nexxtmobile.de 60d2294338 fix(realtime): strip EOU language tags from the LLM context ...

ASR backends like Nemotron-3.5 append end-of-utterance language tags to
transcripts ("<de-DE>", "<en-US>", ...). The committed transcript feeds
the LLM conversation (the user item Transcript field), but the per-utterance
tag is unreliable for short/ambiguous utterances (a German "Hallo." tagged
<en-US>) and the LLM treats it as a language signal, switching the reply
language against the session instructions.

commitUtteranceWithTranscript now strips the EOU tags from the transcript
that goes to appendUserItem/generateResponse (the LLM context). The emitted
input_audio_transcription event keeps the raw transcript (debugging /
client display).

Regression tests: stripEOULanguageTags unit specs + a client-commit spec
asserting the emitted event keeps the raw tag while the LLM user message is
tag-free.

Signed-off-by: nexxtmobile.de <git@nexxtmobile.de>
(cherry picked from commit dff488044c)
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

2026-11-03 14:40:50 +00:00

..

compactcoord

feat(realtime): Semantic VAD EOU token (#10444)

2026-06-30 09:01:22 +02:00

conncoord

feat(realtime): Semantic VAD EOU token (#10444)

2026-06-30 09:01:22 +02:00

coordinator

feat(realtime): Semantic VAD EOU token (#10444)

2026-06-30 09:01:22 +02:00

respcoord

fix(realtime): honor interrupt_response

2026-10-13 08:38:31 +00:00

ttscoord

feat(realtime): Semantic VAD EOU token (#10444)

2026-06-30 09:01:22 +02:00

turncoord

feat(realtime): Semantic VAD EOU token (#10444)

2026-06-30 09:01:22 +02:00

types

fix(realtime): honor interrupt_response

2026-10-13 08:38:31 +00:00

audio_upload_test.go

feat(parakeet-cpp): name speakers from the shared voice registry (#12382)

2026-10-01 08:25:52 +02:00

backend_errors.go

fix(openai): return HTTP errors for pre-stream failures and report per-slot context (#12204)

2026-09-22 17:19:58 +02:00

chat_assistant_gate_test.go

chore: Security hardening (#9719)

2026-05-08 16:25:45 +02:00

chat_assistant_gate.go

chore: Security hardening (#9719)

2026-05-08 16:25:45 +02:00

chat_emit_test.go

fix(streaming): comply with OpenAI usage / stream_options spec (#9815)

2026-05-14 08:53:46 +02:00

chat_emit.go

fix(openai): stream usage non-zero when tools are enabled (#9941)

2026-05-22 10:13:41 +02:00

chat_mcp_auto_gate_auth_test.go

feat(config): let model aliases carry an opt-in profile block (system prompt, MCP, client tools) (#12661)

2026-10-19 23:42:40 +02:00

chat_mcp_auto_gate_test.go

feat(addons): installable add-ons foundation and the Tables add-on (#12702)

2026-10-22 17:57:19 +02:00

chat_mcp_auto_test.go

feat(config): let model aliases carry an opt-in profile block (system prompt, MCP, client tools) (#12661)

2026-10-19 23:42:40 +02:00

chat_stream_reasoning_test.go

fix(streaming/tools): don't leak prefill-misclassified content as trailing reasoning chunk (#10000)

2026-05-26 08:34:26 +02:00

chat_stream_usage_test.go

fix(openai): stream usage non-zero when tools are enabled (#9941)

2026-05-22 10:13:41 +02:00

chat_stream_workers_test.go

fix(streaming/tools): stop healing-marker stubs from gating off content (#9999)

2026-05-25 23:55:35 +02:00

chat_stream_workers.go

fix(reasoning): ignore preclosed prompt markers

2026-09-11 22:44:39 +00:00

chat_test.go

[chat] feat: template.system_messages_after_first — merge or forward late system turns (#11906)

2026-09-09 22:21:32 +02:00

chat.go

feat(addons): installable add-ons foundation and the Tables add-on (#12702)

2026-10-22 17:57:19 +02:00

completion_stream_validation_test.go

fix(completions): reject empty PromptStrings in streaming to avoid index-out-of-range panic (#11028)

