Skip to main content

Minimal stdio-first OCR MCP service powered by PaddleOCR

Project description

Local OCR MCP

基于 FastMCPPaddleOCR 的最小化本地 OCR MCP 服务,只保留 stdio -> ocr_recognize -> PaddleOCR -> 统一响应 envelope 这一条主路径,适合本地 IDE、Agent 和 MCP Client 直接接入。


1. 快速开始

环境要求

  • Python 3.10+
  • uv

1. 直接用 uvx 运行

uvx --from "local-ocr-mcp[paddleocr]" local-ocr-mcp

适合直接接入 MCP 客户端,不需要手动创建虚拟环境。

2. 本地开发运行

git clone <repository-url>
cd local-ocr-mcp
uv sync --extra dev --extra paddleocr
uv run python -m local_ocr_mcp

3. MCP 客户端配置示例

{
  "mcpServers": {
    "local-ocr-mcp": {
      "command": "uvx",
      "args": ["--from", "local-ocr-mcp[paddleocr]", "local-ocr-mcp"]
    }
  }
}

📖 仓库里也保留了一份同步后的配置样例 → mcp_config.json


2. 核心特性

特性 说明
单路径主链路 只保留 stdio -> ocr_recognize -> PaddleOCR,没有传输分支和引擎路由
最小工具面 对外只暴露 ocr_recognizeocr_health_check 两个工具
统一响应契约 所有工具统一返回 status / data / error / meta envelope
固定输入模型 当前只接受本地图片路径 image.path,不支持 URL、base64 或上传流
稳定错误语义 用固定错误码暴露常见失败路径,便于客户端编排与恢复
低认知负担 不保留多引擎、analysis、progress、heartbeat、registry/factory 等扩展层

3. 技术栈

  • Python 3.10+ / uv - 运行时与依赖管理
  • FastMCP - MCP 服务框架
  • PaddleOCR - 固定 OCR 引擎
  • Pillow - 图片合法性校验
  • pytest - 最小契约测试

4. 架构概览

┌──────────────────────────────────────────────────────┐
│                 MCP Client / IDE / Agent            │
├──────────────────────────────────────────────────────┤
│                     stdio transport                  │
├──────────────────────────────────────────────────────┤
│       FastMCP Tools: ocr_recognize / health_check   │
├──────────────────────────────────────────────────────┤
│  RecognitionService            HealthService         │
├──────────────────────────────────────────────────────┤
│                  PaddleOCREngine                    │
├──────────────────────────────────────────────────────┤
│          PaddleOCR / Filesystem / Image Validation   │
└──────────────────────────────────────────────────────┘

当前实现刻意保持单向依赖和短链路:

  • CLI 只负责启动 stdio
  • Tool 层只负责暴露 MCP 接口
  • Service 层负责校验输入并整形成统一 envelope
  • Engine 层只负责调用 PaddleOCR 并归一化结果

5. 工具契约

工具 说明
ocr_recognize 识别本地图片并返回统一 OCR 结果
ocr_health_check 返回轻量健康状态、运行时版本和启动时长

ocr_recognize 请求示例

{
  "image": {
    "path": "./example.png"
  }
}

ocr_recognize 输入约束

  • image 必须是对象
  • image.path 必须是非空字符串
  • 当前只允许 image.path
  • 相对路径会按服务进程当前工作目录解析为绝对路径

ocr_recognize 成功响应示例

{
  "status": "ok",
  "data": {
    "text": "示例文本",
    "boxes": [
      {"x1": 0.0, "y1": 0.0, "x2": 120.0, "y2": 32.0}
    ],
    "confidence": 0.98,
    "engine": "paddleocr",
    "processing_ms": 143
  },
  "error": null,
  "meta": {
    "timestamp": "2026-03-14T00:00:00+00:00",
    "runtime_version": "0.2.0",
    "request_id": "8c4c9d5c-6b4d-4f1f-b6df-4ff80e3ff6b6",
    "resolved_engine": "paddleocr",
    "resolved_image_path": "/abs/path/example.png"
  }
}

ocr_health_check 响应示例

{
  "status": "ok",
  "data": {
    "status": "healthy",
    "uptime_ms": 1234,
    "transport": "stdio",
    "runtime_version": "0.2.0"
  },
  "error": null,
  "meta": {
    "timestamp": "2026-03-14T00:00:00+00:00",
    "runtime_version": "0.2.0"
  }
}

6. 错误语义与运行约束

统一错误结构

{
  "status": "error",
  "data": null,
  "error": {
    "code": "file_not_found|invalid_image|engine_not_available|internal_error",
    "message": "具体错误信息",
    "retryable": false
  },
  "meta": {
    "timestamp": "...",
    "runtime_version": "0.2.0",
    "request_id": "...",
    "resolved_engine": "paddleocr"
  }
}

