BioKnowdes: Mapping NASA’s Life Sciences Knowledge 🚀

"A new way of exploring space biology."

BioKnowdes is an integrated platform designed to help researchers, students, and decision-makers navigate, understand, and uncover connections within NASA’s vast collection of life sciences research. By combining Knowledge Graphs, Heatmaps, and Generative AI, we transform unstructured PDFs into a dynamic universe of data.

🎥 Watch the Demo Video

---

🌌 Project Overview

BioKnowdes bridges the gap between complex scientific research and accessible knowledge. The platform ingests 600+ open-access Space Biology publications and allows users to discover patterns that might otherwise remain buried in massive amounts of textual data.

Key Features

1. 🕸️ Interactive Knowledge Graph

The core of BioKnowdes. It represents research papers as nodes and shared concepts as connections.

• Pipeline: PDF Text Extraction (Apache Tika) → Keyword Extraction (KeyBERT/SciBERT) → Graph Construction (NetworkX).
• Goal: Visualize how studies relate to one another and identify emerging trends.

2. 🔥 Similarity Heatmap

A high-level view that measures the similarity between top research articles.

• Visual: Interactive grid highlighting clusters of related work.
• Utility: Detect gaps in scientific knowledge and understand research concentrations.

3. 🤖 Lumi - AI Research Agent

Lumi is a conversational agent powered by a RAG (Retrieval-Augmented Generation) system.

• Retrieve: Searches the database for relevant papers using semantic ranking.
• Augment: Uses key content from selected papers as context.
• Generate: Produces clear answers cited directly from NASA sources.
• Tech Note: While designed to work with local models (Llama 3.1 via Ollama), the production version utilizes the OpenAI API for performance and stability.

---

🛠️ Technology Stack

• Backend: Python, Flask

• Data Processing: Apache Tika (text extraction), Pandas, NetworkX

• AI & NLP:

  • LangChain (RAG Workflow)
  • OpenAI API (Production LLM) / Ollama (Local Dev)
  • Sentence-Transformers, KeyBERT, SciBERT

• Frontend: HTML, CSS, JavaScript, Pyvis Network (Graph), Plotly (Heatmap)

---

🚀 Installation & Configuration

1. Prerequisites

• Python 3.9+
• Virtual Environment (recommended)

2. Environment Setup

1. Clone and install dependencies:

   git clone https://github.com/your-username/bioknow-nasa.git
   cd bioknow-nasa
   python -m venv venv
   source venv/bin/activate  # Windows: venv\Scripts\activate
   pip install -r requirements.txt

2. Configure Environment Variables: Copy the template and edit with your keys.

   cp env.template .env

   Required variables in .env:

   # OpenAI API (Mandatory for Lumi Agent)
   OPENAI_API_KEY=sk-your-openai-key-here
   
   # Flask Security
   FLASK_SECRET_KEY=your-secret-key-here
   
   # Optional (for future multi-model support)
   ANTHROPIC_API_KEY=...
   GEMINI_API_KEY=...

3. Running the Application

python app.py

Access the application at: http://localhost:5000/

---

🔧 Environment Variable Configuration

📋 Initial Setup

1. Copy the environment variable template:

   cp env.template .env

2. Set your API keys in the file .env:

   # Edit the .env file with your real keys
   nano .env

3. Mandatory variables for production:

   # OpenAI API (mandatory for full functionality)
   OPENAI_API_KEY=sk-your-openai-key-here
   
   # Flask secret key (mandatory for sessions)
   FLASK_SECRET_KEY=your-secret-key-here
   
   # Other APIs (Optional)
   ANTHROPIC_API_KEY=your-anthropic-key-here
   GEMINI_API_KEY=your-gemini-key-here

🔒 Production Security

• ✅ Never commit the file .env (its already on .gitignore)
• ✅ Use different keys for development and production
• ✅ Rotate your keys periodically
• ✅ Set environment variables on your hosting provider

🚀 Production Deployment

For platforms such as Heroku, Railway, or Vercel:

#Configure the environment variables on the provider panel
OPENAI_API_KEY=sk-your-production-key
FLASK_SECRET_KEY=your-production-secret-key
FLASK_ENV=production

✅ Applied Changes

All of the BioKnowdes routes are now under the prefix /ask-lumi:

📋 Route Mapping

Old Route	New Route	Description
/chat	/ask-lumi	Main chat page
/articles	/ask-lumi/articles	Article browser
/settings	/ask-lumi/settings	AI settings
/upload	/ask-lumi/upload	Document upload
/documents	/ask-lumi/documents	List documents
/documents/toggle/:id	/ask-lumi/documents/toggle/:id	Toggle selection
/documents/delete/:id	/ask-lumi/documents/delete/:id	Delete document
/clear	/ask-lumi/clear	Clear session
/external-demo	/ask-lumi/external-demo	External example

