LLM Streaming: SSE for Claude, OpenAI & Gemini APIs

​

Streaming switches

​

Claude streaming

curl -N -X POST "https://llm.apixo.ai/v1/messages" \
  -H "x-api-key: $APIXO_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-opus-4-8",
    "stream": true,
    "max_tokens": 256,
    "messages": [{"role": "user", "content": "Write a short release note."}]
  }'

​

OpenAI Responses streaming

curl -N -X POST "https://llm.apixo.ai/v1/responses" \
  -H "Authorization: Bearer $APIXO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.6-sol",
    "stream": true,
    "input": "Write a short release note."
  }'

​

Gemini streaming

curl --no-buffer -X POST "https://llm.apixo.ai/v1beta/models/gemini-3.5-flash:streamGenerateContent?alt=sse" \
  -H "x-goog-api-key: $APIXO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [
      {
        "role": "user",
        "parts": [{"text": "Write a short release note."}]
      }
    ]
  }'

​

Troubleshooting

• Test non-streaming first if a third-party app fails to read streamed events.
• Confirm the app supports the same provider streaming format.
• Confirm your API key is available as APIXO_API_KEY.
• Retry transient 429, 502, and 504 errors with backoff.

​

Production streaming best practices

• Buffer management: Ensure your HTTP client and reverse proxy (such as Nginx or Cloudflare) disable response buffering so SSE events arrive with minimal latency.
• Client disconnection handling: Implement disconnection handlers (AbortController in JavaScript or the client’s cancellation mechanism) to stop reading the response and release local stream resources when end users disconnect.
• Backpressure handling: Consume and render stream chunks asynchronously to prevent blocking the UI event loop during high-throughput token bursts.

​

Related links

• Chat API Overview
• Authentication
• Supported LLM Models
• Errors and Retry Guidance