Documentation

Streaming in the NexoRouter documentation.

Streaming

Status: Not yet documented as a verified stable public feature.

Some OpenAI-compatible clients can request streaming responses. Do not rely on streaming in production until NexoRouter streaming behavior, usage logs, billing, and client parsing are tested end to end for the selected model.

When to avoid streaming

  • First-time setup.
  • Billing or quota debugging.
  • Tools where the non-streaming request already fails.
  • Environments with proxies that buffer responses.

First test

  1. Confirm non-streaming chat/completions works.
  2. Enable streaming in one client only.
  3. Use a small prompt.
  4. Check that the client receives incremental chunks.
  5. Confirm Usage Logs still show the request with cost and token details.

If it fails

Turn streaming off and verify the same model with a normal request. If the normal request works, treat the issue as client, proxy, or streaming-specific until proven otherwise.

Streaming — NexoRouter