Turn research papers into
cited answers in seconds

The Research Assistant reads your PDFs, understands them with vector search, and answers your questions with real citations β€” no more scrolling through 30 pages to find one fact.

0
Dim vectors
AI capabilities
0
Open source
0
Avg answer time
Scroll
⚑ FastAPI 🧠 Groq LLaMA 3.1 πŸ”Ž Qdrant Vector DB πŸ”— LangChain Agents πŸ“ Sentence Transformers 🐳 Docker πŸ€— HuggingFace Spaces ⚑ FastAPI 🧠 Groq LLaMA 3.1 πŸ”Ž Qdrant Vector DB πŸ”— LangChain Agents πŸ“ Sentence Transformers 🐳 Docker πŸ€— HuggingFace Spaces
✦ Capabilities

Everything you need to
read less, learn more

Five AI-powered tools built on a production RAG pipeline β€” from instant Q&A to full literature reviews.

Cited Q&A

Ask any question about your papers and get answers grounded in the source text, with page-level citations you can verify.

Smart Summaries

Generate brief, comprehensive, or technical summaries with extracted key findings, methodology and limitations.

Literature Reviews

Synthesize a whole topic across papers β€” key themes, research gaps and future directions, generated automatically.

ReAct Agent

An intent-routing LangChain agent picks the right tool for each request and chains steps to reason through complex asks.

Drag & Drop Upload

Drop a PDF and watch it get parsed, chunked, embedded and indexed into vector storage in real time.

Hybrid Retrieval

Semantic + keyword search over 384-dim embeddings finds the most relevant passages, not just keyword matches.

✦ Pipeline

From PDF to answer in four steps

A transparent RAG pipeline β€” no black boxes.

PDF
attention_is_all_you_need.pdf
βœ“
Extracting text Β· metadata Β· sections
πŸ“„
➜
Chunked & embedded into 384-dim vectors
πŸ”Ž β€œWhat is the attention mechanism?”
Hybrid search pulls the most relevant passages
πŸ“„ Attention Is All You Need
LLaMA 3.1 composes a grounded, cited answer
✦ In action

Answers you can
actually trust

Every response is backed by the exact passages it came from. Click a citation to jump straight to the source β€” perfect for literature reviews, thesis work and staying honest with your references.

  • Inline citations with relevance scores
  • Grounded in your documents β€” no hallucinated facts
  • Sub-second retrieval, ~2s end-to-end answers
Open the app β†’
research-assistant/chat
What is the attention mechanism?
Attention lets a model weigh the relevance of every input token when producing each output, replacing recurrence with direct token-to-token connections for better long-range context. πŸ“„ Attention Is All You Need Β· p.3 Β· 0.94
How is it computed?
✦ Built with

A modern, production stack

Async Python backend, managed vector storage and blazing-fast inference.

⚑
FastAPI
Async REST API
🧠
Groq
LLaMA 3.1 inference
πŸ”Ž
Qdrant
Vector search
πŸ”—
LangChain
ReAct agent
πŸ“
MiniLM
384-dim embeddings
🐳
Docker
Containerized
πŸ€—
HF Spaces
Hosting
πŸ“„
PyMuPDF
PDF extraction
✦ Questions

Frequently asked

Yes. The project is fully open source and the free to use.
Currently PDF research papers up to 50MB.
Retrieval-Augmented Generation grounds every answer in retrieved passages from your own documents, and each answer ships with citations so you can verify the source.

Stop skimming.
Start asking.

Upload your first paper and get a cited answer in under two seconds.