DeepOCR

OpenAI-compatible API gateway for DeepSeek web chat with OCR/file upload support.

Features

• OpenAI-compatible API – /v1/chat/completions (streaming + non-streaming), /v1/models, /health
• File upload – images (JPG, PNG, GIF, WebP), documents (PDF, DOCX, XLSX, TXT, CSV) and more
• Vision support – upload images for OCR and visual understanding
• Multi-account – round-robin load balancing across multiple DeepSeek accounts
• Auto file processing – waits for files to be fully processed (PENDING → PARSING → SUCCESS)
• Auto OCR instruction – injects OCR prompt when files are attached
• PoW solving – built-in Proof-of-Work solver (browser-based Web Workers for speed)
• Token management – automatic login, refresh every 10 minutes

Requirements

• Python 3.10+
• Chrome/Chromium (installed automatically by cloakbrowser)

Quick Start

git clone git@github.com:truongsontung/deepocr.git
cd deepocr
python3 -m venv venv && source venv/bin/activate
pip install -r requirements.txt
cp .env.sample .env
# Edit .env with your DeepSeek accounts
python3 deepocr.py

Configuration

Edit .env file:

Variable	Default	Description
API_KEY	sk-deepocr-key	API key for authenticating requests
PORT	8080	Server port
DEEPSEEK_ACCOUNTS	–	Comma-separated email:password pairs

Example .env

API_KEY=sk-deepocr-key
PORT=8080
DEEPSEEK_ACCOUNTS=user1@gmail.com:pass1,user2@gmail.com:pass2

API Usage

Chat Completion (with image)

curl http://localhost:8080/v1/chat/completions \
  -H "Authorization: Bearer sk-deepocr-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-chat",
    "messages": [{
      "role": "user",
      "content": [
        {"type": "text", "text": "Describe this image"},
        {"type": "image_url", "image_url": {"url": "data:image/jpeg;base64,..."}}
      ]
    }]
  }'

Chat Completion (with file upload)

curl http://localhost:8080/v1/chat/completions \
  -H "Authorization: Bearer sk-deepocr-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-chat",
    "messages": [{"role": "user", "content": "Analyze this file"}],
    "files": [{"name": "document.pdf", "data": "<base64>"}]
  }'

Streaming

Add "stream": true to the request body.

List Models

curl http://localhost:8080/v1/models \
  -H "Authorization: Bearer sk-deepocr-key"

Health Check

curl http://localhost:8080/health

Supported Models

Model	Description
deepseek-chat / deepseek-v4-flash	Fast chat (default)
deepseek-v4-pro	Pro model
deepseek-reasoner / deepseek-r1	Reasoning model

Model aliases: gpt-4o → deepseek-v4-flash, gpt-4 → deepseek-v4-flash, o1 → deepseek-reasoner, etc.

Usage Policy

This project is intended for research and educational purposes only.

By using this software, you agree to:

• Use it solely for personal study, research, and learning about AI/API integration
• Not use it for commercial purposes, production services, or any revenue-generating activities
• Not abuse or spam the underlying services
• Comply with all applicable terms of service of DeepSeek and any other integrated platforms
• Respect rate limits and fair usage of the underlying platforms

The authors are not responsible for any misuse of this software. Use at your own risk.

Project Structure

deepocr/
├── deepocr.py          # Main gateway server
├── deepseek_client.py  # DeepSeek API client + browser worker
├── token_manager.py    # Multi-account token rotation
├── config.py           # Configuration loader
├── pow_solver.py       # PoW solver (Python fallback)
├── .env                # Environment variables (secrets)
├── .env.sample         # Example env file
├── .gitignore          # Git ignore rules
├── requirements.txt    # Python dependencies
└── README.md           # This file