imux · 龍imux
AI workbench for macOS

Streaming

Streaming reduces time-to-first-token for long agent answers. Clavue uses the OpenAI Chat Completions streaming protocol (SSE, data: chunks, [DONE]).

Enable streaming

Pass stream: true on /v1/chat/completions. Clients should handle partial delta.content and ignore empty chunks.

curl -N https://api.clavue.com/v1/chat/completions \
  -H "Authorization: Bearer $CLAVUE_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"model":"clavue-2.1-fast","stream":true,"messages":[{"role":"user","content":"List 3 git tips"}]}'

Errors mid-stream

Auth/quota errors usually fail before the first chunk with a JSON error body and requestId. If the stream drops mid-way, retry with the same prompt; include x-imux-request-id when reporting.

Product surfaces

imux Agent Chat, web Chat (/playground), and chat.clavue.com already stream by default for official models.

Next