Akshanshsensei/PDF-Constrained-Conversational-Agent

PDF-Constrained Conversational Agent

This is a production-grade Conversational AI agent designed to answer user queries strictly based on a provided PDF document.

Features

• —Strict Grounding: The agent will only answer questions if the answer exists in the uploaded PDF.
• —Agentic Tool-Use: The LLM can autonomously invoke Python code to calculate exact word counts and extract deterministic statistics without hallucinating.
• —Anti-Hallucination: It explicitly refuses to answer out-of-scope queries.
• —Citations: Generates precise [Source: Page X] citations for every claim.
• —Recursive Memory: Handles long conversations without losing context.
• —Multilingual Support: Capable of cross-lingual RAG using state-of-the-art embedding models.

Usage

Upload a PDF on the left panel, wait for the index to build, and start chatting!