feed · watching · stats · help with your tokens
kwami-labs/kwami
Kwami is a browser TypeScript library for creating 3D interactive AI companions with visual, audio, and speech capabilities.
Kwami solves the problem of creating engaging, interactive AI companions by providing a library that combines visual avatars, voice pipelines, and AI capabilities for developers building web-based conversational agents. Kwami is a browser TypeScript library that renders a WebGL avatar on a canvas and connects it to a LiveKit-backed voice pipeline for real-time interaction.
Similar
- JustVugg/colibri
colibri is a C inference engine that runs very large open mixture-of-experts models on the machine you already have.
- deepseek-ai/dsh-superagent-prototype
dsh-superagent is a research prototype for a proactive coding agent that works on a single GitHub repository using DeepSeek Harness.
- anthropics/FlashProve
FlashProve is a research prototype for zero-knowledge proofs of AI inference that allows providers to prove model identity without revealing weights.
- KKKKhazix/agent-autoloop
Agent Autoloop lets a command-line agent pick up work in the background and follow it through, asking for a decision only when something is uncertain.
- pgEdge/ai-dba-workbench
A PostgreSQL monitoring system that combines traditional and AI-based monitoring for instances v14+ including Supabase and Amazon RDS.
- Anionex/treeAI
TreeAI is a local-first visual workspace for exploring, comparing, and exporting LLM conversation trees across OpenAI-compatible models.
- moesaif/dna-studio
DNA Studio is a self-hosted, open-source AI marketing platform that generates on-brand social media content from a website URL.
- NVIDIA/auto-ontology
An open context layer that turns database metadata into an ontology, so AI agents answer questions about enterprise data with fewer wrong answers.
sources
Claims below include README evidence. A matching quote or worker agreement does not establish that a claim is correct. Category majority among 2 workers. Version 1.
Kwami is a browser TypeScript library for creating 3D interactive AI companions with visual, audio, and speech capabilities.
Kwami is a browser TypeScript library for 3D AI companions.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Kwami solves the problem of creating engaging, interactive AI companions by providing a library that combines visual avatars, voice pipelines, and AI capabilities for developers building web-based conversational agents.
Kwami is a browser TypeScript library for 3D AI companions. Drop a canvas on the page, give the companion a soul, and it can speak, listen, remember, and use tools — with a WebGL avatar that reacts in real time.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Kwami is a browser TypeScript library that renders a WebGL avatar on a canvas and connects it to a LiveKit-backed voice pipeline for real-time interaction.
Kwami is a browser TypeScript library for 3D AI companions. Drop a canvas on the page, give the companion a soul, and it can speak, listen, remember, and use tools — with a WebGL avatar that reacts in real time.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Provides 3D avatars including blob-xyz, black-hole, particles-face, and eye-iris with audio-reactive motion.
3D avatars — blob-xyz, black-hole, particles-face, and eye-iris, with skins, presets, and audio-reactive motion
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Features a voice pipeline that mixes STT, LLM, and TTS or uses a realtime provider via LiveKit.
Voice pipeline — mix-and-match STT / LLM / TTS, or a realtime provider, dispatched through LiveKit
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Supports a provider catalog including Deepgram, OpenAI, Anthropic, Gemini, ElevenLabs, Groq, Cartesia, and AssemblyAI.
Provider catalog — Deepgram, OpenAI, Anthropic, Gemini, ElevenLabs, Groq, Cartesia, AssemblyAI, and more
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Includes soul presets for ready-made personalities like Kaya, Nexus, Phoenix, and Haven with live trait updates.
Soul presets — ready-made personalities (Kaya, Nexus, Phoenix, Haven, …) plus live trait updates
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Offers optional long-term memory so the companion remembers the user across sessions.
Memory — optional long-term context so the companion remembers the user across sessions
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Allows connecting Model Context Protocol (MCP) servers to expose external tools to the agent.
MCP tools — connect Model Context Protocol servers and expose their tools to the agent
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Supports registering native skills that execute behaviors directly in the browser page.
Skills — register native behaviors that run in the page
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Ships as typed ESM with generated .d.ts files and requires three.js as a peer dependency.
Typed ESM — kwami ships as ESM with generated .d.ts; three is a peer dependency
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Provides multiple 3D avatar renderers including blob, black hole, particles, and iris with audio-reactive motion.
3D avatars — blob-xyz, black-hole, particles-face, and eye-iris, with skins, presets, and audio-reactive motion
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Offers a voice pipeline supporting STT, LLM, and TTS mix-and-match via LiveKit for real-time interaction.
Voice pipeline — mix-and-match STT / LLM / TTS, or a realtime provider, dispatched through LiveKit
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Supports a wide catalog of providers including Deepgram, OpenAI, Anthropic, Gemini, ElevenLabs, and Groq.
Provider catalog — Deepgram, OpenAI, Anthropic, Gemini, ElevenLabs, Groq, Cartesia, AssemblyAI, and more
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Includes ready-made personality presets like Kaya and Nexus with the ability to update traits live.
Soul presets — ready-made personalities (Kaya, Nexus, Phoenix, Haven, …) plus live trait updates
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Provides optional long-term memory using a Zep adapter to recall user context across sessions.
Memory — optional long-term context so the companion remembers the user across sessions
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Enables connection to Model Context Protocol servers to expose external tools to the AI agent.
MCP tools — connect Model Context Protocol servers and expose their tools to the agent
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Allows developers to register native skills and behaviors that execute directly within the page.
Skills — register native behaviors that run in the page
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Distributes as typed ESM with generated type definitions and requires three as a peer dependency.
Typed ESM — kwami ships as ESM with generated .d.ts; three is a peer dependency
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: Requires a WebGL2-capable browser to render the avatar.
A WebGL2-capable browser is required.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: Voice, memory, and tool dispatch require a LiveKit-backed agent and a user-controlled token endpoint; the library does not host these services.
Voice, memory, and tool dispatch need a LiveKit-backed agent and a token endpoint you control — the library does not host those for you.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: LLM, TTS, and STT API keys must not be placed in the browser.
Do not put LLM / TTS / STT API keys in the browser.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: Instances must be explicitly disposed to prevent leaking WebGL contexts and LiveKit rooms.
Always dispose() when the canvas leaves the page. Instances stay in a process-wide registry until you do; an unmounted component leaks the WebGL context and the LiveKit room.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: The library does not host voice, memory, or tool dispatch services; users must provide a LiveKit-backed agent and token endpoint.
The avatar is local: new Kwami(canvas) draws immediately. Voice, memory, and tool dispatch need a LiveKit-backed agent and a token endpoint you control — the library does not host those for you.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: Requires a WebGL2-capable browser to function.
A WebGL2-capable browser is required.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: Users must not store LLM, TTS, or STT API keys in the browser; these must be managed on the server.
Do not put LLM / TTS / STT API keys in the browser.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: Avatar displacement runs in a Web Worker but falls back to the main thread if a worker cannot start due to CSP restrictions.
The blob re-shapes its surface in a Web Worker that ships inside the package, so your page's main thread stays free. It falls back to the main thread by itself where a worker cannot start
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: Failing to call dispose() on unmounted components will leak the WebGL context and the LiveKit room.
Instances stay in a process-wide registry until you do; an unmounted component leaks the WebGL context and the LiveKit room.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: Requires three to be installed as a peer dependency; relying on a transitive copy is not supported.
three is a peer dependency. Install it beside kwami; do not rely on a transitive copy.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
ai-ml · web-ui