Inference Overview — dat.ai Docs

Inference

LLM chat via OpenAI-compatible and Ollama-compatible endpoints.

Method	Path	Format
POST	`/v1/chat/completions`	OpenAI-compatible
POST	`/api/chat`	Ollama-compatible

Check the dashboard playground for currently available models. Example: qwen3:1.7b.

Endpoint	Non-streaming	Streaming
`/v1/chat/completions`	JSON	SSE (`text/event-stream`), ends with `data: [DONE]`
`/api/chat`	JSON	NDJSON (`application/x-ndjson`), one JSON object per line

Both endpoints accept a dat.ai-specific extension (not part of OpenAI/Ollama specs):


{
  "datai": {
    "tools": {
      "net": true,
      "fs": true
    },
    "tps": "fast"
  }
}

Field	Description
`datai.tools.net`	Enable HTTP/network tools on the node
`datai.tools.fs`	Enable filesystem tools on the node
`datai.tps`	Throughput preset: `fast` (default), `med`, or `slow`

Rules:

Tools cannot be used with streaming — stream: true + datai.tools returns 400
HTTP wait timeout is capped at 180 seconds