README.md

metadata

title: DeepSeek Proxy
emoji: 🚀
colorFrom: blue
colorTo: purple
sdk: docker
app_port: 7860

🚀 DeepSeek Proxy for Hugging Face

OpenAI-compatible proxy server for DeepSeek web chat — run it on Hugging Face Space for free, then connect VS Code/Cline for unlimited AI coding.

---

⚡ Quick Start (3 Steps)

Step 1: Create HF Space

1. Go to huggingface.co/spaces
2. Click "Create new Space"
3. Settings:
   • Space name: deepseek-proxy (or any name)
   • SDK: Docker
   • Visibility: Private ⚠️ (important for token security)
   • Hardware: CPU free
4. Click "Create Space"

Step 2: Set Your Token

1. In your Space → Settings tab → Secrets
2. Add a secret:
   • Name: DEEPSEEK_TOKEN
   • Value: Your DeepSeek browser token

How to get your token:

1. Open chat.deepseek.com in browser
2. Press F12 → Application → Cookies → chat.deepseek.com
3. Find the token starting with oxMs/... (usually in Authorization or auth-token)
4. Copy and paste into HF Secrets

⚠️ Token expires every ~7 days. Update it weekly.

Step 3: Deploy

Upload these files to your Space:

• app.py
• Dockerfile
• requirements.txt

Or use git:

git clone https://huggingface.co/spaces/YOURNAME/deepseek-proxy
cd deepseek-proxy
# copy files here
git add .
git commit -m "deploy proxy"
git push

Wait for the build (1-2 minutes), then check https://YOURNAME-deepseek-proxy.hf.space/health

---

🔌 VS Code / Cline Setup

1. Install Cline Extension

• VS Code → Extensions → Search "Cline" → Install

2. Configure Cline Settings

Open VS Code settings (Ctrl+,) and add:

{
  "cline.apiConfig": {
    "apiProvider": "openai-compatible",
    "openAiBaseUrl": "https://YOURNAME-deepseek-proxy.hf.space/v1",
    "openAiModelId": "deepseek-chat",
    "openAiApiKey": "hf-any-key"
  }
}

Or use the Cline Settings UI:

1. Click Cline icon in sidebar
2. Settings → API Provider: OpenAI Compatible
3. Base URL: https://YOURNAME-deepseek-proxy.hf.space/v1
4. Model ID: deepseek-chat
5. API Key: hf-any-key (any value works, auth is via your HF Space)

3. Test

Ask Cline: "What is 2+2?" — it should reply in ~5-10 seconds.

---

🧠 Available Models

Model	Use Case
deepseek-chat	Fast coding, general questions
deepseek-reasoner	Deep thinking, debugging, complex logic

---

🔧 Troubleshooting

"Token not configured" error

• Go to Space Settings → Secrets → make sure DEEPSEEK_TOKEN is set
• Rebuild the Space (Factory Rebuild in Settings)

"PoW rejected" or timeout

• DeepSeek may have changed their PoW — check proxy logs in Space
• Try getting a fresh token from browser

Space goes to sleep

• HF Spaces sleep after 48h of inactivity
• Use UptimeRobot to ping /health every 30 minutes
• Or use HF Space's "Permanent" upgrade ($5/month)

Cline timeout errors

• Increase Cline timeout: Cline Settings → Request Timeout → 120 seconds
• The first request is slow (~20s) because of PoW solving
• Subsequent requests are fast (PoW is cached for 5 minutes)

---

📡 API Endpoints

Endpoint	Description
GET /	Status page
GET /health	Health check (for UptimeRobot)
GET /v1/models	List available models
POST /v1/chat/completions	Chat completion (OpenAI format)

---

⚠️ Disclaimer

This uses DeepSeek's web chat via scraping. This is not an official API.

• Personal use only
• Token expires ~7 days
• DeepSeek may block this approach at any time
• Use at your own risk

---

🔗 Links

• DeepSeek Web: chat.deepseek.com
• Hugging Face: huggingface.co/spaces
• Cline Extension: cline.bot

---

Made for free AI coding. Enjoy! 🎉