BlazeInferDocs

Streaming

Receive tokens as Server-Sent Events as they are generated.

Set "stream": true to receive the reply incrementally over Server-Sent Events (SSE). Use streaming for any generation that may take a while, because a non-streaming request whose full answer takes longer than 4 minutes fails.

curl https://blaze.e2enetworks.net/v1/chat/completions \
  -H "Authorization: Bearer $BLAZE_API_KEY" \
  -H "Content-Type: application/json" \
  -N \
  -d '{
    "model": "qwen/qwen3.8-27b",
    "messages": [{"role": "user", "content": "Count from 1 to 10."}],
    "stream": true
  }'

The final usage chunk

BlazeInfer adds a final usage chunk only when the request has no stream_options. If you send stream_options yourself, set include_usage: true; otherwise no usage chunk is sent.

That chunk has an empty choices array, so make sure your code handles it (see the guard in the Python tab above).

finish_reason (for example stop or tool_calls) arrives in its own chunk with an empty delta. Read it to know why generation ended.

Reasoning models stream their reasoning in a separate delta field. The Python and TypeScript examples above print only content. See Reasoning models.

Handle truncated streams

If a stream fails after it has started, the HTTP response is already 200 and the stream simply stops. Treat a stream that ends without a finish_reason as a failed request and retry.

On this page