How yogendrachukka01/adaptiveagent works
AdaptiveAgent is a private-knowledge-base Agentic RAG platform. A user submits a chat query or uploads a document; chat requests enter the LangGraph workflow, retrieve evidence, generate and evaluate an answer, then stream results and review metadata to the frontend. The graph preserves major documented stages and supporting API, retrieval, LLM, and memory capabilities. Most individual node and service implementations were not sampled, so their internal calls are shown only where the README or source index supports them.
Main entry points: builder.py Graph builder · chat.py Chat API · api.ts API client · page.tsx Chat interface
Components 26
Web client
- Chat interface
frontend/src/app/page.tsx - Message display
frontend/src/components/chat/MessageBubble.tsx - API client
frontend/src/lib/api.ts
API and state
- Chat API
backend/app/api/chat.py - Upload API
backend/app/api/upload.py - API key auth
backend/app/core/auth.py - Runtime config
backend/app/core/config.py - API dependencies
backend/app/core/deps.py - Thread state
backend/app/core/threads.py - Agent state
backend/app/models/state.py
Agent workflow
- Graph builder
backend/app/graph/builder.py - Conditional routing
backend/app/graph/edges/conditional.py - Workflow stages
backend/app/graph/nodes - Retrieval stage
backend/app/graph/nodes/retrieval_node.py - Response stage
backend/app/graph/nodes/response_node.py - Evaluation stage
backend/app/graph/nodes/eval_node.py - Refine stage
backend/app/graph/nodes/refine_node.py
AI and retrieval
- Query validation
backend/app/services/validator/validator.py - Hybrid search
backend/app/services/retrieval/hybrid_search.py - LLM providers
backend/app/services/llm.py - Conversation memory
backend/app/services/memory/memory.py - Memory distiller
backend/app/services/memory/memory_worker.py - Vector stores
backend/app/services/retrieval/vector_store
Other components
- User
- Streamed answer
- Knowledge document
Connections 30
- User → Chat interface: asks query
- Chat interface → API client: requests stream
frontend/src/app/page.tsx - API client → Chat API: sends SSE request
- Chat API → Graph builder: invokes graph
backend/app/api/chat.py - Graph builder → Workflow stages: registers stages
backend/app/graph/builder.py - Graph builder → Conditional routing: configures routing
backend/app/graph/builder.py - Graph builder → Retrieval stage: registers retrieval
backend/app/graph/builder.py - Graph builder → Response stage: registers response
backend/app/graph/builder.py - Graph builder → Evaluation stage: registers evaluation
backend/app/graph/builder.py - Graph builder → Refine stage: registers refinement
backend/app/graph/builder.py - Workflow stages → Query validation: validates query
- Retrieval stage → Hybrid search: searches evidence
backend/app/graph/nodes/retrieval_node.py - Retrieval stage → Vector stores: retrieves candidates
- Workflow stages → LLM providers: uses providers
- Response stage → Evaluation stage: evaluates output
backend/app/graph/builder.py - Evaluation stage → Refine stage: routes low scores
backend/app/graph/builder.py - Refine stage → Retrieval stage: retries retrieval
backend/app/graph/builder.py - Chat API → Streamed answer: streams response
- Streamed answer → API client: delivers SSE
- API client → Chat interface: yields events
frontend/src/app/page.tsx - Chat interface → Message display: renders messages
frontend/src/app/page.tsx - User → Knowledge document: uploads document
- Chat interface → API client: uploads file
frontend/src/app/page.tsx - API client → Upload API: posts file
- Chat API → API key auth: checks API key
backend/app/api/chat.py - Chat API → Thread state: uses thread state
backend/app/api/chat.py - Chat API → Conversation memory: stores conversation
backend/app/api/chat.py - Memory distiller → Conversation memory: reads conversations
backend/app/services/memory/memory_worker.py - Memory distiller → LLM providers: extracts facts
backend/app/services/memory/memory_worker.py - Memory distiller → Vector stores: stores distilled facts