LLMRTC Docs

👋 **Hey Product Hunt — I’m Samuel, the founder & chief tinkerer behind LLMRTC!** --- ### Why we built it After countless late-night prototypes I realised gluing WebRTC, STT, LLMs and TTS together felt like weaving four SDKs and a dozen websockets by hand. LLMRTC is my answer: a single TypeScript stack that turns *“speech-in, insight-out”* into <50 ms reality. ([llmrtc.org][1]) ### What LLMRTC does for you * **Real-time voice & vision** — sub-second, bidirectional streams with barge-in and VAD out of the box. * **Provider-agnostic** — flip between OpenAI, Anthropic, Gemini, Bedrock *or* local Ollama/Whisper/Piper with one line change. * **Tool-calling & Playbooks** — define JSON-Schema tools, stitch multi-stage conversations, and let the model drive the flow. * **Hooks + metrics everywhere** — 20+ tap-points plus built-in TTFT, token, and latency tracking so ops teams stay happy. * **Session resilience** — automatic reconnection and graceful degradation when networks (or providers) misbehave. ([llmrtc.org][1]) ### Things you can build today 🔹 Siri-style voice assistants • 🔹 Screen-aware support agents • 🔹 On-device, privacy-first kiosks • 🔹 Multilingual voice bots that cost pennies per conversation. ([llmrtc.org][1]) ### Quick taste ```bash npm i @llmrtc/llmrtc-backend @llmrtc/llmrtc-web-client ``` Spin up the server, connect from the browser, and start talking in <10 lines of code. The docs walk through full examples for cloud **and** local-only stacks. ([llmrtc.org][1]) --- 💬 **I’d love your feedback!** Try the quickstart, star the repo, or just drop a question below—AMA about WebRTC quirks, AI latency hacks, or anything in between. Thanks for checking out our launch! 🎉 [1]: https://www.llmrtc.org/providers... "Welcome to LLMRTC | LLMRTC Docs"