🔬 ResearchAI — AI Research Assistant

Upload papers, ask questions, get summaries, extract contributions, and find related work — all powered by RAG (Retrieval-Augmented Generation).

---

Project Overview

ResearchAI is a Retrieval-Augmented Generation (RAG) system specialized for scientific research. It allows users to build a personal library of academic papers from multiple sources (PDF, ArXiv, DOI, plain text), and then interact with that library using natural language — asking questions, getting summaries, extracting contributions, and finding related work.

🎯 Core Problem Solved Traditional LLMs hallucinate facts and have outdated knowledge. ResearchAI grounds every answer in your actual uploaded papers, providing citations and relevance scores with every response.

✨ Features

Feature	Description
📄 PDF Upload	Drag & drop any research paper
🌐 ArXiv & DOI	Fetch papers directly by ID, URL, or DOI
📝 Plain Text	Paste any abstract or raw text
📊 Summarize	Structured summary: objective / methods / results / conclusion
🔬 Contributions	Extract novel contributions, limitations, and future work
🔗 Related Work	Find connected topics, cited papers, and research gaps
💬 Q&A Chat	Ask anything across your entire library or a single paper
⚡ Streaming	Real-time token-by-token responses via SSE

---

🛠️ Tech Stack

Layer	Technology	Role
Backend	FastAPI	HTTP framework, async, auto-docs
AI / LLM	Gemini 1.5 Flash via LangChain	Answer generation
Embeddings	Google embedding-001	Text → vector conversion
Vector DB	ChromaDB	Local persistent similarity search
PDF Parsing	PyMuPDF (fitz)	Fast PDF text extraction
ArXiv	arxiv Python library	Fetch papers by ID or URL
DOI	CrossRef REST API (free)	Fetch metadata by DOI
Frontend	Vanilla HTML / CSS / JS	Single-page UI served by FastAPI

---

📁 Project Structure

research-ai/
├── app/
│   ├── api/
│   │   ├── papers.py             # Ingest endpoints (PDF, ArXiv, DOI, text)
│   │   └── chat.py               # AI endpoints (summarize, Q&A, contributions)
│   ├── core/
│   │   └── config.py             # Settings loaded from .env
│   ├── services/
│   │   ├── ai/
│   │   │   ├── llm_provider.py   # LLM interface + Gemini implementation
│   │   │   └── rag_service.py    # RAG orchestration (retrieve → generate)
│   │   ├── ingestors/
│   │   │   ├── pdf_ingestor.py   # PyMuPDF parser
│   │   │   ├── arxiv_ingestor.py # ArXiv API client
│   │   │   ├── doi_ingestor.py   # CrossRef API client
│   │   │   └── text_ingestor.py  # Plain text handler
│   │   └── knowledge/
│   │       └── vector_store.py   # ChromaDB wrapper
│   └── main.py                   # FastAPI app + static file serving
├── static/
│   └── index.html                # Web UI (no separate server needed)
├── tests/
│   └── test_research_ai.py       # Unit tests (fully mocked)
├── .env.example                  # Environment variable template
├── .gitignore
└── pyproject.toml

---

🚀 Getting Started

1. Clone the repo

git clone https://github.com/YounesBensafia/research-ai.git
cd research-ai

2. Install dependencies

pip install fastapi uvicorn python-multipart langchain langchain-google-genai \
    langchain-chroma langchain-text-splitters chromadb google-generativeai \
    pydantic-settings python-dotenv pymupdf arxiv httpx pytest pytest-asyncio

3. Configure environment

cp .env.example .env

Open .env and set your Gemini API key:

GEMINI_API_KEY=your_key_here

🔑 Get a free key at → https://aistudio.google.com/app/apikey

4. Run the server

python -m uvicorn app.main:app --reload

Open → http://localhost:8000
Swagger docs → http://localhost:8000/docs

---

🧠 How RAG Works

User Question
     │
     ▼
Embed question → vector
     │
     ▼
Similarity search in ChromaDB (Top-K chunks)
     │
     ▼
Build context from retrieved chunks
     │
     ▼
Inject context into prompt template
     │
     ▼
Gemini generates grounded answer
     │
     ▼
Response + source citations returned

Every answer is grounded in your actual uploaded papers — no hallucinations, full citations.

---

📡 API Reference

Papers — Ingest

Method	Endpoint	Description
POST	/api/papers/upload	Upload a PDF file
POST	/api/papers/arxiv	Ingest from ArXiv URL or ID
POST	/api/papers/doi	Ingest from DOI
POST	/api/papers/text	Ingest plain text or abstract
GET	/api/papers/	List all papers + stats
DELETE	/api/papers/	Remove a paper by source

Chat — AI Features

Method	Endpoint	Description
POST	/api/chat/summarize	Structured paper summary
POST	/api/chat/contributions	Contributions & limitations
POST	/api/chat/related	Related work & topics
POST	/api/chat/ask	Free Q&A (full response)
POST	/api/chat/ask/stream	Streaming Q&A via SSE

Example

# Ingest the Attention paper from ArXiv
curl -X POST http://localhost:8000/api/papers/arxiv \
  -H "Content-Type: application/json" \
  -d '{ "url_or_id": "1706.03762" }'

# Ask a question
curl -X POST http://localhost:8000/api/chat/ask \
  -H "Content-Type: application/json" \
  -d '{ "question": "What is the main contribution of this paper?", "top_k": 5 }'

---

📄 License

MIT — free to use and adapt.

---