How karan27q/ai-agents-orchestrator works
This is an AI workflow automation and multi-agent research platform. Users authenticate, then build and execute DAG workflows or request streamed research; they can also upload documents for embedding and semantic search. The graph preserves the documented workflow, research, RAG, authorization, tracking, and real-time capabilities. Several subsystems are represented from the README and source index rather than sampled implementation; those links are omitted where call ownership is not evidenced.
Main entry points: main.py FastAPI application · app.js Web application · models.py Domain models · workflow_engine.py DAG execution engine
Components 32
Web client
- Web application
frontend/app.js - Sign-in view
frontend/components/auth.js - Dashboard
frontend/components/dashboard.js - Home view
frontend/components/home.js - Workflow editor
frontend/components/workflows.js - Research view
frontend/components/research.js - Files manager
frontend/components/files.js - Search view
frontend/components/search.js
API and access
- FastAPI application
backend/main.py - Authentication routes
backend/routes/auth.py - JWT and role access
backend/auth.py - IAM routes
backend/routes/iam.py - IAM services
backend/iam.py
Workflow and agents
- Workflow routes
backend/routes/workflows.py - DAG execution engine
backend/workflow_engine.py - Agent routes
backend/routes/agents.py - Research agents
backend/agents.py - LLM cache
backend/llm_cache.py - Execution error handling
backend/error_handler.py
Knowledge and search
- File routes
backend/routes/files.py - Vector store
backend/vector_store.py - Search routes
backend/routes/search.py
Persistence and operations
- Domain models
backend/models.py - Relational database
backend/database.py - Task queue
backend/task_queue.py - Task monitoring routes
backend/routes/tasks.py - Notification routes
backend/routes/notifications.py - WebSocket routes
backend/routes/ws.py - Connection manager
backend/ws_manager.py - Observability routes
backend/routes/observability.py
Other components
- Platform user
- Google Gemini
Connections 42
- Platform user → Web application: uses
README.md - Web application → Sign-in view: renders
frontend/app.js - Web application → Dashboard: renders
frontend/app.js - Web application → Home view: renders
frontend/app.js - Web application → Workflow editor: renders
frontend/app.js - Web application → Research view: renders
frontend/app.js - Web application → Files manager: renders
frontend/app.js - Web application → Search view: renders
frontend/app.js - Web application → FastAPI application: calls API
frontend/app.js - FastAPI application → Authentication routes: registers
backend/main.py - FastAPI application → Workflow routes: registers
backend/main.py - FastAPI application → Agent routes: registers
backend/main.py - FastAPI application → File routes: registers
backend/main.py - FastAPI application → Search routes: registers
backend/main.py - FastAPI application → Notification routes: registers
backend/main.py - FastAPI application → WebSocket routes: registers
backend/main.py - FastAPI application → Observability routes: registers
backend/main.py - FastAPI application → Task monitoring routes: registers
backend/main.py - FastAPI application → IAM routes: registers
backend/main.py - Authentication routes → JWT and role access: uses
backend/routes/auth.py - IAM routes → IAM services: uses
backend/routes/iam.py - Workflow routes → DAG execution engine: executes
backend/routes/workflows.py - Agent routes → Research agents: invokes
backend/routes/agents.py - Research agents → LLM cache: uses
backend/agents.py - File routes → Vector store: indexes files
backend/routes/files.py - File routes → Task queue: enqueues work
backend/routes/files.py - Search routes → Vector store: searches vectors
backend/routes/search.py - Task monitoring routes → Task queue: queries tasks
backend/routes/tasks.py - Notification routes → Connection manager: sends messages
backend/routes/notifications.py - WebSocket routes → Connection manager: manages sockets
backend/routes/ws.py - DAG execution engine → Execution error handling: uses
backend/workflow_engine.py - JWT and role access → Domain models: queries users
backend/auth.py - IAM services → Domain models: uses
backend/iam.py - DAG execution engine → Domain models: loads workflow
backend/workflow_engine.py - Vector store → Domain models: uses chunks
backend/vector_store.py - Task queue → Domain models: stores task state
backend/task_queue.py - Authentication routes → Relational database: opens session
backend/routes/auth.py - Workflow routes → Relational database: opens session
backend/routes/workflows.py - Relational database → Domain models: defines base
backend/database.py - DAG execution engine → Google Gemini: generates text
- Research agents → Google Gemini: researches with
- Research view → Agent routes: requests research