docs/archive/nebius-setup.md

Nebius Configuration Guide

🚀 Quick Setup

Your app is now configured to use Nebius AI instead of OpenAI! Here's how to get it running:

1. Set Environment Variables

Create a .env file in the root directory:

# Required: Nebius AI API Key
NEBIUS_API_KEY=your_nebius_api_key_here

# Optional: Modal for advanced processing
MODAL_TOKEN_ID=your_modal_token_id
MODAL_TOKEN_SECRET=your_modal_token_secret

# Optional: GitHub for repository search
GITHUB_TOKEN=your_github_token_here

2. Supported Models

The app now uses these Nebius models:

🧠 Default Model: DeepSeek-R1-0528

• Model ID: deepseek-ai/DeepSeek-R1-0528
• Use Cases: Chat completions, document analysis, query enhancement
• Features: Advanced reasoning, code understanding, research synthesis

🌟 Alternative Models Available:

• Qwen: Qwen/Qwen3-235B-A22B
• Llama Nemotron: nvidia/Llama-3_1-Nemotron-Ultra-253B-v1

3. Start the Application

npm run dev

The app will start at http://localhost:5000

🔧 What Changed

✅ Removed OpenAI Dependencies

• ❌ OPENAI_API_KEY no longer required
• ❌ openai package dependency removed from routes
• ✅ All AI operations now use Nebius

✅ Enhanced AI Capabilities

• Document Analysis: Summary, classification, key points, quality scoring
• Query Enhancement: Smart query improvement with intent analysis
• Research Synthesis: Multi-document analysis and insights
• Embeddings: Vector generation for semantic search
• Chat Completions: Conversational AI for explanations

✅ API Endpoints Updated

• /api/explain - Uses DeepSeek for document explanations
• /api/ai-search - Enhanced AI-powered search
• /api/analyze-document - Document analysis with DeepSeek
• /api/enhance-query - Query improvement
• /api/embeddings - Vector embeddings via Nebius
• /api/health - Health check includes Nebius status

🎯 Features Now Available

AI-Enhanced Search Interface

• Smart query enhancement with purple sparkles ✨
• Real-time AI suggestions and improvements
• Dual search modes: Traditional + AI-enhanced

Document Analysis Tools

• Summary: Concise document summaries
• Classification: Automatic document categorization
• Key Points: Extract important findings
• Quality Score: Assess document credibility

Advanced AI Features

• Research Synthesis: Analyze multiple documents together
• Embedding Generation: Create vector representations
• Relevance Scoring: AI-powered result ranking

🔍 Model Selection

When to Use Each Model:

DeepSeek-R1-0528 (Default)

• General-purpose tasks
• Code analysis and programming
• Research and academic content
• Reasoning-heavy tasks

Qwen3-235B-A22B

• Multilingual content
• Large-scale document processing
• Complex reasoning tasks

Llama-3_1-Nemotron-Ultra-253B-v1

• Creative writing and content generation
• Instruction following
• Advanced conversational AI

🚨 Troubleshooting

API Key Issues

# Check if your Nebius API key is working
curl -H "Authorization: Bearer YOUR_API_KEY" \
     -H "Content-Type: application/json" \
     https://api.studio.nebius.ai/v1/models

Health Check

Visit http://localhost:5000/api/health to verify all services are working.

Common Issues

1. Missing API Key: Ensure NEBIUS_API_KEY is set in your .env file
2. Rate Limits: Nebius may have rate limits - check your usage
3. Model Names: Use exact model IDs as shown above

📊 Cost Benefits

• Lower Costs: Nebius often provides better pricing than OpenAI
• Better Performance: DeepSeek models are optimized for reasoning tasks
• More Control: Choose specific models for different use cases

Your KnowledgeBridge app is now powered by cutting-edge Nebius AI models! 🎉