How dhinakaran311/legal_research_assistant works
The product is an Indian-law research assistant: users search or chat in the web app, requests pass through the backend to the AI engine, and results combine legal-document retrieval, graph references, and answer synthesis. The documented data-ingestion branch collects and loads legal material into the retrieval stores. Source excerpts show the query pipeline and its agent stages; several broader capabilities, including adaptive RAG and GraphQL, are documented but their runtime internals were not sampled.
Main entry points: agentic_pipeline.py Agentic pipeline · index.js Backend API · pipeline_factory.py Pipeline factory · chat.py Chat route
Components 27
Web experience
- Search and chat
frontend/app/search/page.tsx - Chat interface
frontend/app/chat/page.tsx
Backend services
- Backend API
backend/src/index.js - Authentication
backend/src/api/auth.js - Chat service
backend/src/services/chatService.js - AI engine client
backend/src/services/aiEngineClient.js - User database
backend/prisma/seed.js
AI research
- AI API
ai_engine/src/main.py - Query routes
ai_engine/src/routes/query.py - Adaptive RAG
ai_engine/src/pipelines/adaptive_rag.py - Chat route
ai_engine/src/routes/chat.py - Pipeline factory
ai_engine/src/pipeline_factory.py - Agentic pipeline
ai_engine/src/agents/agentic_pipeline.py - Query planning
ai_engine/src/agents/planner_agent.py - Local research
ai_engine/src/agents/local_research_agent.py - Web research
ai_engine/src/agents/web_research_agent.py - Conflict checking
ai_engine/src/agents/conflict_checker_agent.py - Answer synthesis
ai_engine/src/agents/synthesis_agent.py - Conversation memory
ai_engine/src/conversation/memory.py - Query resolver
ai_engine/src/conversation/query_resolver.py
Legal knowledge
- Vector store
ai_engine/src/vectorstore/chroma_client.py - Legal knowledge graph
ai_engine/src/graph/neo4j_client.py
Data ingestion
- Legal sources
data_ingestion/sources/multi_act_scraper.py - Vector data loader
data_ingestion/loaders/load_multi_act_data.py - Graph data loader
data_ingestion/loaders/load_multi_act_to_neo4j.py
Other components
- Researcher
- Legal websites
Connections 31
- Researcher → Search and chat: searches
- Researcher → Chat interface: asks questions
- Search and chat → Backend API: sends requests
- Chat interface → Backend API: sends requests
- Backend API → Authentication: authenticates
- Backend API → Chat service: handles chat
- Chat service → AI engine client: requests answers
- Backend API → User database: stores user data
- Backend API → AI API: calls over HTTP
- AI API → Query routes: registers routes
ai_engine/src/main.py - AI API → Chat route: registers routes
ai_engine/src/main.py - Query routes → Pipeline factory: gets pipeline
ai_engine/src/routes/query.py - Query routes → Adaptive RAG: offers adaptive search
- Chat route → Pipeline factory: gets pipeline
ai_engine/src/routes/chat.py - Chat route → Conversation memory: uses history
ai_engine/src/routes/chat.py - Chat route → Query resolver: resolves references
ai_engine/src/routes/chat.py - Query resolver → Conversation memory: reads turns
ai_engine/src/conversation/query_resolver.py - Pipeline factory → Agentic pipeline: creates pipeline
ai_engine/src/pipeline_factory.py - Pipeline factory → Vector store: creates client
ai_engine/src/pipeline_factory.py - Pipeline factory → Legal knowledge graph: creates client
ai_engine/src/pipeline_factory.py - Agentic pipeline → Query planning: plans query
ai_engine/src/agents/agentic_pipeline.py - Agentic pipeline → Local research: retrieves local evidence
ai_engine/src/agents/agentic_pipeline.py - Agentic pipeline → Web research: escalates research
ai_engine/src/agents/agentic_pipeline.py - Agentic pipeline → Conflict checking: checks conflicts
ai_engine/src/agents/agentic_pipeline.py - Agentic pipeline → Answer synthesis: synthesizes answer
ai_engine/src/agents/agentic_pipeline.py - Local research → Vector store: searches documents
ai_engine/src/agents/local_research_agent.py - Local research → Legal knowledge graph: retrieves graph facts
- Web research → Legal websites: fetches legal material
ai_engine/src/agents/web_research_agent.py - Legal sources → Legal websites: collects documents
- Vector data loader → Vector store: loads documents
- Graph data loader → Legal knowledge graph: loads relationships
data_ingestion/loaders/load_multi_act_to_neo4j.py