Streaming responses

Display tokens as they arrive using stream_text.

Need ultra-low latency output? Use stream_text - an async iterator that yields text deltas in real time.

1. Prepare the script

   stream.py

   import asyncio
   from ai_sdk import openai, stream_text
   
   model = openai("gpt-4.1-mini")
   
   async def main():
   result = stream_text(
   model=model,
   prompt="Write a short poem about the sea where each line rhymes.",
   on_chunk=lambda d: print(d, end="", flush=True),
   )
   
       # Alternatively await the full text:
       full = await result.text()
       print("\n---\nFull text:")
       print(full)
   
   asyncio.run(main())
   
   
2. Run it

   python stream.py
   

   You’ll see tokens appear immediately instead of buffering the full response.

stream_text accepts the same arguments as generate_text plus optional callbacks:

• on_chunk(delta) - each text delta
• on_error(exc) - exceptions while streaming
• on_finish(full_text) - once complete

stream_object is similar, but yields objects instead of text.

stream_text is similar, but yields objects instead of text.