiaomi-OCR-0-GGUF)，以及 [ByronLeeee 的 BF16 NInfer 转换版](https://huggingfac

Network accessline 49

| [`evaluation_pipeline/`](evaluation_pipeline/) | 整页 / 两阶段推理与评测复现 | | [`postprocess/`](postprocess/) | 表格转换、公式/文本组装与重复内容处理 | | [Hugging Face model](https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0) | huggingface模型卡片、示例和基准图表

Network accessline 50

| [`postprocess/`](postprocess/) | 表格转换、公式/文本组装与重复内容处理 | | [Hugging Face model](https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0) | huggingface模型卡片、示例和基准图表 | [Model Scope model](https://www.modelscope.cn/models/SeerRay-Lab/Xiaomi-OCR-0) | mo

Network accessline 51

| [Hugging Face model](https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0) | huggingface模型卡片、示例和基准图表

Network accessline 52

| [Model Scope model](https://www.modelscope.cn/models/SeerRay-Lab/Xiaomi-OCR-0) | modelscope模型卡片、示例和基准图表|

Network accessline 53

| [Hugging Face Space](https://huggingface.co/spaces/SeerRay-Lab/Xiaomi-OCR-0/tree/main) | Space 资源和项目页源码 |

Network accessline 70

Read and execute https://raw.githubusercontent.com/SeerRay-Lab/Xiaomi-OCR-0/main/SKILL.md

Network accessline 84

git clone https://github.com/SeerRay-Lab/Xiaomi-OCR-0.git

Network accessline 90

不安装 Agent Skill 或 MCP 服务也能运行网页 Demo。Demo 会向本机的 SGLang 或 vLLM 服务 `http://127.0.0.1:8000/v1` 发送请求。

Network accessline 96

- [SGLang 安装指南](https://docs.sglang.ai/get_started/install.html) · [支持的模型](https://docs.sglang.io/docs/supported-models/multimodal_language_models.md)

Network accessline 97

- [vLLM GPU 安装指南](https://docs.vllm.ai/en/latest/getting_started/installation/gpu/) · [Qwen3.5 模型支持](https://docs.vllm.ai/en/latest/api/vllm/model_executor/models/qwen3_5/)

Network accessline 132

打开 <http://127.0.0.1:8787>。界面支持中英文切换，右上角可选择语言；英文直达链接为 <http://127.0.0.1:8787/?lang=en>。同一网页支持整页/区域文档解析、KIE、VQA 和 PDF 解析。上述 `requirements.txt` 已包含图像、PDF 和公共后处理所需依赖。区域模式可选依赖 PaddlePaddle/PaddleX 与 PP-DocLayoutV3 权重。依赖说明见 [demo/README.md](demo

Network accessline 136

[社区 BF16 GGUF 包](https://ollama.com/longwayxu/xiaomi-ocr-0:bf16)包含语言模型和视觉投影器，约 1.8 GB。安装 [Ollama](https://ollama.com/download) 并保持本地服务运行，然后拉取模型：

Network accessline 150

export XIAOMI_OCR_BACKEND=ollama export XIAOMI_OCR_MODEL=longwayxu/xiaomi-ocr-0:bf16 export XIAOMI_OCR_LOCAL_URL=http://127.0.0.1:11434/v1

Network accessline 151

export XIAOMI_OCR_MODEL=longwayxu/xiaomi-ocr-0:bf16 export XIAOMI_OCR_LOCAL_URL=http://127.0.0.1:11434/v1 python demo/server.py --port 8787

Network accessline 152

export XIAOMI_OCR_LOCAL_URL=http://127.0.0.1:11434/v1

Network accessline 156

打开 <http://127.0.0.1:8787>。整页 OCR、PDF 解析、KIE 和 VQA 复用现有流程；区域模式还需安装下文的版面分析依赖。直接安装 MCP 和转换细节见 [Ollama 配置与验证说明](OLLAMA.md)。

Network accessline 160

区域识别额外需要 PaddlePaddle、PaddleX 和 PP-DocLayoutV3。请在运行 Demo 或 MCP 服务的 Python 环境中，先按 [PaddlePaddle 官方安装教程](https://paddlepaddle.github.io/PaddleX/latest/installation/paddlepaddle_install.html)选择适合操作系统、Python 和 CPU/GPU 的版本，再参考 [PaddleX 官方安装教程](h

Network accessline 170

感谢社区对xiaomi-ocr-0的支持, [prithivMLmods/Xiaomi-OCR-0-GGUF](https://huggingface.co/prithivMLmods/Xiaomi-OCR-0-GGUF) 和 [xzl01/Xiaomi-OCR-0-GGUF](https://github.com/xzl01/Xiaomi-OCR-0-GGUF) 提供了GGUF量化权重, [ByronLeeee/Xiaomi-OCR-0-Ninfer](https://hu

