How datahub-project/analytics-agent works
Analytics Agent is a web application for asking warehouse questions in natural language and receiving SQL-backed results, charts, and conversational follow-ups. The main flow is browser chat to the chat API and LangGraph agent, which draws on configured LLMs and optional DataHub context, invokes a selected SQL or MCP engine, and streams responses back to the UI. The product also manages conversations, connections, OAuth, and settings. Relationships are limited to README statements and the supplied excerpts or source-index references; unsampled chat and UI internals are not assumed.
Main entry points: graph.py Agent graph · repository.py Persistence repositories · factory.py Engine registry · settings.py Settings API
Components 24
Web interface
- React app
frontend/src/App.tsx - Chat experience
frontend/src/components/Chat/ChatView.tsx - Setup wizard
frontend/src/components/Onboarding/OnboardingWizard.tsx - Settings and connections
frontend/src/components/Settings/SettingsModal.tsx
API and persistence
- Chat endpoint
backend/src/analytics_agent/api/chat.py - Settings API
backend/src/analytics_agent/api/settings.py - OAuth auth
backend/src/analytics_agent/api/oauth.py - Conversation API
backend/src/analytics_agent/api/conversations.py - Application database
backend/src/analytics_agent/db/base.py - Persistence repositories
backend/src/analytics_agent/db/repository.py
Agent and context
- Agent graph
backend/src/analytics_agent/agent/graph.py - LLM providers
backend/src/analytics_agent/agent/llm.py - Chart generation
backend/src/analytics_agent/agent/chart_generator.py - DataHub skills
backend/src/analytics_agent/skills/datahub_skills.py - DataHub context
backend/src/analytics_agent/context/datahub.py - MCP context platform
backend/src/analytics_agent/context/mcp_platform.py
Data connections
- Engine registry
backend/src/analytics_agent/engines/factory.py - SQL query engine
backend/src/analytics_agent/engines/sqlalchemy/engine.py - MCP query engine
backend/src/analytics_agent/engines/mcp/engine.py - Warehouse connectors
connectors
Other components
- Analyst
- Data warehouse
- LLM service
- DataHub service
Connections 26
- Analyst → React app: uses
README.md - React app → Chat experience: renders
frontend/src/App.tsx - React app → Settings and connections: opens
frontend/src/App.tsx - Chat experience → Chat endpoint: sends chat
- Settings and connections → Settings API: configures
- Setup wizard → Settings API: configures
- Chat endpoint → Agent graph: runs
- Chat endpoint → Persistence repositories: persists conversation
- Conversation API → Persistence repositories: reads and writes
- Settings API → Persistence repositories: reads and writes
backend/src/analytics_agent/api/settings.py - OAuth auth → Persistence repositories: stores credentials
backend/src/analytics_agent/api/oauth.py - Persistence repositories → Application database: queries
backend/src/analytics_agent/db/repository.py - Agent graph → LLM providers: requests completions
- Agent graph → Engine registry: selects engine
- Agent graph → DataHub skills: uses skills
- Agent graph → Chart generation: generates charts
- LLM providers → LLM service: calls provider
- DataHub skills → DataHub context: uses context
backend/src/analytics_agent/skills/datahub_skills.py - DataHub context → DataHub service: queries metadata
- Engine registry → SQL query engine: creates
backend/src/analytics_agent/engines/factory.py - Engine registry → MCP query engine: creates
backend/src/analytics_agent/engines/factory.py - Engine registry → Warehouse connectors: configures
backend/src/analytics_agent/engines/factory.py - SQL query engine → Data warehouse: executes SQL
README.md - MCP query engine → Warehouse connectors: discovers tools
backend/src/analytics_agent/engines/mcp/engine.py - Warehouse connectors → Data warehouse: queries data
- Settings API → MCP context platform: tests connection
backend/src/analytics_agent/api/settings.py