An Agentic AI Research Assistant powered by Retrieval-Augmented Generation (RAG)
MedAgent-X is a full-stack AI application designed to accelerate clinical research. Instead of manually navigating hundreds of pages, researchers can upload a clinical PDF, ask questions in the chatbot, generate knowledge graphs, and export AI-generated presentation slides.
- Upload clinical research papers in PDF format.
- Robust document parsing using PyPDF with defensive error handling.
- Automatically extracts and processes textual content from complex academic papers.
- Ask questions about uploaded research papers.
- Uses a custom Retrieval-Augmented Generation (RAG) pipeline.
- Retrieves the most relevant document chunks before generating responses.
- Explicitly cites the corresponding page number for every answer.
Automatically converts unstructured medical findings into interactive Mermaid.js dependency graphs for better understanding.
Generate a presentation summarizing the uploaded paper, with the option of converting the slides into a .PPTX file.
The application automatically rejects non-medical PDFs using LLM-based classification, ensuring that only domain-specific research papers are accepted.
Instead of relying on external vector databases, MedAgent-X implements an in-memory semantic retrieval engine using NumPy. MedAgent-X includes a defensive extraction pipeline using extensive exception handling that gracefully skips problematic pages without interrupting document processing.
Rather than accepting free-form AI responses, MedAgent-X prompts the LLM to generate structured JSON. This predictable schema is directly mapped into PowerPoint templates using python-pptx, enabling reliable presentation generation.
- Frontend hosted on Vercel
- Backend deployed on Render
- Communication exclusively through REST APIs
- No Google Cloud SDK dependencies
- Raw Gemini REST implementation for lightweight deployment
- Vanilla JavaScript
- Tailwind CSS
- PDF.js
- Mermaid.js
- Python
- FastAPI
- NumPy
- PyPDF
- python-pptx
- Google Gemini 2.5 Flash API
- Retrieval-Augmented Generation (RAG)
- Custom Semantic Search
- Prompt Engineering
- Vercel (Frontend)
- Render (Backend)