2026-07-22 09:29:11 +02:00

completion.go

fix(completions): reject empty PromptStrings in streaming to avoid index-out-of-range panic (#11028)

2026-07-22 09:29:11 +02:00

constants.go

fix(openai): stop max_tokens streaming retry loop on reasoning models (#9716) (#10448)

2026-06-30 09:01:53 +02:00

diarization_test.go

feat(voice): name registered speakers through a bundle speaker component

2026-10-05 23:45:14 +00:00

diarization.go

feat(diarization): return sound events with include_sounds (#12544)

2026-10-07 14:10:45 +02:00

edit.go

feat(middleware): Model routing, PII filtering, Cloud model proxies (#9802)

2026-05-25 09:28:27 +02:00

embeddings.go

feat(router): make KNN a first-class classifier with a persisted, curated corpus (#10652)

2026-08-18 09:37:43 +02:00

evidence_test.go

feat(addons): installable add-ons foundation and the Tables add-on (#12702)

2026-10-22 17:57:19 +02:00

evidence.go

feat(config): let model aliases carry an opt-in profile block (system prompt, MCP, client tools) (#12661)

2026-10-19 23:42:40 +02:00

image_test.go

Fix image upload processing and img2img pipeline in diffusers backend (#8879)

2026-03-11 08:05:50 +01:00

image.go

[httpclient] validate redirect targets and pin dialed addresses for user URLs (#12853)

2026-10-23 11:54:42 +02:00

inference_test.go

feat: interleaved thinking with tool calls (reasoning_content alias + Anthropic thinking blocks) (#10744)

2026-07-08 16:45:43 +00:00

inference.go

fix(openai): stop max_tokens streaming retry loop on reasoning models (#9716) (#10448)

2026-06-30 09:01:53 +02:00

inpainting_test.go

feat(distributed): gated X-LocalAI-Node response header (middleware + wrapper) (#9976)

2026-05-25 10:51:48 +02:00

inpainting.go

feat(distributed): gated X-LocalAI-Node response header (middleware + wrapper) (#9976)

2026-05-25 10:51:48 +02:00

list_capabilities_test.go

feat(images): add ESRGAN upscaling and Studio workflow (#12620)

2026-10-11 00:30:09 +02:00

list_capabilities.go

feat(images): add ESRGAN upscaling and Studio workflow (#12620)

2026-10-11 00:30:09 +02:00

list.go

feat(api): add GET /v1/models/capabilities endpoint (#10687)

2026-07-05 08:51:55 +02:00

moderations_test.go

feat(api): add text moderation endpoint (#11316)

2026-08-03 18:03:46 +02:00

moderations.go

feat(api): add text moderation endpoint (#11316)

2026-08-03 18:03:46 +02:00

openai_suite_test.go

Fix image upload processing and img2img pipeline in diffusers backend (#8879)

2026-03-11 08:05:50 +01:00

realtime_chained_novad_test.go

feat(realtime): LFM3-Audio speech-to-speech on llama.cpp, remove liquid-audio (#12958)

2026-10-29 22:06:41 +01:00

realtime_chunker_test.go

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

2026-06-11 08:43:12 +01:00

realtime_chunker.go

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

2026-06-11 08:43:12 +01:00

realtime_classifier_test.go

feat(failover): switch realtime pipeline stages per call

2026-09-27 07:42:20 +00:00

realtime_classifier.go

feat(classifier/VAD): support voice control on low power devices (#10804)

2026-07-29 12:50:22 +02:00

realtime_client_commit_test.go

fix(realtime): end a client response cleanly when its commit fails

2026-10-13 08:44:23 +00:00

realtime_commit_order_test.go

refactor(realtime): gate client responses on the commit tail

2026-10-13 08:44:22 +00:00

realtime_compactcoord.go

feat(realtime): Semantic VAD EOU token (#10444)

2026-06-30 09:01:22 +02:00

realtime_compaction_test.go

feat(realtime): Semantic VAD EOU token (#10444)

2026-06-30 09:01:22 +02:00

realtime_compaction.go

feat(realtime): Semantic VAD EOU token (#10444)

2026-06-30 09:01:22 +02:00

realtime_conncoord.go

feat(realtime): LFM3-Audio speech-to-speech on llama.cpp, remove liquid-audio (#12958)

