How zda25m005-netizen/agentic-ai-os works
Enterprise Agentic AI OS provides agent-driven research and career workflows through a FastAPI service, with a browser interface, retrieval, persistent missions and memory, and report generation. The documented request path is a client submitting a goal, the agent planning and using tools, then returning an answer; a separate RAG path ingests documents and produces cited answers. The graph also preserves the documented GraphRAG, observability, feedback, and fine-tuning capabilities. Some subsystem links are grounded only in README or import references; unsampled implementation details are not inferred.
Main entry points: main.py FastAPI service · graph.py Agent graph · runtime.py Mission runtime · worker.py Mission worker
Components 24
Client Interface
- Web workspace
frontend/app - FastAPI service
app/api/main.py - Knowledge API client
frontend/app/lib/knowledgeApi.ts
Agent Orchestration
- Agent graph
app/agents/graph.py - Planner executor critic
app/agents - Tool use
app/agents/tool_loop.py - Tool registry
app/tools/registry.py
Knowledge Retrieval
- Document ingestion
app/rag/ingest.py - Hybrid retrieval
app/rag/retriever.py - Vector store
app/rag/vectorstore.py - GraphRAG
app/graph
Missions And Reports
- Mission endpoints
app/api/missions.py - Mission worker
app/missions/worker.py - Mission runtime
app/missions/runtime.py - Mission repository
app/missions/repository.py - Evidence analysis
app/analysis/pipeline.py - Report rendering
app/exec/latex/pipeline.py - Career services
app
Platform Services
- Long-term memory
app/memory/orchestrator.py - LLM client
app/core/llm.py - Feedback loop
app/feedback - Metrics and tracing
app/obs - LoRA pipeline
app/finetune
Other components
- Client
Connections 24
- Client → Web workspace: uses
README.md - Client → FastAPI service: calls
README.md - FastAPI service → Agent graph: runs
app/api/main.py - Agent graph → Planner executor critic: dispatches
app/agents/graph.py - Agent graph → Tool use: uses
- Tool use → Tool registry: invokes
- FastAPI service → Hybrid retrieval: answers with RAG
README.md - Document ingestion → Vector store: writes vectors
app/rag/ingest.py - Hybrid retrieval → Vector store: searches
app/rag/retriever.py - FastAPI service → GraphRAG: offers GraphRAG
README.md - FastAPI service → Mission endpoints: registers routes
app/api/main.py - Mission worker → Mission runtime: ticks
app/missions/worker.py - Mission worker → Mission repository: lists missions
app/missions/worker.py - Mission runtime → Mission repository: persists state
- Mission runtime → Evidence analysis: builds evidence
- Evidence analysis → Report rendering: supplies artifact
- Agent graph → Long-term memory: uses memory
app/agents/graph.py - Planner executor critic → LLM client: calls
app/agents/graph.py - FastAPI service → Feedback loop: records feedback
README.md - FastAPI service → Metrics and tracing: exposes metrics
README.md - LLM client → Metrics and tracing: records usage
app/core/llm.py - FastAPI service → Career services: serves career routes
app/api/main.py - FastAPI service → LoRA pipeline: supports fine-tuned serving
README.md - Knowledge API client → Web workspace: serves knowledge UI
frontend/app/lib/knowledgeApi.ts