Mistral Chat Completions And Streaming

Agent streams /v1/chat/completions with stream=true and max_tokens=256 for a long answer; the final SSE chunk reports finish_reason='length'.

Detect finish_reason='length' on the terminal chunk and treat the answer as truncated — surface a partial-completion to the caller or continue by appending the assistant text and re-requesting. Never present truncated output as complete.

Agent consumes the SSE stream from /v1/chat/completions and stops reading at the first empty data line instead of the terminal 'data: [DONE]' marker.

Parse the SSE stream until the explicit 'data: [DONE]' sentinel; accumulate delta.content across chunks and only finalize on the terminator. Do not treat keep-alive or blank lines as end-of-stream.

Operator sets stop=['\n\nUser:'] expecting generation to halt before the next turn marker; the returned text still contains the marker.

Configure stop sequences correctly and verify finish_reason='stop'; the stop string itself is not included in the output, so do not also strip it manually and corrupt the text. Confirm the marker matches exactly.