Network accessline 175

![OmniDocBench v1.6 参数效率](https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0/resolve/main/assets/omni-parameters.png)

Network accessline 177

![OCR 导向 VQA 参数效率](https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0/resolve/main/assets/vqa-parameters.png)

Network accessline 179

![Real5-OmniDocBench 参数效率](https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0/resolve/main/assets/real5-parameters.png)

Network accessline 181

![Wild-OmniDocBench 参数效率](https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0/resolve/main/assets/wild-parameters.png)

Network accessline 224

archivePrefix={arXiv},       primaryClass={cs.CV},       url={https://arxiv.org/abs/2609.36136},

Network accessline 225

primaryClass={cs.CV},       url={https://arxiv.org/abs/2609.36136}, }

Network accessline 226
```
url={https://arxiv.org/abs/2609.36136},
```

`SKILL.md`

Network accessline 11

Read and execute https://raw.githubusercontent.com/SeerRay-Lab/Xiaomi-OCR-0/main/SKILL.md

Network accessline 30

The runtime imports shared `evaluation_pipeline/` and `postprocess/` modules from the full repository. Reuse a complete checkout or clone `https://github.com/SeerRay-Lab/Xiaomi-OCR-0.git`

Network accessline 31

Reuse a complete checkout or clone `https://github.com/SeerRay-Lab/Xiaomi-OCR-0.git`

`THIRD_PARTY_NOTICES.md`

Network accessline 12

- License: MIT; the complete upstream license is included in   [demo/static/katex/LICENSE](demo/static/katex/LICENSE). - Source: <https://github.com/KaTeX/KaTeX/tree/v0.16.11>.

Network accessline 13

[demo/static/katex/LICENSE](demo/static/katex/LICENSE). - Source: <https://github.com/KaTeX/KaTeX/tree/v0.16.11>. - Distribution: <https://registry.npmjs.org/katex/-/katex-0.16.11.tgz>.

Network accessline 14

- Source: <https://github.com/KaTeX/KaTeX/tree/v0.16.11>.

Network accessline 15

- Distribution: <https://registry.npmjs.org/katex/-/katex-0.16.11.tgz>.

`demo/README.md`

Network accessline 16

Open <http://127.0.0.1:8787>. Set `XIAOMI_OCR_LOCAL_URL` if the inference server uses another local port. The demo only accepts loopback inference URLs.

Network accessline 41

- [PaddlePaddle installation guide](https://paddlepaddle.github.io/PaddleX/latest/en/installation/paddlepaddle_install.html)

Network accessline 42

- [PaddleX installation guide](https://paddlepaddle.github.io/PaddleX/latest/en/installation/installation.html)

Network accessline 43

- [PP-DocLayoutV3 model card and weights](https://huggingface.co/PaddlePaddle/PP-DocLayoutV3)

Network accessline 49

If your inference runtime uses a custom served model name, set `XIAOMI_OCR_MODEL` to the exact ID returned by its `/v1/models` endpoint. `XIAOMI_OCR_LOCAL_URL` defaults to `http://127.0.0.1:8000/v1`. Both variables must be set in the termin

Network accessline 50

to the exact ID returned by its `/v1/models` endpoint. `XIAOMI_OCR_LOCAL_URL` defaults to `http://127.0.0.1:8000/v1`. Both variables must be set in the terminal that launches the demo. The connection indicator checks the configured model vi

Network accessline 51

defaults to `http://127.0.0.1:8000/v1`. Both variables must be set in the terminal

`demo/server.py`

Executable scriptline 1
```
#!/usr/bin/env python3
```

Possible obfuscationline 62

normalized = mcp_pipeline()._image_png(base64.b64decode(image_b64, validate=True))

Network accessline 85
```
[Credential-related source omitted]
```
Access to secretsline 85
```
[Credential-related source omitted]
```
Network accessline 86
```
[Credential-related source omitted]
```
Access to secretsline 86
```
[Credential-related source omitted]
```
Access to secretsline 87
```
[Credential-related source omitted]
```

Network accessline 152

base = base.removesuffix("/chat/completions").rstrip("/")     url = base + ("/models" if base.endswith("/v1") else "/v1/models")     with urllib.request.urlopen(url, timeout=3) as response:

Network accessline 153

url = base + ("/models" if base.endswith("/v1") else "/v1/models")     with urllib.request.urlopen(url, timeout=3) as response:         models = json.load(response).get("data", [])

Network accessline 154

with urllib.request.urlopen(url, timeout=3) as response:

Possible obfuscationline 168
```
raw = base64.b64decode(image_b64)
```

Network accessline 438

"Accept": "application/json",     }     req = urllib.request.Request(url, data=data, headers=headers, method="POST")

Network accessline 439

}     req = urllib.request.Request(url, data=data, headers=headers, method="POST")     t0 = time.perf_counter()

