OpenAI Streaming Chat Completions

Source: https://github.com/openai/openai-python/blob/main/README.md?plain=1, https://deepwiki.com/openai/openai-python/4.1.2-streaming-chat-completions

Purpose

Deliver Chat Completions responses incrementally via Server-Sent Events (SSE), enabling real-time display of partial results. Supports raw chunk iteration and a higher-level stream manager with semantic events, automatic state accumulation, and incremental JSON parsing for structured outputs.

When to Use

• Real-time chat UIs that display tokens as they arrive
• Long responses where full-completion latency is unacceptable
• Streaming tool calls where incremental fragments matter
• Structured output streaming with progressive JSON parsing
• Any scenario needing low time-to-first-token

Prefer .stream() (the stream manager) for automatic accumulation and semantic events. Use raw stream=True only when you need chunk-by-chunk control.

Basic Usage