Streaming

Set stream: true and the response is a server-sent event stream in the OpenAI format, terminated by data: [DONE].

curl -N https://api.tiny-router.dev/v1/chat/completions \
  -H "Authorization: Bearer $TR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "~deepseek/deepseek-v4-flash-latest",
    "stream": true,
    "messages": [{"role": "user", "content": "Count to five."}]
  }'

What we do not do to your stream

• Bytes are forwarded as they arrive and flushed on every read. Nothing is buffered to re-encode it.
• Events are not reordered, merged or reformatted. Metering reads the same buffer that is being written to you.
• Upstream keep-alive comments are passed through, so a long first token does not look like a dead connection.

Disconnecting

Closing the connection cancels the upstream request, which is what stops the provider billing us and you. Settlement still runs, on a context detached from your request, so the tokens already generated are charged and the rest of the hold is released.