2026-10-29 22:06:41 +01:00

realtime_doubles_test.go

feat(realtime): LFM3-Audio speech-to-speech on llama.cpp, remove liquid-audio (#12958)

2026-10-29 22:06:41 +01:00

realtime_eou_langtag_test.go

fix(realtime): strip EOU language tags from the LLM context

2026-11-03 14:40:50 +00:00

realtime_failover_test.go

feat(tts): stream Qwen text and add assistant preset (#12770)

2026-10-19 12:30:52 +02:00

realtime_failover.go

refactor: name the realtime pipeline stages with constants

2026-09-27 07:42:21 +00:00

realtime_gate_test.go

feat(realtime): LFM3-Audio speech-to-speech on llama.cpp, remove liquid-audio (#12958)

2026-10-29 22:06:41 +01:00

realtime_interrupt_test.go

feat(realtime): LFM3-Audio speech-to-speech on llama.cpp, remove liquid-audio (#12958)

2026-10-29 22:06:41 +01:00

realtime_lfm3_gallery_test.go

feat(realtime): LFM3-Audio speech-to-speech on llama.cpp, remove liquid-audio (#12958)

2026-10-29 22:06:41 +01:00

realtime_live_voices_test.go

feat(voice): name registered speakers through a bundle speaker component

2026-10-05 23:45:14 +00:00

realtime_modality_test.go

fix(realtime): accept legacy 'modalities' alias for output_modalities (fixes #11103) (#11104)

2026-07-26 23:09:30 +02:00

realtime_model.go

feat(realtime): LFM3-Audio speech-to-speech on llama.cpp, remove liquid-audio (#12958)

2026-10-29 22:06:41 +01:00

realtime_reasoning_test.go

feat: forward reasoning_effort to the backend so jinja models honor it (#10184)

2026-06-05 13:45:43 +00:00

realtime_reasoning.go

feat: forward reasoning_effort to the backend so jinja models honor it (#10184)

2026-06-05 13:45:43 +00:00

realtime_respcoord.go

fix(realtime): honor interrupt_response

2026-10-13 08:38:31 +00:00

realtime_s2s_test.go

feat(realtime): LFM3-Audio speech-to-speech on llama.cpp, remove liquid-audio (#12958)

2026-10-29 22:06:41 +01:00

realtime_s2s.go

feat(realtime): LFM3-Audio speech-to-speech on llama.cpp, remove liquid-audio (#12958)

2026-10-29 22:06:41 +01:00

realtime_semantic_vad_test.go

realtime: announce speakers in transcription-only sessions (#12871)

2026-10-24 00:07:02 +02:00

realtime_semantic_vad.go

feat(realtime): LFM3-Audio speech-to-speech on llama.cpp, remove liquid-audio (#12958)

2026-10-29 22:06:41 +01:00

realtime_sound_detection_test.go

fix(realtime): detach VAD-commit transcription from barge-in cancellation (#12446)

2026-10-05 01:27:40 +02:00

realtime_sound_detection.go

feat(ced): sound-event classification backend (CED audio tagger) (#10425)

2026-06-22 01:00:28 +02:00

realtime_speaker_event_test.go

feat(realtime): speaker-aware conversations - surface identity to client and LLM (#10424)

2026-06-21 21:07:10 +02:00

realtime_speech_test.go

feat(tts): stream Qwen text and add assistant preset (#12770)

2026-10-19 12:30:52 +02:00

realtime_speech.go

feat(tts): stream Qwen text and add assistant preset (#12770)

2026-10-19 12:30:52 +02:00

realtime_stream_test.go

feat(realtime): LFM3-Audio speech-to-speech on llama.cpp, remove liquid-audio (#12958)

2026-10-29 22:06:41 +01:00

realtime_stream.go

feat(tts): stream Qwen text and add assistant preset (#12770)

2026-10-19 12:30:52 +02:00

realtime_thinking_test.go

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

2026-06-11 08:43:12 +01:00

realtime_thinking.go

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

2026-06-11 08:43:12 +01:00

realtime_transcription_test.go

realtime: announce speakers in transcription-only sessions (#12871)

2026-10-24 00:07:02 +02:00

realtime_transcription.go

realtime: announce speakers in transcription-only sessions (#12871)