Network accessline 440

req = urllib.request.Request(url, data=data, headers=headers, method="POST")

Network accessline 441

t0 = time.perf_counter()     try:         with urllib.request.urlopen(req, timeout=180) as resp:

Network accessline 442

try:         with urllib.request.urlopen(req, timeout=180) as resp:             raw = resp.read()

Network accessline 443

with urllib.request.urlopen(req, timeout=180) as resp:

Network accessline 989

t0 = time.perf_counter()         try:             upstream_req = urllib.request.Request(

Network accessline 990

try:             upstream_req = urllib.request.Request(                 url, data=json.dumps(payload).encode("utf-8"),

Network accessline 991
```
upstream_req = urllib.request.Request(
```

Network accessline 993

headers=headers, method="POST",             )             upstream = urllib.request.urlopen(upstream_req, timeout=300)

Network accessline 994

)             upstream = urllib.request.urlopen(upstream_req, timeout=300)         except urllib.error.HTTPError as e:

Network accessline 995

upstream = urllib.request.urlopen(upstream_req, timeout=300)

Possible obfuscationline 1096

png_bytes = base64.b64decode(normalize_image_b64(image_b64))

Possible obfuscationline 1140

png_bytes = base64.b64decode(normalize_image_b64(image_b64))

Possible obfuscationline 1255

data = base64.b64decode(b64, validate=True)

Access to secretsline 1284
```
[Credential-related source omitted]
```

Network accessline 1297

BOOK_TMP.mkdir(parents=True, exist_ok=True)     httpd = ThreadingHTTPServer((args.host, args.port), Handler)     url = f"http://{args.host}:{args.port}/"

Network accessline 1298

httpd = ThreadingHTTPServer((args.host, args.port), Handler)     url = f"http://{args.host}:{args.port}/"     print(f"Xiaomi-OCR-0 Demo  →  {url}")

Network accessline 1299

url = f"http://{args.host}:{args.port}/"

`demo/static/app.js`

Executable scriptline 1
```
Executable file
```

Network accessline 934

let res;   try {     res = await fetch("/api/ocr/stream", {

Network accessline 935

try {     res = await fetch("/api/ocr/stream", {       method: "POST",

Network accessline 936
```
res = await fetch("/api/ocr/stream", {
```

Network accessline 1008

const timer = setTimeout(() => ac.abort(), 120000);   try {     const res = await fetch("/api/parse/regions", {

Network accessline 1009

try {     const res = await fetch("/api/parse/regions", {       method: "POST",

Network accessline 1010

const res = await fetch("/api/parse/regions", {

Network accessline 1035

const item = itemById(state.activeId);   if (!item || !item.image) throw new Error("无可用图像");   const res = await fetch(item.image);

Network accessline 1036

if (!item || !item.image) throw new Error("无可用图像");   const res = await fetch(item.image);   if (!res.ok) throw new Error("图像加载失败");

Network accessline 1037
```
const res = await fetch(item.image);
```

Network accessline 1048

const c = cfg();   const pdfData = await fileToDataUrl(file);   const start = await fetch("/api/book", {

Network accessline 1049

const pdfData = await fileToDataUrl(file);   const start = await fetch("/api/book", {     method: "POST", headers: { "Content-Type": "application/json" },

Network accessline 1050

const start = await fetch("/api/book", {

Network accessline 1059

while (true) {     await new Promise(function (resolve) { setTimeout(resolve, 800); });     const res = await fetch("/api/book/" + encodeURIComponent(started.id));

Network accessline 1060

await new Promise(function (resolve) { setTimeout(resolve, 800); });     const res = await fetch("/api/book/" + encodeURIComponent(started.id));     status = await res.json();

Network accessline 1061

const res = await fetch("/api/book/" + encodeURIComponent(started.id));

Network accessline 1073

if (status.status !== "done" && !status.has_markdown) throw new Error(status.error || "PDF 解析未完成");   $("taskProgressFill").style.width = "100%";   const rawRes = await fetch("/api/book/" + encodeURIComponent(started.id) + "/raw");

Network accessline 1074

$("taskProgressFill").style.width = "100%";   const rawRes = await fetch("/api/book/" + encodeURIComponent(started.id) + "/raw");   const raw = await rawRes.json();

Network accessline 1075

const rawRes = await fetch("/api/book/" + encodeURIComponent(started.id) + "/raw");

Network accessline 1206

function loadShowcase() {   return fetch("/showcase.json")

Network accessline 1207
```
return fetch("/showcase.json")
```

Network accessline 1302

async function checkHealth() {   try {     const res = await fetch("/api/health");

Network accessline 1303

try {     const res = await fetch("/api/health");     if (!res.ok) throw new Error("服务检查失败");

