Your Chill AI Workspace
Chat. Create. Chill. All in one place.
At a Glance · Features · Getting Started · Architecture · Contributing
Chaya is a self-hosted AI workspace that brings together intelligent chat, media creation, and ambient vibes under a neon-noir cyberpunk interface. Connect your favorite LLMs, plug in MCP tools, define custom personas — and do it all with lofi beats playing in the background.
Multi-agent chat with real-time thinking chains and MCP tool execution
Orchestrate conversations across GPT-4o, Claude, Gemini, DeepSeek, Grok, and local Ollama models. Run multi-agent roundtables, watch thinking chains unfold in real time, and let MCP tools handle the heavy lifting.
- Multi-provider LLM — OpenAI / Anthropic / Google / DeepSeek / Grok / Nvidia / Ollama
- Multi-agent roundtable — multiple AI actors collaborating in a single session
- MCP tool integration — auto-discovery, OAuth, real-time execution logs
- Thinking chain visualization — transparent reasoning, step by step
- Skill packs — reusable prompt-based capabilities attached to agents
- TTS — ElevenLabs-powered text-to-speech on any message
- Discord bridge — per-channel actors with independent memory
Unified model registry — manage tokens, toggle models, one place
Generate images and videos with natural language. Paste reference images, pick aspect ratios, batch-generate variants, and push results straight to Google Drive.
- Text-to-image / Image-to-image — powered by Gemini, with more providers pluggable
- Video generation — Veo, Runway (expandable)
- Gallery timeline — browse outputs organized by date
- Google Drive sync — one-click upload to your cloud
Describe your scene, pick a ratio, hit generate
A built-in lofi / ambient radio that streams YouTube live stations. Browse, search, favorite — and keep the music playing while you work in any other tab. Because good vibes are not optional.
- YouTube live streams — curated lofi, jazz, synthwave, and more
- Global player — music persists across page navigation
- Search & favorites — build your own station list
- Mini bar — always-visible playback controls in the header
Stay in the zone. Lofi beats never stop.
Define AI personalities with custom system prompts, voice presets, memory triggers, and behavioral toggles. Each persona is a full character profile — not just a system message.
A handcrafted neon cyberpunk design system. Pure black backgrounds, glowing green accents, mist-pink highlights, and a three-tier card elevation system built for focus.
| Dependency | Version | Purpose |
|---|---|---|
| Node.js | 18+ | Frontend runtime |
| Python | 3.10+ | Backend runtime |
| MySQL | 8.0+ | Data storage |
| Redis | 6.0+ | Cache & messaging |
git clone https://github.com/your-org/chaya.git
cd chaya
# Backend
cd backend
python -m venv venv && source venv/bin/activate
pip install -r requirements.txt
python app.py
# Frontend (new terminal)
cd front
pnpm install && pnpm devOpen http://localhost:5177 and you're in.
- Add LLM providers — Settings > Model Registry > add your API keys
- Connect MCP servers — Settings > MCP > paste server URLs
- Set up a persona — Persona tab > create or import character profiles
- Start chatting — pick an agent, pick a model, go
chaya/
├── backend/ # Flask API server
│ ├── api/ # REST endpoints
│ ├── services/
│ │ ├── actor/ # Agent actor model & lifecycle
│ │ ├── providers/ # LLM provider adapters
│ │ ├── mcp/ # MCP client & auto-selection
│ │ ├── chill/ # YouTube ambient service
│ │ └── media/ # Image & video generation
│ └── app.py
│
├── front/ # React 19 + Vite 7 + TypeScript
│ └── src/
│ ├── components/ # UI — Workflow, Chatu, Chill, Agents...
│ ├── services/ # API clients, context, state
│ └── index.css # Niho design system tokens
│
├── docs/ # Additional documentation
└── electron/ # Electron shell (experimental)
Frontend: React 19, Vite 7, TypeScript, Tailwind CSS, Radix UI, MCP SDK
Backend: Flask, Python 3.10+, MySQL, Redis, Actor-based agent runtime
Contributions welcome. Fork, branch, PR — the usual flow.
git checkout -b feature/your-idea
git commit -m "add: your idea"
git push origin feature/your-ideaThen open a Pull Request.
Built with good vibes and too much coffee.




