How athena-cyber1/athena works
Athéna is a conversational cyber assistant whose main workflow sends a user’s chat through the Next.js API to a Python agent pipeline. The agent plans and performs actions, checks evidence, and formats a final response; LLM completion is supporting rather than authoritative. The documented product also includes file ingestion, training endpoints, MCP tools, browser automation, and optional local-PC command execution. Several links among unsampled subsystems are documented only at the service/API level, so the graph avoids asserting their internal call paths.
Main entry points: agent.py Agent loop · serveur.py HTTP service · route.ts Chat API · store.py Memory and RAG
Components 24
Web application
- Chat interface
src/app/page.tsx - Chat API
src/app/api/chat/route.ts - Origin security
src/lib/secu.ts - File API
src/app/api/files/route.ts - MCP API
src/app/api/mcp/route.ts - Training API
src/app/api/entrainer/route.ts - Browser API
src/app/api/browser/route.ts - PC command API
src/app/api/exec/route.ts
Agent pipeline
- Agent loop
mini-services/llm-chat/athena/agent/agent.py - Planning and policy
mini-services/llm-chat/athena/agent/planner.py - Observation
mini-services/llm-chat/athena/agent/observer.py - Agent state
mini-services/llm-chat/athena/agent/state.py - Sub-agent
mini-services/llm-chat/athena/agent/sous_agent.py - Skills
mini-services/llm-chat/athena/agent/skills.py - Response voice
mini-services/llm-chat/athena/agent/voix.py
Knowledge and tools
- Memory and RAG
mini-services/llm-chat/athena/memory/store.py - MCP tools
mini-services/llm-chat/athena/tools/mcp.py
Local services
- HTTP service
mini-services/llm-chat/serveur.py - LLM engine
mini-services/llm-chat/athena/llm/engine.py - LLM bridge
mini-services/llm-bridge/index.ts - Local agent
mini-services/local-agent/index.js
Verification
- Verification checks
mini-services/llm-chat/athena/verification/canari.py
Other components
- User
- Model providers
Connections 26
- User → Chat interface: sends chat
README.md - Chat interface → Chat API: submits messages
- Chat API → Origin security: checks origin
src/app/api/chat/route.ts - Chat API → HTTP service: forwards chat
README.md - HTTP service → Agent loop: calls run_agent
mini-services/llm-chat/serveur.py - Agent loop → Planning and policy: uses planning
mini-services/llm-chat/athena/agent/agent.py - Agent loop → Observation: uses observer
mini-services/llm-chat/athena/agent/agent.py - Agent loop → Agent state: uses state
mini-services/llm-chat/athena/agent/agent.py - Agent loop → Sub-agent: uses sub-agent
mini-services/llm-chat/athena/agent/agent.py - Agent loop → Skills: uses skills
mini-services/llm-chat/athena/agent/agent.py - Agent loop → Response voice: formats response
mini-services/llm-chat/athena/agent/agent.py - Agent loop → LLM engine: requests completion
- LLM engine → LLM bridge: requests completion
mini-services/llm-chat/athena/llm/engine.py - LLM bridge → Model providers: routes model requests
README.md - Agent loop → Memory and RAG: reads and records
mini-services/llm-chat/athena/agent/agent.py - Agent loop → Verification checks: checks output
mini-services/llm-chat/athena/agent/agent.py - File API → HTTP service: proxies file requests
- HTTP service → Memory and RAG: stores file records
mini-services/llm-chat/serveur.py - MCP API → HTTP service: proxies MCP requests
- HTTP service → MCP tools: exposes MCP operations
mini-services/llm-chat/serveur.py - Agent loop → MCP tools: uses MCP tools
mini-services/llm-chat/athena/agent/agent.py - Training API → HTTP service: proxies training
- HTTP service → Memory and RAG: records training pairs
mini-services/llm-chat/serveur.py - Browser API → Local agent: proxies browser actions
src/app/api/browser/route.ts - User → PC command API: confirms command
README.md - PC command API → Local agent: forwards command
README.md