How naureenfathima/blockseblock-ai-bootcamp works
This repository is a cumulative bootcamp project for building a domain-specific AI assistant, rather than a single standalone feature. Its principal workflow runs from a user request through chat or smart routing, optional session history and document retrieval, and an LLM response; agent actions, MCP tools, voice, and image analysis extend that workflow. The graph combines sampled endpoint wiring with README-documented capabilities; unsampled feature internals are represented only where the tree supports their paths.
Main entry points: main.py Assistant API · agent.py Agent loop · ingestion.py Document ingestion · factory.py Provider selection
Components 23
User experience
- Web interface
ui/app.js - Assistant API
week-4-launch/feature-11-production-design/solution/main.py - Chat and schemas
shared/models.py
Knowledge and memory
- Session history
shared/session_store.py - Document ingestion
shared/ingestion.py - Document store
shared/document_store.py - Vector search
shared/vector_store.py - Retrieval memory
shared/retrieval_memory.py
AI and actions
- Smart routing
shared/router.py - Agent loop
shared/agent.py - Task planner
shared/planner.py - Local tools
shared/tools.py - MCP client
shared/mcp_client.py - Domain MCP tools
shared/domain_mcp_server.py - Agent task state
shared/task_store.py - LLM interface
shared/llm_client.py - Provider selection
shared/providers/factory.py - Model providers
shared/providers - Voice and vision
shared/llm_client.py
Operations
- Observability and limits
shared/middleware.py - Metrics
shared/metrics.py
Other components
- User
- Model services
Connections 25
- User → Web interface: uses
- Web interface → Assistant API: sends requests
- Assistant API → Chat and schemas: validates
- Assistant API → Session history: reads and writes
- Assistant API → Document ingestion: ingests uploads
- Document ingestion → Document store: stores documents
- Document ingestion → Vector search: indexes chunks
- Assistant API → Vector search: searches
- Assistant API → Smart routing: classifies queries
- Smart routing → Vector search: selects retrieval
- Assistant API → Retrieval memory: logs retrieval
- Assistant API → Agent loop: runs tasks
- Agent loop → Task planner: plans steps
- Agent loop → Local tools: calls tools
- Agent loop → MCP client: uses remote tools
- MCP client → Domain MCP tools: dispatches tools
- Assistant API → Agent task state: reads task status
- Assistant API → LLM interface: requests completion
- LLM interface → Provider selection: selects provider
- Provider selection → Model providers: creates provider
- Model providers → Model services: calls model
- Assistant API → Voice and vision: handles voice and images
- Voice and vision → Provider selection: selects modality provider
- Assistant API → Observability and limits: applies safeguards
- Assistant API → Metrics: reports metrics