livekit-agents@1.8.7

## Realtime E2E latency

You can now measure realtime replies from the last detected user speech to first assistant playback when you configure local VAD. Read the latency from assistant message metrics:

```python
@session.on("conversation_item_added")
def on_message(event):
    if event.item.type == "message" and event.item.role == "assistant":
        print(event.item.metrics.get("e2e_latency"))
```

## New providers and models

- **Claude Opus 5, Opus 5.5 and Sonnet 5.5**: model support through the Anthropic SDK upgrade.

## Breaking changes

- **Telemetry**: fallback attributes `lk.fallback.label` and `lk.fallback.index` move to adapter request spans; update trace queries for the flattened hierarchy.
- **Anthropic**: custom SDK HTTP clients must use `httpx2`, not `httpx`.
- **OpenTelemetry**: replace the removed Events API with `LogRecord(event_name=...)`; custom `Logger` subclasses must implement `enabled()`.

## Behavior changes

- **Dependencies**: now requires `openai>=3.19,<4`, `mcp>=2.2,<3`, `websockets>=17.1,<17.2`, and OpenTelemetry `>=1.45,<1.46`; Anthropic requires `>=1.8,<2`, and turn detection requires `onnxruntime>=1.30,<1.31`.
- **TLS**: contexts are cached per trust configuration; restart workers after replacing a CA bundle at the same path.
- **Tools**: tools still running at interruption timeout now continue asynchronously and deliver deferred results.
- **Google**: instruction-free realtime replies are allowed on immutable-context models; unavailable timing is reported as unknown.
- **ElevenLabs**: STT client errors now retain their status and fail without retries.

