Bug regarding MINIMAX API in Livekit #35
Description
Basic Information - Models Used
speech-02-hd
Basic Information - Scenario Description
minimax tts api not working with livekit infrastructure
Is this bug known and solvable?
- I have followed the GitHub READMEs for
Minimax-MCPandMinimax-MCP-JS. - I have checked the official Minimax documentation and existing GitHub issues，but found no solution.
Information about environment
OS: Windows
Tech Stack:
Python - Backend
Livekit - Agent Framework
Trace-ID in the request head
start_segment()
Description
Logs:
{"sentence": "Thank you for
calling RiseWell Therapy!"}
…
{"sentence": " This is
Harper. How can I assist you
today?"}
… ERROR livekit.agents …
raise RuntimeError(
RuntimeError: start_segment()
called before the previous segment
was ended
Observed behavior:
LLM generates multiple sentence segments concurrently.
Because segments arrive too close together, the framework detects overlap: a new start_segment() is called before the previous one is properly ended/flushed.
Result: Error triggers immediately on the first/second segment → no audio plays at all (not even the first sentence).