🔌 API Endpoints

Old Route	New Route	Method	Description
/chat (POST)	/ask-lumi/chat	POST	AI Chat
/api/load-articles	/ask-lumi/api/load-articles	POST	Load articles
/api/articles-list	/ask-lumi/api/articles-list	GET	List articles
/api/search-articles	/ask-lumi/api/search-articles	POST	Search articles
/api/settings	/ask-lumi/api/settings	GET/POST	AI Config

🌐 Original Routes (Kept)

These are the original project routes that were not changed:

• / - Home with graph
• /graph - Graph search
• /heatmap - Thermal analysis
• /get_graph_data - Graph data
• /api/keywords - Keywords list

📁 Updated Files

Backend:

• ✅ app.py - All of the BioKnowdes routes altered
• ✅ CORS updated to /ask-lumi/api/*

Templates:

• ✅ templates/index.html - Menu links
• ✅ templates/chat.html - Navigation links
• ✅ templates/articles.html - Link "Back to Chat"
• ✅ templates/settings.html - Link "Back to Chat"

JavaScript:

• ✅ static/js/bioknowdes.js - All fetch calls
• ✅ static/js/articles.js - API calls and redirections
• ✅ static/js/settings.js - API calls and redirections

Outros:

• ✅ external_button_example.html - API URL and redirect

🚀 How to Access

Main Pages:

Home (Graph):        http://localhost:5000/
AI Chat:             http://localhost:5000/ask-lumi
Browse Articles:     http://localhost:5000/ask-lumi/articles
Settings:            http://localhost:5000/ask-lumi/settings
Heatmap:             http://localhost:5000/heatmap

API Endpoints:

# AI Chat
POST http://localhost:5000/ask-lumi/chat
{
  "message": "Your question here"
}

# Load articles
POST http://localhost:5000/ask-lumi/api/load-articles
{
  "filenames": ["article1_text.txt", "article2_text.txt"]
}

# List all articles
GET http://localhost:5000/ask-lumi/api/articles-list

# Settings
GET http://localhost:5000/ask-lumi/api/settings
POST http://localhost:5000/ask-lumi/api/settings
{
  "provider": "openai",
  "openai": {
    "api_key": "...",
    "model": "gpt-4"
  }
}

🧪 Testing

Execute the server:

cd /home/jabs/development/personal/Nasa-Space-Apps_25/bioknow-nasa
source venv/bin/activate
python app.py

Test each route:

# Original home
curl http://localhost:5000/

# Chat (must return HTML)
curl http://localhost:5000/ask-lumi

# Articles API
curl http://localhost:5000/ask-lumi/api/articles-list

📊 Navigation Structure

┌─────────────┐
│    Home     │ (Graph - /)
│   (Graph)   │
└──────┬──────┘
       │
       ├─────────────┐
       │             │
┌──────▼──────┐  ┌──▼────────┐
│   Heatmap   │  │  Ask-Lumi │ (/ask-lumi)
│ (/heatmap)  │  │  (Chat)   │
└─────────────┘  └─────┬─────┘
                       │
                       ├────────────┬──────────────┐
                       │            │              │
                ┌──────▼──────┐ ┌──▼────────┐ ┌──▼─────────┐
                │   Articles  │ │ Settings  │ │  External  │
                │  (/articles)│ │(/settings)│ │   Demo     │
                └─────────────┘ └───────────┘ └────────────┘

🔄 Workflow

1. User acess home (/)

   • Sees graph visualization
   • Dropdown menu with all options

2. Clicks "AI Chat" → Goes to /ask-lumi

   • Can load documents
   • Can go to /ask-lumi/articles or /ask-lumi/settings

3. Loads articles on /ask-lumi/articles

   • Selects articles
   • Clicks "Load Selected"
   • API: POST /ask-lumi/api/load-articles
   • Redirects to /ask-lumi with session_id

4. Asks questions on the chat

   • API: POST /ask-lumi/chat
   • AI answers based on documents

🎯 Benefits of the New Structure

✅ Organization: All BioKnowdes routes grouped under /ask-lumi

✅ Separation: Clear distinction between:

• Graph visualization (original)
• AI and document analysis (BioKnowdes)

✅ Scalability: Each to add new features under /ask-lumi

✅ External API: CORS configured to /ask-lumi/api/*

📝 Important Notes

1. Session ID: Always included in API responses

2. CORS: Allowed only for /ask-lumi/api/*, for security

3. Navigation: All templates have updated links

4. redirects: After loading articles or saving settings, redirects to /ask-lumi

🐛 Troubleshooting

404 Error in routes

# Verify if its accessing with the correct prefix
# WRONG:  /articles
# CORRECT: /ask-lumi/articles

CORS Error

# Make sure its requisition goes to /ask-lumi/api/*
# CORS is configured only for this prefix

Session not persisting

# Verify if the session_id is being passed correctly
# On the URL after loading articles

---

Full Update! 🎉

All routes are now organized under /ask-lumi.