DWA

DeepSeek Web API

A single Go binary that exposes the OpenAI and Anthropic chat APIs on top of the DeepSeek web app.

_{Real capture: the live server answering real curl calls to both endpoints. Play / download the MP4.}

---

What it does

DWA talks to the unofficial chat.deepseek.com web API and re-exposes it through two standard interfaces:

• OpenAI - POST /v1/chat/completions
• Anthropic - POST /v1/messages

Point any OpenAI or Anthropic SDK at http://127.0.0.1:8000 and it just works. Streaming and non-streaming are both supported, and the proof-of-work challenge that the web app requires is solved locally with the original WASM module.

How it works

flowchart LR
    A[Your client<br/>OpenAI / Anthropic SDK] -->|HTTP| B[DWA server]
    B --> C[convert<br/>format translation]
    B --> D[pow<br/>WASM solver]
    B --> E[client<br/>DeepSeek HTTP + SSE]
    D -->|x-ds-pow-response| E
    E -->|streaming SSE| F[(chat.deepseek.com)]
    F -->|tokens| E
    E -->|chunks| B
    B -->|response| A

Loading

Each request fetches a POW challenge and a HIF token in parallel, solves the challenge through sha3_pow.wasm, then streams the completion back and converts it to the format the client asked for. Conversations reuse a cached DeepSeek session so multi-turn chats keep their context.

Quick start

Requirements: Go 1.22+ and a DeepSeek account.

# 1. Set your token
cp .env.example .env
#    then edit .env and paste your DEEPSEEK_TOKEN

# 2. Run
go run .

On Windows you can just double-click start.bat. It builds the binary if needed and launches the server.

 DWA - DeepSeek Web API
 OpenAI:    http://127.0.0.1:8000/v1/chat/completions
 Anthropic: http://127.0.0.1:8000/v1/messages
 Models:    http://127.0.0.1:8000/v1/models

Getting your token

The token is the Bearer value the web app sends, not a cookie.

1. Log in at chat.deepseek.com and send any message.
2. Open DevTools (F12) and go to the Network tab.
3. Find the request to /api/v0/chat/completion.
4. Under Request Headers, copy the value after Authorization: Bearer.
5. Paste it into .env as DEEPSEEK_TOKEN.

The token is tied to your session and expires over time. When you start seeing 40003 invalid token, grab a fresh one the same way.

Endpoints

Method	Path	Format	Notes
GET	/v1/models	-	Lists the available model IDs
POST	/v1/chat/completions	OpenAI	Streaming and non-streaming
POST	/v1/messages	Anthropic	Streaming and non-streaming

Response previews

OpenAI - non-streaming

curl http://127.0.0.1:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model":"deepseek-chat","messages":[{"role":"user","content":"Say hi in one word"}]}'

{
  "id": "chatcmpl-18b4adb1e8094f38",
  "object": "chat.completion",
  "created": 1780238735,
  "model": "deepseek-chat",
  "choices": [
    {
      "index": 0,
      "message": { "role": "assistant", "content": "Hi" },
      "finish_reason": "stop"
    }
  ],
  "usage": { "prompt_tokens": 0, "completion_tokens": 0, "total_tokens": 0 }
}

OpenAI - streaming ("stream": true)

data: {"choices":[{"delta":{"content":"1","role":"assistant"},"finish_reason":null,"index":0}],"model":"deepseek-chat","object":"chat.completion.chunk"}

data: {"choices":[{"delta":{"content":" ","role":"assistant"},"finish_reason":null,"index":0}],"model":"deepseek-chat","object":"chat.completion.chunk"}

data: {"choices":[{"delta":{"content":"2","role":"assistant"},"finish_reason":null,"index":0}],"model":"deepseek-chat","object":"chat.completion.chunk"}

data: [DONE]

Anthropic

curl http://127.0.0.1:8000/v1/messages \
  -H "Content-Type: application/json" \
  -d '{"model":"deepseek-chat","max_tokens":100,"messages":[{"role":"user","content":"Say hello in one word"}]}'

{
  "id": "msg_18b4adb4b6943064",
  "type": "message",
  "role": "assistant",
  "model": "deepseek-chat",
  "content": [{ "type": "text", "text": "Hello" }],
  "stop_reason": "end_turn",
  "stop_sequence": null,
  "usage": { "input_tokens": 1, "output_tokens": 1 }
}

Project structure

DWA/
├── main.go                 entry point: loads .env, parses flags, starts server
├── sha3_pow.wasm           the POW module from the DeepSeek web app
├── start.bat               Windows launcher (builds into bin/ then runs)
├── .env.example            token template
├── bin/                    compiled binary lands here (gitignored)
└── internal/
    ├── pow/                WASM proof-of-work solver (wazero)
    ├── client/             DeepSeek HTTP client, sessions, SSE streaming
    ├── convert/            OpenAI / Anthropic <-> DeepSeek translation
    └── server/             HTTP handlers and session cache

Configuration

Variable	Default	Description
DEEPSEEK_TOKEN	(required)	Bearer token from the web app
HOST	127.0.0.1	Bind address
PORT	8000	Bind port

Flags override the environment: go run . -host 0.0.0.0 -port 9000.

Notes and limitations

• This uses an unofficial API, so it can break whenever the web app changes.
• The token is a session token and expires. There is no automatic refresh.
• usage token counts are placeholders. The web API does not report them.
• Tool calls are passed through a small XML convention and parsed back out, so behaviour depends on the model following the format.

License

See LICENSE.