What's Changed
- debug QOL by @theomonnom in #2150
- use the forkserver on Linux by default by @theomonnom in #2238
- update to use property. Also make consistent by @milo157 in #2236
- livekit-plugins-aws: update the dependencies to allow for versions greater than specified version by @mike-r-mclaughlin in #2244
- fix quick rotation when flush with empty audio by @longcw in #2251
- add image_encode_options for google realtime model by @longcw in #2249
- Gemini Live improvements & bugfixes by @davidzhao in #2247
- 11labs: improve resilience, do not fail when unexpected payload is received by @davidzhao in #2255
- feat: support cached prompt metrics for Gemini LLM by @mateuszkulpa in #2256
- fix google-cloud-texttospeech and anthropic min versions by @longcw in #2268
- fix raw function tool for google gemini by @longcw in #2270
- support simple word and sentence split for character-based language by @longcw in #2263
- remove leading whitespace from gemini user transcripts by @davidzhao in #2254
- deepgram: default to aura-2 for TTS by @davidzhao in #2272
- add realtime model metrics by @longcw in #2275
- revert buffer changes to AudioStreamDecoder by @davidzhao in #2286
- feat: Add max_completion_tokens parameter to OpenAI LLM by @dorlanpabon in #2258
- fix gather not retrieved error in stt by @longcw in #2276
- correctly pass timeout as connect timeout instead of total by @davidzhao in #2285
- Partial fix for Neuphonic bug detailed in issue #2281. by @adnansiddiquei in #2282
- Fix Neuphonic CI Tests. by @adnansiddiquei in #2287
- feat: allow to switch between Gemini and Vertex AI using env vars by @mateuszkulpa in #2292
Full Changelog: https://github.com/livekit/agents/compare/livekit-agent@1.0.20...livekit-agents@1.0.21