BioTrek

An AI-powered knowledge engine for NASA space biology research. Ask questions about 600+ research papers, get answers with auto-generated charts, and browse an insights dashboard — all through a conversational interface backed by RAG.

Built for the Build a Space Biology Knowledge Engine challenge.

Showcase

---

Features

• Conversational RAG chatbot — ask natural language questions; answers are grounded in 600+ ingested NASA papers with source citations
• Automatic chart rendering — detects chartable data in AI responses (JSON blocks, Markdown tables, CSV, bullet lists, [CHART:type] directives) and renders Line, Bar, Pie, Area, or Scatter charts automatically
• Source citations panel — slide-out sidebar with matched documents, content snippets, and direct links to original PDFs/HTML files
• Chat history — create, rename, load, and delete past conversations with full message persistence per user
• Insights dashboard — publication statistics with interactive Recharts visualizations across 9 research domains
• Off-topic gating — dual-signal filtering (vector distance threshold + LLM refusal pattern matching) suppresses hallucinated sources on irrelevant queries
• Auth — JWT + refresh token rotation, proactive silent refresh 2 min before expiry, 401-interceptor fallback, role-based access control
• Dark/light mode — theme toggle persisted to localStorage

---

Architecture

Single Node.js process — no separate Python service. The entire RAG pipeline runs in TypeScript via LangChain.js and external APIs.

---

Tech Stack

Backend

Layer	Technology
Framework	NestJS 11 + TypeScript
ORM	TypeORM → PostgreSQL
Auth	JWT + refresh token rotation (Passport)
Embeddings	HuggingFace Inference API (BAAI/bge-small-en-v1.5)
LLM	Groq API (llama-3.3-70b-versatile, temp 0.3)
Vector store	ChromaDB via LangChain.js
RAG	@langchain/core · @langchain/community · @langchain/groq
Doc parsing	pdf-parse (PDF)
API docs	Swagger / OpenAPI at /docs

Frontend

Layer	Technology
Framework	React 19 + TypeScript
Build	Vite 7
Styling	Tailwind CSS + shadcn/ui + Radix UI
Charts	Recharts 3
3D / Shaders	Three.js (custom GLSL fragment shader)
Markdown	react-markdown + rehype-highlight
Routing	React Router 7
Toasts	Sonner

---

Setup

Prerequisites

• Node.js ≥ 18
• PostgreSQL ≥ 14
• ChromaDB (pip install chromadb) + Python 3.9+
• Groq API key (free tier available)

1. Clone

git clone <repo-url>
cd BioTrek

2. PostgreSQL

Linux:

sudo apt update && sudo apt install postgresql postgresql-contrib
sudo systemctl start postgresql && sudo systemctl enable postgresql
sudo -u postgres psql -c "CREATE DATABASE biotrek;"

macOS:

brew install postgresql@16 && brew services start postgresql@16
createdb biotrek

Windows: Download from postgresql.org, create a database named biotrek owned by the default postgres user.

3. ChromaDB

pip install chromadb
chroma run --path ./back-end/data/database   # keep running in a separate terminal

4. Backend

cd back-end
cp env-example-relational .env

update the .env file with your PostgreSQL credentials, Groq API key, and any other overrides, then run:

npm install
npm run migration:run
npm run seed:run:relational
npm run start:dev

• API: http://localhost:3000/api/v1
• Swagger: http://localhost:3000/docs

First run: if no ChromaDB collection is found, the vector store builds automatically from the documents in CHATBOT_DATA_PATH. This calls the HuggingFace API to embed all chunks and may take a few minutes.

5. Frontend

cd front-end
npm install
npm run dev

App: http://localhost:5173

To override the API base URL:

# front-end/.env
VITE_API_URL=http://localhost:3000/api/v1

---

Scripts

Backend (back-end/)

Script	Description
npm run start:dev	Dev server with hot-reload
npm run build	Compile TypeScript to dist/
npm run start:prod	Run compiled output
npm run migration:generate	Generate a new TypeORM migration
npm run migration:run	Apply pending migrations
npm run migration:revert	Revert last migration
npm run seed:run:relational	Seed the database
npm run lint	ESLint

Frontend (front-end/)

Script	Description
npm run dev	Dev server with HMR
npm run build	Production build (outputs to dist/)
npm run preview	Preview production build locally
npm run lint	ESLint

---

Thanks for checking out BioTrek! Feel free to explore the code, run it locally, and reach out if you have any questions or feedback.