Issues

HTTP 402 maps to ErrorCategory.QUOTA, but QUOTA is non-permanent so built-in ServiceSwitcher failover never triggers

#5608

· mannyb223 opened

on Sep 2, 2026

Deepgram Flux TTS: Interrupt is sent without playback_offset, so SpeechInterrupted never carries text_spoken

#5605

· mannyb223 opened

on Sep 2, 2026

FunctionSchema has no strict field — OpenAI Responses tools always send strict: null

#5604

· mannyb223 opened

on Sep 2, 2026

ElevenLabsDialogueTTSService: socket closed with 1008 "voices can only be set in the first message", losing unspoken text

#5601

· Sayan-Rezolve opened

on Sep 2, 2026

Context summarisation drops a leading developer-role prompt (and ignores input_text parts)

#5595

· mannyb223 opened

on Sep 2, 2026

FastAPIWebsocketTransport: `write_audio_frame` skips `_write_audio_sleep` when the serializer buffers, so output drains 3.5x faster than playout

#5592

· dileep-chagam opened

on Sep 2, 2026

OpenAIResponsesLLMService (websocket): response.failed discards error.code/response id, and the session is never recovered — later turns inherit a possibly-bad connection

#5575

· mannyb223 opened

on Sep 1, 2026

Token-mode text transforms cannot match across token boundaries — bounded lookback (LiveKit-style) instead of full sentence aggregation

#5574

· mannyb223 opened

on Sep 1, 2026

TTSService speaking pause also holds upstream frames, delaying the next LLM inference until the current utterance finishes

#5572

· achalringg opened

on Sep 1, 2026

LLMContextSummarizer: messages_since_summary never resets and counts the injected summary message

#5504

· jamsea opened

on Aug 31, 2026

A @job handler that raises sends no ERROR response, so job() with default timeout=None waits forever

#5493

· scttbnsn opened

on Aug 29, 2026

SyncParallelPipeline delivers SystemFrames (incl. InputAudioRawFrame) twice and retains copies in branch queues

#5492

· scttbnsn opened

on Aug 29, 2026

Navigation Menu

Uh oh!

All issues

HTTP 402 maps to ErrorCategory.QUOTA, but QUOTA is non-permanent so built-in ServiceSwitcher failover never triggers

Deepgram Flux TTS: Interrupt is sent without playback_offset, so SpeechInterrupted never carries text_spoken

FunctionSchema has no strict field — OpenAI Responses tools always send strict: null

ElevenLabsDialogueTTSService: socket closed with 1008 "voices can only be set in the first message", losing unspoken text

Context summarisation drops a leading developer-role prompt (and ignores input_text parts)

FastAPIWebsocketTransport: `write_audio_frame` skips `_write_audio_sleep` when the serializer buffers, so output drains 3.5x faster than playout

OpenAIResponsesLLMService (websocket): response.failed discards error.code/response id, and the session is never recovered — later turns inherit a possibly-bad connection

Token-mode text transforms cannot match across token boundaries — bounded lookback (LiveKit-style) instead of full sentence aggregation

TTSService speaking pause also holds upstream frames, delaying the next LLM inference until the current utterance finishes

LLMContextSummarizer: messages_since_summary never resets and counts the injected summary message

A @job handler that raises sends no ERROR response, so job() with default timeout=None waits forever

SyncParallelPipeline delivers SystemFrames (incl. InputAudioRawFrame) twice and retains copies in branch queues

Uh oh!

Search results