Network accessline 1304
```
const res = await fetch("/api/health");
```

`demo/static/i18n.js`

Executable scriptline 1
```
Executable file
```

`demo/static/img/xiaomi-logo.svg`

Network accessline 1

<?xml version="1.0" encoding="utf-8"?> <!-- Generator: Adobe Illustrator 16.0.0, SVG Export Plug-In . SVG Version: 6.00 Build 0)  --> <!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">

Network accessline 2

<!-- Generator: Adobe Illustrator 16.0.0, SVG Export Plug-In . SVG Version: 6.00 Build 0)  --> <!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd"> <svg version="1.1" id="Layer_1" xmlns="http://

Network accessline 3

<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">

Network accessline 4

<svg version="1.1" id="Layer_1" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" x="0px" y="0px"

`demo/static/index.html`

Network accessline 16

<div class="top-right">       <div class="conn" role="status"><span class="dot off" id="connDot"></span><span id="statusText">检查服务…</span></div>       <a class="source-link" href="https://github.com/SeerRay-Lab/Xiaomi-OCR-0" target="_blank"

Network accessline 17

<div class="conn" role="status"><span class="dot off" id="connDot"></span><span id="statusText">检查服务…</span></div>       <a class="source-link" href="https://github.com/SeerRay-Lab/Xiaomi-OCR-0" target="_blank" rel="noreferrer">GitHub ↗</a>

Network accessline 18

<a class="source-link" href="https://github.com/SeerRay-Lab/Xiaomi-OCR-0" target="_blank" rel="noreferrer">GitHub ↗</a>

Network accessline 19

<a class="source-link hf-link" href="https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0" target="_blank" rel="noreferrer" aria-label="Hugging Face 模型主页"><span class="hf-full">Hugging Face</span><span class="hf-short">HF</span> ↗</a>

Network accessline 42

<label class="task-input" id="kieFieldsWrap" hidden><span>想提取哪些字段？<small>逗号分隔，留空提取主要信息</small></span><input id="kieFields" type="text" placeholder="例如：日期、商户、总额、税额" /></label>         <label class="task-input" id="vqaQuestionWrap" hidden><sp

Network accessline 43

<label class="task-input" id="vqaQuestionWrap" hidden><span>关于这张图，你想了解什么？</span><input id="vqaQuestion" type="text" placeholder="例如：文档中的总金额是多少？" /></label>         <div class="task-note" id="regionPromptNote" hidden><span>区域模式适合印刷文档，将自动检测版面

Network accessline 44

<div class="task-note" id="regionPromptNote" hidden><span>区域模式适合印刷文档，将自动检测版面并分别识别文字、表格和公式。</span> <span id="regionDependencyStatus" role="status">正在检查区域解析依赖…</span><span id="regionInstallHint" hidden> 需要安装 <a href="https://paddlepaddle.gith

Network accessline 63

<section class="community" aria-labelledby="communityTitle">       <div class="community-copy"><span class="eyebrow">BUILT FOR EVERYONE</span><h2 id="communityTitle">科技不是高高在上，<br />而是服务于每一个人。</h2><p>如果它帮到了你，欢迎点亮 GitHub 的 Star 和 Hugging Face

Network accessline 64

<div class="community-copy"><span class="eyebrow">BUILT FOR EVERYONE</span><h2 id="communityTitle">科技不是高高在上，<br />而是服务于每一个人。</h2><p>如果它帮到了你，欢迎点亮 GitHub 的 Star 和 Hugging Face 的 Like。<br />如果还不够好，也请留下 Issue，让我们持续改进。</p></div>       <div class

Network accessline 65

<div class="community-actions"><div class="support-links"><a href="https://github.com/SeerRay-Lab/Xiaomi-OCR-0" target="_blank" rel="noreferrer"><span aria-hidden="true">☆</span> GitHub Star <span aria-hidden="true">↗</span></a><a href="htt

`demo/static/katex/katex.min.js`

Possible obfuscationline 1
```
[Credential-related source omitted]
```
Executable scriptline 1
```
[Credential-related source omitted]
```

`evaluation_pipeline/README.md`

Network accessline 12

The evaluation environment follows the [Environment Setup and Running instructions in OmniDocBench](https://github.com/opendatalab/OmniDocBench/blob/main/README.md#environment-setup-and-running). The main components are:

`evaluation_pipeline/README_zh.md`

Network accessline 12

评测环境参考 [OmniDocBench 的 Environment Setup and Running 说明](https://github.com/opendatalab/OmniDocBench/blob/main/README.md#environment-setup-and-running)，主要组件如下：

`evaluation_pipeline/init.py`

Executable scriptline 1
```
Executable file
```

`evaluation_pipeline/cropping.py`

Executable scriptline 1
```
Executable file
```

`evaluation_pipeline/e2e_img2md.py`

Executable scriptline 1
```
#!/usr/bin/env python3
```

Network accessline 173

url = f"http://127.0.0.1:{port}/v1/models"     no_proxy = urllib.request.ProxyHandler({})

