feat(realtime): stream the LLM / TTS / transcription pipeline stages#10176

Merged

richiejp merged 16 commits into

masterfrom

feat/realtime-streaming

Jun 11, 2026

Commits on Jun 10, 2026

feat(realtime): pipeline streaming + disable_thinking config

mudler
authored and
richiejp
committed
feat(realtime): sentence segmenter for streamed LLM->TTS pipelining

mudler
authored and
richiejp
committed
feat(realtime): streaming TTS/transcription methods on Model interface

mudler
authored and
richiejp
committed
feat(realtime): emitSpeech with flag-gated streaming TTS

mudler
authored and
richiejp
committed
feat(realtime): route response audio through emitSpeech (streaming TTS)

mudler
authored and
richiejp
committed
feat(realtime): streaming transcription text deltas

mudler
authored and
richiejp
committed
feat(realtime): pipeline disable_thinking maps to enable_thinking off

mudler
authored and
richiejp
committed
feat(realtime): speechStreamer for token-streamed LLM->TTS

mudler
authored and
richiejp
committed
feat(realtime): wire streamLLMResponse for token-streamed replies

mudler
authored and
richiejp
committed
docs(realtime): document pipeline streaming + disable_thinking

mudler
authored and
richiejp
committed
fix(realtime): register pipeline streaming/thinking config fields

mudler
authored and
richiejp
committed
fix(realtime): always strip reasoning from spoken output

mudler
authored and
richiejp
committed
fix(realtime): clean TTS temp path before read (gosec G304)

mudler
authored and
richiejp
committed
refactor(realtime): buffer whole message for TTS, drop sentence segmenter

mudler
authored and
richiejp
committed
feat(realtime): stream tool-call turns via tokenizer-template autoparser

mudler
authored and
richiejp
committed
feat(realtime): script-aware clause chunking + streamed-reply fixes
richiejp
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

feat(realtime): stream the LLM / TTS / transcription pipeline stages#10176

feat(realtime): stream the LLM / TTS / transcription pipeline stages#10176
richiejp merged 16 commits into
masterfrom
feat/realtime-streaming

Commits on Jun 10, 2026

feat(realtime): pipeline streaming + disable_thinking config

feat(realtime): sentence segmenter for streamed LLM->TTS pipelining

feat(realtime): streaming TTS/transcription methods on Model interface

feat(realtime): emitSpeech with flag-gated streaming TTS

feat(realtime): route response audio through emitSpeech (streaming TTS)

feat(realtime): streaming transcription text deltas

feat(realtime): pipeline disable_thinking maps to enable_thinking off

feat(realtime): speechStreamer for token-streamed LLM->TTS

feat(realtime): wire streamLLMResponse for token-streamed replies

docs(realtime): document pipeline streaming + disable_thinking

fix(realtime): register pipeline streaming/thinking config fields

fix(realtime): always strip reasoning from spoken output

fix(realtime): clean TTS temp path before read (gosec G304)

refactor(realtime): buffer whole message for TTS, drop sentence segmenter

feat(realtime): stream tool-call turns via tokenizer-template autoparser

feat(realtime): script-aware clause chunking + streamed-reply fixes