错误码说明

  • file_not_foundimage.path 指向的文件不存在
  • invalid_image:请求结构不合法,或文件存在但不是有效图片
  • engine_not_availablePaddleOCR 未安装或初始化失败
  • internal_error:识别链路内部出现未归类异常

当前运行约束

  • 只支持 stdio
  • 只支持 PaddleOCR
  • 不会自动切换引擎
  • 不支持 streamable-http
  • 不支持 analysis、progress、heartbeat
  • 不支持 URL、base64、上传流等非文件输入

PaddleOCR 兼容行为

内部固定使用 PaddleOCR,但对其 Python API 保留最小兼容:

  • 优先走较新的 predict()
  • 如果安装版本只支持旧接口,则回退到 ocr()
  • 对外继续返回统一响应结构

7. 配置与脚本

可选环境变量

变量 说明 默认值
PADDLEOCR_LANG PaddleOCR 初始化语言 ch
LOG_LEVEL 日志级别 INFO

示例:

PADDLEOCR_LANG=ch LOG_LEVEL=INFO uv run python -m local_ocr_mcp

辅助脚本

当前 scripts/ 目录只保留和最小实现一致的两个脚本:

  • uv run python scripts/list_tools.py
  • uv run python scripts/recognize_image.py path/to/image.png --json

📖 详细说明 → scripts/README.md


8. 开发与验证

常用命令:

uv run python -m pytest -q
uv run python -m local_ocr_mcp --help
uv run python scripts/list_tools.py
uv run python scripts/recognize_image.py path/to/image.png --json

CI 当前只覆盖最小必需验证:

  • 编译检查
  • 单元测试
  • CLI --help
  • stdio 启动烟测

9. 常见问题

1. 服务启动后为什么一直不退出?

因为它在等待 MCP 客户端通过 stdio 调用工具。这是正常行为。

2. 为什么返回 engine_not_available

通常说明 PaddleOCR 或底层依赖没有正确安装。优先重新同步依赖:

uv sync --extra paddleocr

如果你是通过 uvx 启动,确认命令里包含:

uvx --from local-ocr-mcp[paddleocr] local-ocr-mcp

3. 为什么返回 invalid_image

常见原因:

  • 请求里没有 image.path
  • image.path 不是字符串
  • 文件存在,但不是合法图片

4. 相对路径为什么找不到文件?

相对路径是相对于服务进程当前工作目录解析的。最稳妥的做法是直接传绝对路径。

5. 为什么不保留多引擎和 HTTP?

因为当前版本明确选择“核心最小化”路线,只保留本地 OCR 主链路,不为未来扩展提前保留复杂结构。


最后更新: 2026-03-14
License: MIT

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

local_ocr_mcp-0.2.0.tar.gz (10.9 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

local_ocr_mcp-0.2.0-py3-none-any.whl (15.2 kB view details)

Uploaded Python 3

File details

Details for the file local_ocr_mcp-0.2.0.tar.gz.

File metadata

  • Download URL: local_ocr_mcp-0.2.0.tar.gz
  • Upload date:
  • Size: 10.9 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.9.17 {"installer":{"name":"uv","version":"0.9.17","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Linux Mint","version":"22.2","id":"zara","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

File hashes

Hashes for local_ocr_mcp-0.2.0.tar.gz
Algorithm Hash digest
SHA256 cab40adbbb8eaa9907be10447e553eebd70f4ba7b46371a3662b4f3035912bb1
MD5 e404524654f5a6332f46258eaffdaf50
BLAKE2b-256 d5741256a0834ff31ccd98eaa525e09487554919f7e7a4f19aba60ddd27880ad

See more details on using hashes here.

File details

Details for the file local_ocr_mcp-0.2.0-py3-none-any.whl.

File metadata

  • Download URL: local_ocr_mcp-0.2.0-py3-none-any.whl
  • Upload date:
  • Size: 15.2 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.9.17 {"installer":{"name":"uv","version":"0.9.17","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Linux Mint","version":"22.2","id":"zara","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

File hashes

Hashes for local_ocr_mcp-0.2.0-py3-none-any.whl
Algorithm Hash digest
SHA256 f437ba5b516cd2f9ea4f83886360e38a3abb139ec79d3dd3be748cf88543129f
MD5 6b560ef0b4c33ae0acde5505e0e375dd
BLAKE2b-256 88059daa7c0264a06b446cb216759a45a9f67415fdfaf4b5733a9206fbae6189

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page