How kibrom1/bridgemind works
BridgeMind is a requirements-stage design for a multi-tenant agent that answers questions across databases, APIs, documents, and live web sources. The intended user flow is a query through the chat interface, handled by an orchestrator that can use retrieval and specialized SQL, API, document, or web capabilities; responses should include source citations. The README names these components and describes their goals, but no implementation excerpts are available, so links not explicitly documented are marked unverified or omitted.
Components 20
Query experience
- Web interface
- Chat route
- Agent orchestrator
- Retrieval
Data access
- SQL tools route
- SQL agent
- DB connector manager
- API tools route
- API agent
- API connector manager
Knowledge sources
- File agent
- Document source manager
- Web connector manager
- Web agent
Platform storage
- PostgreSQL
Other components
- User
- LLM provider
- Connected APIs
- Live web sources
- Object storage
Connections 15
- User → Web interface: asks questions
- Web interface → Chat route: sends chat
- Chat route → Agent orchestrator: handles queries
- Agent orchestrator → Retrieval: retrieves evidence
- Agent orchestrator → LLM provider: uses tools
- SQL tools route → SQL agent: runs SQL
- SQL agent → DB connector manager: routes queries
- API tools route → API agent: executes API tools
- API agent → API connector manager: discovers connectors
README.md - API agent → Connected APIs: sends requests
- File agent → Document source manager: searches sources
- Web agent → Web connector manager: uses connectors
- Web agent → Live web sources: accesses live data
- Retrieval → PostgreSQL: uses indexed data
- File agent → Object storage: reads documents