🧠 What it does: A user gives a natural language question like: “Compare the 2024 Indian election results with 2019 and summarize key policy differences.”
This app will: Retrieve relevant documents (e.g., Wikipedia, News, PDFs) Use Prompting to clarify user intent if needed Generate a structured output: bullet points, tables, summaries Use Function Calling to: -Call web search APIs (or a local DB) -Summarize PDFs -Plot charts if needed
🧱 Tech Stack Breakdown 💡 Prompting: Smart prompt templates: “You are a research expert. Answer in a structured JSON format.” "Ask clarification if the user’s query is ambiguous."
🔍 RAG (Retrieval-Augmented Generation): Use LangChain or LlamaIndex Load documents (PDFs, websites, YouTube transcripts) Vector DB: FAISS or Pinecone
📦 Structured Output: Ask LLM to return: json Copy Edit { "summary": "...", "key_points": ["...", "..."], "source_links": ["..."] } Display in clean UI with sections
🛠️ Function Calling (Tool Use):
Use OpenAI's function_call or LangChain tools Sample functions: -search_web(query) -load_pdf(url) -generate_chart(data_type, range)
📱 User Interface (React frontend): -Query input -Response section (summary, key points, table) -Source viewer