feat: streaming LLM responses#84

Open

TumCucTom wants to merge 10 commits intoMiniMax-AI:mainfrom

TumCucTom:feat/streaming-llm

Commits on Apr 5, 2026

feat(llm): add streaming support for LLM responses

TumCucTom
and
claude
committed
feat(agent): streaming LLM output in agent loop with --no-stream fallback

TumCucTom
and
claude
committed
fix(llm): add missing await in LLMClient.generate_stream
TumCucTom
committed
fix(agent): await generate_stream coroutine before iterating
TumCucTom
committed
fix(llm): remove erroneous await in generate_stream wrapper
TumCucTom
committed
fix(anthropic): remove invalid stream=True from messages.stream() call
TumCucTom
committed
fix(llm): use getattr for stop_reason/finish_reason on raw streaming events
TumCucTom
committed
fix(anthropic): parse SDK streaming events correctly

TumCucTom
and
claude
committed
fix streaming: force unbuffered stdout + sys.stdout.write for live output

TumCucTom
and
claude
committed
fix streaming duplicate events in AnthropicClient

TumCucTom
and
claude
committed