Free DeepSeek

DeepSeek Web Chat → OpenAI / Claude / Gemini Compatible API

---

Free DeepSeek converts DeepSeek Web chat capabilities into OpenAI, Claude, and Gemini compatible APIs. Built with a Go backend and a React WebUI admin panel (source in webui/, builds to static/admin).

Features

Feature	Description
OpenAI Compatible	GET /v1/models, POST /v1/chat/completions, POST /v1/responses, POST /v1/embeddings, POST /v1/files
Claude Compatible	GET /anthropic/v1/models, POST /anthropic/v1/messages
Gemini Compatible	POST /v1beta/models/{model}:generateContent, streamGenerateContent
Multi-Account Pool	Automatic token refresh, email & mobile login support
Concurrency Control	Per-account in-flight limits + wait queues
DeepSeek PoW	High-performance pure Go implementation
Tool Calling	Leak prevention, structured incremental output
Admin API	Config management, hot-reload settings, proxy management, batch testing, session cleanup, import/export
WebUI Dashboard	Single-page admin app with dark mode at /admin
Health Probes	GET /healthz (liveness), GET /readyz (readiness)

Supported Models

Model Type	Model ID	Thinking	Search
default	deepseek-v4-flash	On by default, controllable	No
default	deepseek-v4-flash-nothinking	Always off	No
expert	deepseek-v4-pro	On by default, controllable	No
expert	deepseek-v4-pro-nothinking	Always off	No
default	deepseek-v4-flash-search	On by default, controllable	Yes
default	deepseek-v4-flash-search-nothinking	Always off	Yes
expert	deepseek-v4-pro-search	On by default, controllable	Yes
expert	deepseek-v4-pro-search-nothinking	Always off	Yes
vision	deepseek-v4-vision	On by default, controllable	No
vision	deepseek-v4-vision-nothinking	Always off	No

Common aliases (e.g. gpt-4.1, gpt-5, o3, claude-*, gemini-*) are also supported and mapped automatically.

Quick Start

Option 1: Run from Source

Prerequisites: Go 1.24+, Node.js 20.19+ (for WebUI build only)

Install Go — download from go.dev/dl or use your package manager:

# Linux (amd64)
wget https://go.dev/dl/go1.24.3.linux-amd64.tar.gz
sudo tar -C /usr/local -xzf go1.24.3.linux-amd64.tar.gz
echo 'export PATH=$PATH:/usr/local/go/bin' >> ~/.bashrc && source ~/.bashrc
go version

# macOS
brew install go

# Windows — download the MSI from https://go.dev/dl/ and run the installer

Then clone, configure, and run:

git clone https://github.com/MrFadiAi/free-deepseek.git
cd free-deepseek
cp config.example.json config.json
# Edit config.json with your DeepSeek account credentials
go run ./cmd/ds2api

Default URL: http://127.0.0.1:5001

The server binds to 0.0.0.0:5001, so LAN devices can access it via your internal IP.

WebUI auto-build: On first launch, if static/admin doesn't exist, the server will automatically run npm ci and npm run build.

Option 2: Docker

cp .env.example .env
cp config.example.json config.json
# Edit .env (set DS2API_ADMIN_KEY) and config.json (add your accounts)
docker-compose up -d

Option 3: Vercel

1. Fork this repository
2. Import in Vercel
3. Set environment variables (DS2API_ADMIN_KEY, DS2API_CONFIG_JSON)
4. Deploy

Configuration

Copy config.example.json to config.json and fill in your details:

• keys / api_keys: Client access keys for API authentication
• accounts: DeepSeek accounts (email or mobile login)
• model_aliases: Custom model name mappings
• runtime: Concurrency, queue, and token refresh settings (hot-reloadable via admin)
• auto_delete.mode: Remote session cleanup after requests (none / single / all)
• current_input_file: Independent split strategy for long contexts
• thinking_injection: Thinking enhancement prompt injection (enabled by default)

Environment Variables

Variable	Description
DS2API_ADMIN_KEY	Admin panel access key (required)
DS2API_CONFIG_PATH	Path to config.json file
DS2API_CONFIG_JSON	Inline config JSON (or Base64-encoded)
PORT	Server port (default: 5001)

See config.example.json for the full schema.

Authentication

Mode	Description
Managed Account	Pass a key from config.keys via Bearer or x-api-key header — the server auto-selects an account
Direct Token	Pass a raw DeepSeek token not in config.keys — used directly

Optional header X-Ds2-Target-Account: target a specific account by email or mobile. Gemini routes also accept x-goog-api-key or ?key= / ?api_key= query parameters.

Concurrency Model

Per-account concurrency = DS2API_ACCOUNT_MAX_INFLIGHT (default: 2)
Recommended concurrency = accounts × per-account limit
Queue limit = DS2API_ACCOUNT_MAX_QUEUE (default: recommended concurrency)
429 threshold = in-flight + queue ≈ accounts × 4

• When in-flight slots are full, requests enter the wait queue (no immediate 429)
• 429 is only returned when the total capacity is exceeded
• GET /admin/queue/status returns real-time concurrency state

License

This project is licensed under the GNU Affero General Public License v3.0 — see LICENSE.