Network accessline 179

return False, int(time.time() - start)         try:             req = urllib.request.Request(url, method="GET")

Network accessline 180

try:             req = urllib.request.Request(url, method="GET")             with opener.open(req, timeout=5) as resp:

Network accessline 181

req = urllib.request.Request(url, method="GET")

Access to secretsline 227
```
[Credential-related source omitted]
```
Network accessline 727
```
[Credential-related source omitted]
```
Network accessline 728
```
[Credential-related source omitted]
```
Access to secretsline 728
```
[Credential-related source omitted]
```

`evaluation_pipeline/layout_detect_paddlex.py`

Executable scriptline 1
```
#!/usr/bin/env python3
```
Access to secretsline 165
```
[Credential-related source omitted]
```

`evaluation_pipeline/twostage_img2md.py`

Executable scriptline 1
```
#!/usr/bin/env python3
```

Network accessline 439

def wait_for_service(port, proc, timeout=1200, interval=5):     url = f"http://127.0.0.1:{port}/v1/models"     no_proxy = urllib.request.ProxyHandler({})

Network accessline 440

url = f"http://127.0.0.1:{port}/v1/models"     no_proxy = urllib.request.ProxyHandler({})     opener = urllib.request.build_opener(no_proxy)

Network accessline 446

return False, int(time.time() - start)         try:             req = urllib.request.Request(url, method="GET")

Network accessline 447

try:             req = urllib.request.Request(url, method="GET")             with opener.open(req, timeout=5) as resp:

Network accessline 448

req = urllib.request.Request(url, method="GET")

Access to secretsline 479
```
[Credential-related source omitted]
```
Network accessline 766
```
[Credential-related source omitted]
```
Network accessline 767
```
[Credential-related source omitted]
```
Network accessline 768
```
[Credential-related source omitted]
```
Access to secretsline 768
```
[Credential-related source omitted]
```

`example_pics/receipt.md`

Dangerous shell operationline 25
```
Total Incl. GST@6% RM 5.90
```
Dangerous shell operationline 27
```
CASH RM 5.90
```

`examples/mcp-config.json`

Network accessline 7

],       "env": {         "XIAOMI_OCR_LOCAL_URL": "http://127.0.0.1:8000/v1",

Network accessline 8

"env": {         "XIAOMI_OCR_LOCAL_URL": "http://127.0.0.1:8000/v1",         "XIAOMI_OCR_MODEL": "SeerRay-Lab/Xiaomi-OCR-0"

Network accessline 9

"XIAOMI_OCR_LOCAL_URL": "http://127.0.0.1:8000/v1",

`postprocess/init.py`

Executable scriptline 1
```
Executable file
```

`postprocess/assemble_markdown.py`

Executable scriptline 1
```
#!/usr/bin/env python3
```

`postprocess/document.py`

Executable scriptline 1
```
Executable file
```

`postprocess/markdown_merge.py`

Executable scriptline 1
```
#!/usr/bin/env python3
```

`postprocess/otsl.py`

Executable scriptline 1
```
Executable file
```

`postprocess/region.py`

Executable scriptline 1
```
Executable file
```

`postprocess/repetition_guard.py`

Executable scriptline 1
```
#!/usr/bin/env python3
```

`skills/xiaomi-ocr/SKILL.md`

Network accessline 45

| Service | `XIAOMI_OCR_BACKEND` | `XIAOMI_OCR_LOCAL_URL` (default) | `XIAOMI_OCR_MODEL` (default) | | --- | --- | --- | --- | | SGLang / vLLM | `openai` | `http://127.0.0.1:8000/v1` | `SeerRay-Lab/Xiaomi-OCR-0` |

Network accessline 46

| --- | --- | --- | --- | | SGLang / vLLM | `openai` | `http://127.0.0.1:8000/v1` | `SeerRay-Lab/Xiaomi-OCR-0` | | Ollama | `ollama` | `http://127.0.0.1:11434/v1` | `longwayxu/xiaomi-ocr-0:bf16` |

Network accessline 47

| SGLang / vLLM | `openai` | `http://127.0.0.1:8000/v1` | `SeerRay-Lab/Xiaomi-OCR-0` |

Network accessline 48

| Ollama | `ollama` | `http://127.0.0.1:11434/v1` | `longwayxu/xiaomi-ocr-0:bf16` |

Network accessline 101

Check the OS, Python version, CPU/GPU support and available GPU memory against the official    [PaddlePaddle installation guide](https://paddlepaddle.github.io/PaddleX/latest/en/installation/paddlepaddle_install.html)

Network accessline 102

[PaddlePaddle installation guide](https://paddlepaddle.github.io/PaddleX/latest/en/installation/paddlepaddle_install.html)

Network accessline 103

and [PaddleX installation guide](https://paddlepaddle.github.io/PaddleX/latest/en/installation/installation.html).

