How palakkumari19/multi-agent-research-assistant works
The product is a Streamlit research assistant. A user submits a query; the app retrieves similar in-memory research, invokes the LangGraph workflow, displays the plan, findings, critique, and report, and saves the result to MongoDB and FAISS. The workflow explicitly dispatches Planner → Researcher → Critic → Writer. The README also documents history management and PDF export. The graph follows shown calls and documented capabilities; unsampled implementation details are not inferred.
Main entry points: app.py Streamlit App · workflow.py LangGraph Workflow · llm.py LLM Client · mongo_store.py MongoDB Store
Components 18
User Interface
- Streamlit App
app.py - PDF Generator
utils/pdf_generator.py
Workflow Orchestration
- LangGraph Workflow
graph/workflow.py - Research State
graph/state.py
Research Agents
- Planner Agent
agents/planner.py - Researcher Agent
agents/researcher.py - Critic Agent
agents/critic.py - Writer Agent
agents/writer.py - Web Search
utils/search.py
AI and Search Services
- LLM Client
utils/llm.py - Groq API
- Tavily Search
- Secret Configuration
utils/config.py
Research Memory
- MongoDB Store
memory/mongo_store.py - MongoDB Atlas
- FAISS Memory
memory/faiss_store.py - Embedding Model
utils/embeddings.py
Other components
- Research User
Connections 23
- Research User → Streamlit App: submits query
app.py - Streamlit App → LangGraph Workflow: invokes
app.py - LangGraph Workflow → Research State: uses state
graph/workflow.py - LangGraph Workflow → Planner Agent: dispatches
graph/workflow.py - LangGraph Workflow → Researcher Agent: dispatches
graph/workflow.py - LangGraph Workflow → Critic Agent: dispatches
graph/workflow.py - LangGraph Workflow → Writer Agent: dispatches
graph/workflow.py - Planner Agent → LLM Client: invokes
agents/planner.py - Researcher Agent → Web Search: searches questions
agents/researcher.py - Web Search → Tavily Search: queries
utils/search.py - Critic Agent → LLM Client: invokes
agents/critic.py - Writer Agent → LLM Client: invokes
agents/writer.py - LLM Client → Groq API: runs inference
utils/llm.py - LLM Client → Secret Configuration: gets API key
utils/llm.py - Web Search → Secret Configuration: gets API key
utils/search.py - MongoDB Store → Secret Configuration: gets connection URI
memory/mongo_store.py - Streamlit App → FAISS Memory: retrieves similar
app.py - Streamlit App → FAISS Memory: stores research
app.py - FAISS Memory → Embedding Model: embeds queries
memory/faiss_store.py - Streamlit App → MongoDB Store: loads and deletes history
app.py - Streamlit App → MongoDB Store: saves research
app.py - MongoDB Store → MongoDB Atlas: reads and writes
memory/mongo_store.py - Streamlit App → PDF Generator: generates report PDF
app.py