# deepseek-flash-openrouter-gb10

An OpenAI-compatible chat endpoint for **DeepSeek V4.1 Flash** (OpenRouter model
`deepseek/deepseek-v4.1-flash`). It only provides the model: run opencode (or any
OpenAI-compatible client) on your own machine and point it here. Your files and tools
stay on your machine; only the model calls come to this capsule.

No GPU. The container's only egress is `openrouter.ai:443`, through the m64 gate,
which adds the OpenRouter key: the capsule never holds it.

## Routes

All paths are relative to the capsule's URL (e.g. `http://<tailnet-ip>:<port>`).

| Method | Path | What |
| --- | --- | --- |
| POST | `/v1/chat/completions` | OpenAI chat completions, streaming or not |
| GET | `/v1/models` | one model: `deepseek-v4.1-flash` |
| GET | `/status` | phase (`ready` when usable) |
| GET | `/agent` | this document plus live status |
| GET | `/human/` | landing page: opencode setup and a quick-test chat |

### POST /v1/chat/completions

Body: `application/json` only (415 otherwise). Accepted fields:

- `messages` (required), `tools`, `tool_choice`, `response_format`, `max_tokens`,
  `stream`, `capsule`;
- `model`, `user`, `stream_options` are accepted and ignored or passed on; `n` must be 1.
  The model is always DeepSeek V4.1 Flash, whatever `model` says.

Any other field (e.g. `temperature`, `top_p`) is refused with 400 naming the field.

`stream: true` returns Server-Sent Events (`data: {...}` chunks, ending with
`data: [DONE]`), relayed as OpenRouter sends them. OpenRouter may interleave SSE comment
lines (`: OPENROUTER PROCESSING`) while the model thinks; SSE clients ignore them.
Non-streaming responses are OpenRouter's JSON, plus `params_used`.

Errors from OpenRouter (rate limits, upstream failures) keep their status and body.

```
curl -s $URL/v1/chat/completions -H 'Content-Type: application/json' \
  -d '{"messages":[{"role":"user","content":"Say hi"}],"max_tokens":50}'

curl -N $URL/v1/chat/completions -H 'Content-Type: application/json' \
  -d '{"messages":[{"role":"user","content":"Count to 5"}],"stream":true}'
```

## opencode on your machine

`~/.config/opencode/opencode.json` (global) or `opencode.json` in the project root:

```json
{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "dsflash": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "DeepSeek V4.1 Flash (capsule)",
      "options": { "baseURL": "<capsule URL>/v1" },
      "models": { "deepseek-v4.1-flash": { "name": "DeepSeek V4.1 Flash", "tool_call": true } }
    }
  },
  "model": "dsflash/deepseek-v4.1-flash"
}
```

Or, on a machine with m64: `m64 deepseek-flash-openrouter-gb10 --ui opencode`
serves the capsule and starts opencode with a generated config pointing at it.

## Access

On the tailnet (the default bind) there is no token: anyone on the tailnet can use the
endpoint and so spend the OpenRouter key. With a token (`--host 0.0.0.0` etc.), send
`Authorization: Bearer <token>` (opencode: `"apiKey": "<token>"` in the provider options).