Network accessline 111

Load PP-DocLayoutV3 locally (the first use may download its    [weights](https://huggingface.co/PaddlePaddle/PP-DocLayoutV3)), then verify one real

Network accessline 112

[weights](https://huggingface.co/PaddlePaddle/PP-DocLayoutV3)), then verify one real

Network accessline 118

Official runtime references: [SGLang Qwen3.5 guide](https://github.com/sgl-project/sglang/blob/main/docs_new/cookbook/autoregressive/Qwen/Qwen3.5.mdx) ·

Network accessline 119

[SGLang Qwen3.5 guide](https://github.com/sgl-project/sglang/blob/main/docs_new/cookbook/autoregressive/Qwen/Qwen3.5.mdx) ·

Network accessline 120

[vLLM Qwen3.5 model implementation](https://docs.vllm.ai/en/latest/api/vllm/model_executor/models/qwen3_5/) ·

Network accessline 121

[vLLM supported models](https://docs.vllm.ai/en/latest/models/supported_models/).

`skills/xiaomi-ocr/install.sh`

Executable scriptline 1
```
#!/usr/bin/env bash
```

Network accessline 5

SKILL_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" LOCAL_URL="${XIAOMI_OCR_LOCAL_URL:-http://127.0.0.1:8000/v1}"

Network accessline 6

LOCAL_URL="${XIAOMI_OCR_LOCAL_URL:-http://127.0.0.1:8000/v1}"

Network accessline 28

if [[ "$LOCAL_URL" != http://127.0.0.1:* && "$LOCAL_URL" != http://localhost:* && "$LOCAL_URL" != http://\[::1\]:* ]]; then

Network accessline 52

fi export XIAOMI_OCR_REPO="$REPO_DIR" "$PY" - "$SKILL_DIR/mcp-config.json" "$PY" "$SKILL_DIR/mcp_ocr_server.py" "$LOCAL_URL" "$MODEL" "$REPO_DIR" <<'PYCONFIG'

Network accessline 53

export XIAOMI_OCR_REPO="$REPO_DIR" "$PY" - "$SKILL_DIR/mcp-config.json" "$PY" "$SKILL_DIR/mcp_ocr_server.py" "$LOCAL_URL" "$MODEL" "$REPO_DIR" <<'PYCONFIG' import json, sys

Network accessline 54

"$PY" - "$SKILL_DIR/mcp-config.json" "$PY" "$SKILL_DIR/mcp_ocr_server.py" "$LOCAL_URL" "$MODEL" "$REPO_DIR" <<'PYCONFIG'

`skills/xiaomi-ocr/mcp_ocr_server.py`

Executable scriptline 1
```
#!/usr/bin/env python3
```
Access to secretsline 26
```
[Credential-related source omitted]
```
Network accessline 32
```
[Credential-related source omitted]
```
Access to secretsline 32
```
[Credential-related source omitted]
```
Access to secretsline 33
```
[Credential-related source omitted]
```
Access to secretsline 34
```
[Credential-related source omitted]
```
Access to secretsline 35
```
[Credential-related source omitted]
```
Access to secretsline 36
```
[Credential-related source omitted]
```

Network accessline 64

parsed = urllib.parse.urlparse(base)     if parsed.scheme != "http" or parsed.hostname not in {"127.0.0.1", "localhost", "::1"}:         raise RuntimeError("XIAOMI_OCR_LOCAL_URL must point to a local http://127.0.0.1 inference server")

Network accessline 65

if parsed.scheme != "http" or parsed.hostname not in {"127.0.0.1", "localhost", "::1"}:         raise RuntimeError("XIAOMI_OCR_LOCAL_URL must point to a local http://127.0.0.1 inference server")     return base

Network accessline 66

raise RuntimeError("XIAOMI_OCR_LOCAL_URL must point to a local http://127.0.0.1 inference server")

Network accessline 70

def _http_json(url: str, payload: dict, timeout: int = 300) -> dict:     req = urllib.request.Request(url, data=json.dumps(payload).encode(),

Network accessline 71

req = urllib.request.Request(url, data=json.dumps(payload).encode(),

Network accessline 72

headers={"Content-Type": "application/json"}, method="POST")     try:         with urllib.request.urlopen(req, timeout=timeout) as resp:

Network accessline 73

try:         with urllib.request.urlopen(req, timeout=timeout) as resp:             return json.loads(resp.read())

Network accessline 74

with urllib.request.urlopen(req, timeout=timeout) as resp:

Network accessline 115

return p.read_bytes()     if url:         with urllib.request.urlopen(url, timeout=30) as response:

Network accessline 116

if url:         with urllib.request.urlopen(url, timeout=30) as response:             return response.read()

