Kayori v2

Kayori v2 is an async, adapter-based AI companion built around LangGraph, platform runtimes, and an orchestrated message pipeline.

What Kayori Does

Kayori is an intelligent conversational agent that connects to multiple platforms (Discord, Telegram, Webhook) and provides:

• Natural Conversations - ReAct-based reasoning with context-aware responses using LangGraph
• Tool Execution - Real-world actions like weather lookup, Spotify control, reminders, and web search
• Emotional Intelligence - Mood analysis across 28 emotions that adapts responses based on user sentiment
• Memory - Short-term conversation history plus long-term episodic memory via vector databases
• Proactive Behavior - Scheduler-driven actions including curiosity-based engagement and mood-triggered responses
• Audio Support - Full speech-to-text (Whisper) and text-to-speech (EdgeTTS) pipeline

Think of it as a personal AI assistant that lives in your chat platforms, remembers conversations, understands emotions, and can take actions on your behalf.

Architecture Overview

Message Flow:

Input → Gateway BUS → Orchestrator → Agent → Output Sink → Response

Core Components:

• Gateway BUS - Central message bus decoupling input from processing
• Orchestrator - Manages state and agent execution
• Agent - LangGraph ReAct agent with tool access
• Output Sink - Routes responses (direct or multi-platform)
• Scheduler - Drives proactive behaviors

Features

• Multi-Platform: Discord, Telegram, Webhook runtimes
• Audio Pipeline: Whisper STT + EdgeTTS
• Memory Systems: Short-term, Episodic (Pinecone), Graph (Neo4j)
• Mood Engine: 28 emotion dimensions with dynamic analysis
• Tools: Weather, Reminder, Spotify, Tavily Search + MCP tools
• Scheduler: Fuzzy/precise scheduling, curiosity triggers, mood thresholds

Quick Start

Prerequisites: Python 3.14+, API keys for services you enable

Environment (.env):

API_KEY=your_groq_api_key
DISCORD_BOT_TOKEN=your_discord_token
DISCORD_USER_ID=your_user_id

Install:

uv sync  # or: pip install .

Run:

python examples/main.py

Project Structure

src/
├── adapters/    # Input/Output, Bus, Memory, State, Scheduler
├── agent/       # ReAct agent service
├── core/        # Orchestrator, OutputSink, MoodEngine, Scheduler
├── tools/       # Built-in tools
├── mcp/         # MCP tool integrations
└── shared_types/# Models and protocols

License

MIT. See LICENSE.