How akshayetukuri/nexus-multi-agent-ai works
NEXUS accepts requests through a web or terminal interface, classifies them with an LLM-backed router, and returns a response from one of four role-specific agents. Notes requests can also read or persist notes. The README describes Groq for the web version and Ollama for the terminal version, but the sampled source sends routing requests to local Ollama and agent responses to Groq from both interfaces; the graph reflects that code. The agent features are documented, but their implementations are not separately sampled.
Main entry points: app.py Web interface · main.py Terminal interface · router.py Request router · agent.py Agent behavior
Components 14
User interfaces
- Web interface
app.py - Terminal interface
main.py
Intent routing
- Request router
router.py
Specialist agents
- Agent instances
agents_setup.py - Agent behavior
agent.py - Calendar agent
agents_setup.py - Email agent
agents_setup.py - Task agent
agents_setup.py - Notes agent
agents_setup.py
Notes persistence
- Notes storage
notes_storage.py - Persistent notes
Model services
- Ollama service
- Groq API
Other components
- User
Connections 13
- User → Web interface: submits request
- User → Terminal interface: enters request
- Web interface → Request router: routes request
- Terminal interface → Request router: routes request
- Request router → Ollama service: classifies request
- Request router → Agent instances: selects agent
- Agent instances → Agent behavior: creates agents
- Web interface → Agent behavior: requests response
- Terminal interface → Agent behavior: requests response
- Agent behavior → Groq API: generates response
- Web interface → Notes storage: reads or saves notes
- Terminal interface → Notes storage: reads or saves notes
- Notes storage → Persistent notes: reads and writes