Network accessline 117

with urllib.request.urlopen(url, timeout=30) as response:

Possible obfuscationline 121

raw = base64.b64decode(b64.split(",", 1)[-1], validate=True)

Access to secretsline 155
```
[Credential-related source omitted]
```
Access to secretsline 156
```
[Credential-related source omitted]
```
Access to secretsline 157
```
[Credential-related source omitted]
```
Access to secretsline 158
```
[Credential-related source omitted]
```

`skills/xiaomi-ocr/scripts/otsl_convert.py`

Executable scriptline 1
```
#!/usr/bin/env python3
```

`skills/xiaomi-ocr/scripts/register_codex_mcp.py`

Executable scriptline 1
```
#!/usr/bin/env python3
```
Access to secretsline 40
```
[Credential-related source omitted]
```
Access to secretsline 43
```
[Credential-related source omitted]
```
Network accessline 47
```
[Credential-related source omitted]
```
Access to secretsline 47
```
[Credential-related source omitted]
```
Network accessline 48
```
[Credential-related source omitted]
```
Access to secretsline 48
```
[Credential-related source omitted]
```

Network accessline 49

if not endpoint.startswith(("http://127.0.0.1:", "http://localhost:", "http://[::1]:")):

Access to secretsline 55
```
[Credential-related source omitted]
```

`skills/xiaomi-ocr/scripts/selftest.py`

Executable scriptline 1
```
#!/usr/bin/env python3
```

How the signals fit together

Facts from a static read of the files at the pinned commit. They describe what the code does, not why.

Some files could not be read; the facts below still hold.

Signal confidence: 257 facts

Network access in prose of documentationINSTALL.md:29Low confidence
Network access in prose of documentationINSTALL.md:30Low confidence
Network access in prose of documentationINSTALL.md:31Low confidence
Network access in prose of documentationINSTALL.md:53Low confidence
Network access in prose of documentationINSTALL.md:62Low confidence
Network access in code of documentationINSTALL.md:73Low confidence
Network access in code of documentationINSTALL.md:81Low confidence
Network access in prose of documentationINSTALL.md:92Low confidence
Network access in prose of documentationINSTALL.md:94Low confidence
Network access in prose of documentationINSTALL.md:103Low confidence
Network access in prose of documentationINSTALL.md:104Low confidence
Network access in prose of documentationINSTALL.md:105Low confidence
Network access in code of documentationOLLAMA.md:41Low confidence
Network access in code of documentationOLLAMA.md:41Low confidence
Network access in code of documentationOLLAMA.md:42Low confidence
Network access in code of documentationOLLAMA.md:42Low confidence
Network access in code of documentationOLLAMA.md:43Low confidence
Network access in code of documentationOLLAMA.md:43Low confidence
Network access in code of documentationOLLAMA.md:59Low confidence
Network access in code of documentationOLLAMA.md:59Low confidence
Network access in code of documentationOLLAMA.md:60Low confidence
Network access in code of documentationOLLAMA.md:60Low confidence
Network access in prose of documentationREADME.md:5Low confidence
Network access in prose of documentationREADME.md:6Low confidence
Network access in prose of documentationREADME.md:7Low confidence
Network access in prose of documentationREADME.md:8Low confidence
Network access in prose of documentationREADME.md:9Low confidence
Network access in prose of documentationREADME.md:10Low confidence
Network access in prose of documentationREADME.md:11Low confidence
Network access in prose of documentationREADME.md:12Low confidence
Network access in prose of documentationREADME.md:32Low confidence
Network access in prose of documentationREADME.md:52Low confidence
Network access in prose of documentationREADME.md:53Low confidence
Network access in prose of documentationREADME.md:54Low confidence
Network access in prose of documentationREADME.md:55Low confidence
Network access in prose of documentationREADME.md:56Low confidence
Network access in code of documentationREADME.md:73Low confidence
Network access in code of documentationREADME.md:90Low confidence
Network access in prose of documentationREADME.md:96Low confidence
Network access in prose of documentationREADME.md:102Low confidence
Network access in prose of documentationREADME.md:103Low confidence
Network access in prose of documentationREADME.md:138Low confidence
Network access in prose of documentationREADME.md:142Low confidence
Network access in code of documentationREADME.md:156Low confidence
Network access in code of documentationREADME.md:157Low confidence
Network access in code of documentationREADME.md:158Low confidence
Network access in prose of documentationREADME.md:162Low confidence
Network access in prose of documentationREADME.md:168Low confidence
Network access in prose of documentationREADME.md:169Low confidence
Network access in prose of documentationREADME.md:170Low confidence

and 207 more

Works with

Agent compatibility, from the files in this repository
Agent	Status	Evidence	Install to
Claude Code	Compatible	Standard SKILL.md SKILL.md	`.claude/skills/<skill-name>/`
Codex	Compatible	Standard SKILL.md SKILL.md	`.agents/skills/<skill-name>/`
Cursor	Compatible	Standard SKILL.md SKILL.md	`.cursor/skills/<skill-name>/`
Gemini CLI	Compatible	Standard SKILL.md SKILL.md	`.gemini/skills/<skill-name>/`
GitHub Copilot	Compatible	Standard SKILL.md SKILL.md	`.github/skills/<skill-name>/`
OpenCode	Compatible	Standard SKILL.md SKILL.md	`.opencode/skills/<skill-name>/`

Native The repository ships a manifest, folder or frontmatter field for this agent.
Compatible A standard SKILL.md that this agent can read.
Unknown No evidence either way.