2026-10-24 00:07:02 +02:00

realtime_transport_webrtc.go

fix(realtime): raise WebRTC data-channel max-message-size + keep sendLoop alive (#10407)

2026-06-19 21:36:25 +02:00

realtime_transport_ws.go

feat(realtime): WebRTC support (#8790)

2026-03-13 21:37:15 +01:00

realtime_transport.go

feat(realtime): WebRTC support (#8790)

2026-03-13 21:37:15 +01:00

realtime_tts_buffered_test.go

feat(tts): stream Qwen text and add assistant preset (#12770)

2026-10-19 12:30:52 +02:00

realtime_tts_live_test.go

feat(tts): stream Qwen text and add assistant preset (#12770)

2026-10-19 12:30:52 +02:00

realtime_tts_live.go

feat(tts): stream Qwen text and add assistant preset (#12770)

2026-10-19 12:30:52 +02:00

realtime_tts_pipeline_test.go

feat(realtime): Semantic VAD EOU token (#10444)

2026-06-30 09:01:22 +02:00

realtime_tts_pipeline.go

feat(realtime): Semantic VAD EOU token (#10444)

2026-06-30 09:01:22 +02:00

realtime_turncoord.go

refactor(realtime): gate client responses on the commit tail

2026-10-13 08:44:22 +00:00

realtime_vad_buffer_test.go

feat(realtime): Semantic VAD EOU token (#10444)

2026-06-30 09:01:22 +02:00

realtime_vad_tick_test.go

feat(classifier/VAD): support voice control on low power devices (#10804)

2026-07-29 12:50:22 +02:00

realtime_voice_profile_test.go

fix(realtime): support voice profile switching (#11948)

2026-09-09 10:38:23 +02:00

realtime_voicegate_bundle_test.go

feat(config): let parakeet-cpp declare the speaker_recognition usecase

2026-10-05 23:46:06 +00:00

realtime_voicegate_integration_test.go

realtime: announce speakers in transcription-only sessions (#12871)

2026-10-24 00:07:02 +02:00

realtime_voicegate_test.go

feat(parakeet-cpp): name speakers from the shared voice registry (#12382)

2026-10-01 08:25:52 +02:00

realtime_voicegate.go

feat(classifier/VAD): support voice control on low power devices (#10804)

2026-07-29 12:50:22 +02:00

realtime_warmup_test.go

feat(realtime): eager blocking pipeline warm-up + /backend/load API (#10662)

2026-07-03 18:00:37 +02:00

realtime_webrtc_ice_test.go

fix(realtime): keep the ICE interface allow-list working with a fixed UDP port (#11466)

2026-08-11 18:37:49 +02:00

realtime_webrtc_ice.go

fix(realtime): keep the ICE interface allow-list working with a fixed UDP port (#11466)

2026-08-11 18:37:49 +02:00

realtime_webrtc_request_test.go

fix(realtime): accept GA WebRTC signaling (#11778)

2026-08-29 21:28:37 +02:00

realtime_webrtc_sctp_test.go

fix(realtime): raise WebRTC data-channel max-message-size + keep sendLoop alive (#10407)

2026-06-19 21:36:25 +02:00

realtime_webrtc_sctp.go

fix(realtime): raise WebRTC data-channel max-message-size + keep sendLoop alive (#10407)

2026-06-19 21:36:25 +02:00

realtime_webrtc.go

fix(realtime): accept GA WebRTC signaling (#11778)

2026-08-29 21:28:37 +02:00

realtime.go

fix(realtime): strip EOU language tags from the LLM context

2026-11-03 14:40:50 +00:00

sound_classification.go

fix(openai): answer 400 for a missing or malformed audio upload

2026-09-27 07:42:21 +00:00

transcription_resolve_test.go

fix(transcription): honor model-config language/translate + OpenAI language form field (#10731)

2026-07-07 21:16:25 +00:00

transcription.go

fix(audio): record transcription usage and traces (#12546)

2026-10-07 16:57:23 +02:00

upscale_test.go

feat(images): add ESRGAN upscaling and Studio workflow (#12620)

2026-10-11 00:30:09 +02:00

upscale.go

feat(images): add ESRGAN upscaling and Studio workflow (#12620)

2026-10-11 00:30:09 +02:00