Recording shutdown, VAD resizing, interruption teardown, chat summaries, structured outputs, multilingual sentence splitting, and tool discovery are more reliable.
## What's Changed
* fix(deps): update third party deps by @renovate[bot] in https://github.com/livekit/agents/pull/5489
* fix(deps): update third party deps (major) by @renovate[bot] in https://github.com/livekit/agents/pull/6268
* fix(upliftai): declare the python-socketio dependency by @zaidqureshi2 in https://github.com/livekit/agents/pull/7567
* fix(google): only block generate_reply when instructions passed to immutable-context model by @d42f in https://github.com/livekit/agents/pull/5715
* flatten fallback/stream adapter spans, name model on attempts by @davidzhao in https://github.com/livekit/agents/pull/7723
* fix(elevenlabs): fail fast on 4xx errors in STT recognize() by @pradneshfernandez in https://github.com/livekit/agents/pull/7701
* fix(http_context): build the TLS context once per process, not once per job by @zdurm in https://github.com/livekit/agents/pull/7617
* fix(ivr): reimplement TfidfLoopDetector with numpy to remove scikit-learn dependency by @Mugunth007 in https://github.com/livekit/agents/pull/6867
* fix: preserve nullable literal response schemas by @liangzhechen655 in https://github.com/livekit/agents/pull/7715
* fix: discard resampler history when resetting audio buffer by @liangzhechen655 in https://github.com/livekit/agents/pull/7713
* soniox: add reduce_silence option to TTS by @razvanvoica in https://github.com/livekit/agents/pull/6798
* fix(google): report unknown timing for results without word offsets by @15054538509 in https://github.com/livekit/agents/pull/7687
* fix: measure e2e latency for realtime agents by @chenghao-mou in https://github.com/livekit/agents/pull/7719
* Only send modern transcriptions to client sdks which can auto back convert modern -> legacy transcriptions internally by @1egoman in https://github.com/livekit/agents/pull/7240
* fix: reference real tool name in ToolProxyToolset description and errors by @BlueX888 in https://github.com/livekit/agents/pull/7734
* fix(voice): don't hang or buffer audio when the recorder encoder fails by @ANDROIDHASSAN in https://github.com/livekit/agents/pull/7741
* fix(vad): clamp cursor when resizing speech buffer by @gudcks0305 in https://github.com/livekit/agents/pull/7566
* fix(tokenize): split sentences on Arabic, Urdu and Devanagari terminators by @ahmad65saad in https://github.com/livekit/agents/pull/7509
* fix(interruption): tolerate websocket teardown race on session.close write by @BhaskarKapri07 in https://github.com/livekit/agents/pull/6893
* fix(telemetry): count a watchdog starved after it woke inside the stall window by @zdurm in https://github.com/livekit/agents/pull/7612
* fix(llm): keep earlier summaries when summarizing chat context again by @RaphaelFakhri in https://github.com/livekit/agents/pull/7721
* feat(deepgram): make Flux audio chunk duration configurable by @dnaumenko in https://github.com/livekit/agents/pull/7205
* feat(llm): report prompt-cache write tokens as cache_creation_tokens by @adrian-cowham in https://github.com/livekit/agents/pull/7710
* docs: fix broken DeepWiki badge in README by @longcw in https://github.com/livekit/agents/pull/7770
* fix: split underscore terms in BM25 description and param tokens by @BlueX888 in https://github.com/livekit/agents/pull/7733
* fix(voice): answer every tool return after ctx.update() by @longcw in https://github.com/livekit/agents/pull/7777
* fix(voice): release tools still running when an interrupted speech times out by @longcw in https://github.com/livekit/agents/pull/7759
* ci: draft release notes in the release PR by @longcw in https://github.com/livekit/agents/pull/7779
* fix: preserve reply_required for invalid tool results by @chenghao-mou in https://github.com/livekit/agents/pull/7434
* feat(worker): add preload_modules to forkserver preload by @tanish111 in https://github.com/livekit/agents/pull/7761
* fix: flush buffered tts audio before playback drains by @chenghao-mou in https://github.com/livekit/agents/pull/7692
* feat(llm): add CacheBreakpoint content marker for OpenAI prompt caching by @adrian-cowham in https://github.com/livekit/agents/pull/7459
* feat(inference): send prompt cache breakpoints for GPT-5.6 and later by @adrian-cowham in https://github.com/livekit/agents/pull/7461
* feat(openai): send prompt cache breakpoints for GPT-5.6 and later by @adrian-cowham in https://github.com/livekit/agents/pull/7464
* feat(voice): end the instructions message with a cache breakpoint each turn by @adrian-cowham in https://github.com/livekit/agents/pull/7475
* feat(llm): add Instructions(dynamic=...) for per-call text behind a cache breakpoint by @adrian-cowham in https://github.com/livekit/agents/pull/7478
* feat(openai): send prompt cache breakpoints from the Responses LLM by @adrian-cowham in https://github.com/livekit/agents/pull/7656
* feat(inference): add openai/gpt-live-transcribe STT by @shawnfeldman in https://github.com/livekit/agents/pull/7479
* feat(fishaudio): support drama-3 model by @cshape in https://github.com/livekit/agents/pull/7799
* gemini 4 argon by @tinalenguyen in https://github.com/livekit/agents/pull/7790
* feat(inference): add language_hint to DeepgramFluxOptions by @adrian-cowham in https://github.com/livekit/agents/pull/7797
* fix(rime): use model-specific WebSocket endpoints by @naszzz in https://github.com/livekit/agents/pull/7766
* livekit-agents@1.8.7 by @github-actions[bot] in https://github.com/livekit/agents/pull/7792

## New Contributors
* @d42f made their first contribution in https://github.com/livekit/agents/pull/5715
* @pradneshfernandez made their first contribution in https://github.com/livekit/agents/pull/7701
* @Mugunth007 made their first contribution in https://github.com/livekit/agents/pull/6867
* @liangzhechen655 made their first contribution in https://github.com/livekit/agents/pull/7715
* @razvanvoica made their first contribution in https://github.com/livekit/agents/pull/6798
* @15054538509 made their first contribution in https://github.com/livekit/agents/pull/7687
* @ANDROIDHASSAN made their first contribution in https://github.com/livekit/agents/pull/7741
* @gudcks0305 made their first contribution in https://github.com/livekit/agents/pull/7566
* @BhaskarKapri07 made their first contribution in https://github.com/livekit/agents/pull/6893
* @dnaumenko made their first contribution in https://github.com/livekit/agents/pull/7205

**Full Changelog**: https://github.com/livekit/agents/compare/livekit-agents@1.8.6...livekit-agents@1.8.7