How gzwebsj-163/velthyn works
Velthyn is a multi-channel AI agent platform with a macOS Electron client. Its principal workflow accepts a message through an IM, Web, or terminal channel, bridges it into shared chat/session orchestration, uses model and agent capabilities, then returns a response through the channel. The README documents the major agent subsystems and integrations; sampled source confirms the desktop starts a local backend and loads its chat URL. Unsampled implementation details are represented only where documented, and links cite the README when its architecture description is the evidence.
Main entry points: service.mo Chat orchestration · bot_factory.mo Model routing · channel IM and Web channels · web Local chat backend
Components 20
Channels and Client
- IM and Web channels
server/channel - macOS desktop client
desktop/main.js
Agent Runtime
- Chat orchestration
server/agent/chat/service.mo - Agent tools
server/agent/tools/tool_manager.mo - Memory
server/agent/memory/service.mo - Knowledge RAG
server/agent/knowledge/service.mo - Agent skills
server/agent/skills/service.mo - Self evolution
server/agent/evolution/executor.mo - Prompt building
server/agent/prompt/builder.mo - Workspace
server/agent/workspace/service.mo
Models and Extensions
- Model routing
server/models/bot_factory.mo - MCP servers
server/agent/tools/mcp/mcp_client.mo - Plugins
server/plugins/plugin_manager.mo - Global skill library
server/skills - Voice services
server/voice/factory.mo - Translation
server/translate/factory.mo
Service Platform
- Channel bridge
server/bridge - Local chat backend
server/channel/web - Server application
server/app.mo
Other components
- Chat user
Connections 21
- Chat user → IM and Web channels: sends message
README.md - IM and Web channels → Channel bridge: routes input
README.md - Channel bridge → Chat orchestration: dispatches session
README.md - Chat orchestration → Model routing: requests generation
README.md - Chat orchestration → Agent tools: calls tools
README.md - Chat orchestration → Memory: uses memory
README.md - Chat orchestration → Knowledge RAG: retrieves knowledge
README.md - Chat orchestration → Agent skills: loads skills
README.md - Chat orchestration → Self evolution: invokes evolution
README.md - Chat orchestration → Prompt building: builds prompts
- Agent tools → MCP servers: exposes MCP tools
README.md - Agent skills → Global skill library: loads library
- Chat orchestration → Workspace: uses workspace
- Server application → IM and Web channels: hosts channels
- macOS desktop client → Local chat backend: starts backend
desktop/main.js - macOS desktop client → Local chat backend: loads chat URL
desktop/main.js - Local chat backend → Chat orchestration: serves agent chat
- Model routing → Voice services: supports voice replies
README.md - Model routing → Translation: supports translation
- Server application → Plugins: loads plugins
- Chat orchestration → Plugins: uses extensions