waveterm/aiprompts/openai-streaming-text.md

For just text streaming, you only need to handle these 3 core events:

Essential Events

1. response.created

{
  "type": "response.created",
  "response": {
    "id": "resp_abc123",
    "created_at": 1640995200,
    "model": "gpt-5"
  }
}

Purpose: Initialize response tracking (like Anthropic's message_start)

2. response.output_text.delta

{
  "type": "response.output_text.delta",
  "item_id": "msg_abc123",
  "delta": "Hello, how can I"
}

Purpose: Stream text chunks (like Anthropic's text_delta)

3. response.completed

{
  "type": "response.completed",
  "response": {
    "usage": {
      "input_tokens": 100,
      "output_tokens": 200
    }
  }
}

Purpose: Finalize response (like Anthropic's message_stop)

Optional but Recommended

4. error

{
  "type": "error",
  "code": "rate_limit_exceeded",
  "message": "Rate limit exceeded"
}

Purpose: Handle errors gracefully

---

That's it for basic text streaming! You can ignore all the response.output_item.added/done, tool calling, reasoning, and annotation events if you just want simple text responses.

Your Go implementation would be:

1. Parse SSE stream
2. Switch on event.type
3. Handle these 4 event types
4. Accumulate text from delta fields
5. Emit to your existing SSE handler

Much simpler than the full implementation.