Derived automatically from paths, manifests and frontmatter. Not tested on each agent.

Explore alternatives and comparisons

Add this badge to your README

It updates every night and links to this page.

[![xiaomi-research/Xiaomi-OCR-0 on hotskills](https://hotskills.dev/badge/xiaomi-research-xiaomi-ocr-0.svg)](https://hotskills.dev/skill/xiaomi-research-xiaomi-ocr-0/)

xiaomi-research/Xiaomi-OCR-0

History

SKILL.md

skills/xiaomi-ocr/SKILL.md

Share this page

Heat

Stars over time

Individual skills in this repository

How it works

xiaomi-ocr-setup

Frontmatter

SKILL.md

Reference documents

The playbook

Scripts executed separately

Files

xiaomi-ocr

Frontmatter

SKILL.md

Reference documents

The playbook

Scripts executed separately

Files

Skills included

Repository facts

Rank over the last 12 weeks

Before you install

INSTALL.md

LICENSE

OLLAMA.md

README.md

README_zh.md

SKILL.md

THIRD_PARTY_NOTICES.md

demo/README.md

demo/server.py

demo/static/app.js

demo/static/i18n.js

demo/static/img/xiaomi-logo.svg

demo/static/index.html

demo/static/katex/katex.min.js

evaluation_pipeline/README.md

evaluation_pipeline/README_zh.md

evaluation_pipeline/__init__.py

evaluation_pipeline/cropping.py

evaluation_pipeline/e2e_img2md.py

evaluation_pipeline/layout_detect_paddlex.py

evaluation_pipeline/twostage_img2md.py

example_pics/receipt.md

examples/mcp-config.json

postprocess/__init__.py

postprocess/assemble_markdown.py

postprocess/document.py

postprocess/markdown_merge.py

postprocess/otsl.py

postprocess/region.py

postprocess/repetition_guard.py

skills/xiaomi-ocr/SKILL.md

skills/xiaomi-ocr/install.sh

skills/xiaomi-ocr/mcp_ocr_server.py

skills/xiaomi-ocr/scripts/otsl_convert.py

skills/xiaomi-ocr/scripts/register_codex_mcp.py

skills/xiaomi-ocr/scripts/selftest.py

How the signals fit together

Works with

You might also use

mvanhorn/last30days-skill

K-Dense-AI/scientific-agent-skills

wanshuiyin/Auto-claude-code-research-in-sleep

Imbad0202/academic-research-skills-codex

HKUSTDial/Supervisor-Skills

Master-cai/Research-Paper-Writing-Skills

Explore alternatives and comparisons

Add this badge to your README

`SKILL.md`

`skills/xiaomi-ocr/SKILL.md`

`INSTALL.md`

`LICENSE`

`OLLAMA.md`

`README.md`

`README_zh.md`

`SKILL.md`

`THIRD_PARTY_NOTICES.md`

`demo/README.md`

`demo/server.py`

`demo/static/app.js`

`demo/static/i18n.js`

`demo/static/img/xiaomi-logo.svg`

`demo/static/index.html`

`demo/static/katex/katex.min.js`

`evaluation_pipeline/README.md`

`evaluation_pipeline/README_zh.md`

`evaluation_pipeline/init.py`

`evaluation_pipeline/cropping.py`

`evaluation_pipeline/e2e_img2md.py`

`evaluation_pipeline/layout_detect_paddlex.py`

`evaluation_pipeline/twostage_img2md.py`

`example_pics/receipt.md`

`examples/mcp-config.json`

`postprocess/init.py`

`postprocess/assemble_markdown.py`

`postprocess/document.py`

`postprocess/markdown_merge.py`

`postprocess/otsl.py`

`postprocess/region.py`

`postprocess/repetition_guard.py`

`skills/xiaomi-ocr/SKILL.md`

`skills/xiaomi-ocr/install.sh`

`skills/xiaomi-ocr/mcp_ocr_server.py`

`skills/xiaomi-ocr/scripts/otsl_convert.py`

`skills/xiaomi-ocr/scripts/register_codex_mcp.py`

`skills/xiaomi-ocr/scripts/selftest.py`