Building Your Own AI Search Engine: Vibe Coding Guide

"You are SearchArchitectGPT. Help me build a minimal AI search engine inspired by Perplexica.
 
PROJECT REQUIREMENTS:
- Next.js 14 with TypeScript and Tailwind CSS
- LangChain for AI orchestration
- OpenAI for LLM and embeddings
- SearXNG for meta-search (Docker)
- Streaming responses with citations
- Semantic result reranking
 
ARCHITECTURE GOALS:
1. Query processing with context awareness
2. SearXNG integration for web search
3. Embedding-based result reranking
4. Streaming answer generation with citations
5. Simple, clean UI with real-time updates
 
TASKS:
1. Set up Next.js project with required dependencies
2. Create docker-compose.yml for SearXNG
3. Build basic project structure:
   - /app/api/search/route.ts (main search endpoint)
   - /lib/search.ts (search logic)
   - /lib/rerank.ts (semantic reranking)
   - /components/SearchInterface.tsx (UI)
   - /components/StreamingAnswer.tsx (real-time display)
 
OUTPUT: Complete file structure and package.json with exact dependencies needed."

"You are SearchLogicGPT. Build the core search functionality for my AI search engine.
 
CONTEXT: Next.js app with LangChain, OpenAI, and SearXNG running on localhost:4000
 
REQUIREMENTS:
1. Query processing chain that:
   - Analyzes conversation history for context
   - Determines if web search is needed (skip for greetings)
   - Optimizes natural language queries for search engines
   - Handles both general queries and direct URL processing
 
2. SearXNG integration that:
   - Sends optimized queries to localhost:4000/search
   - Handles different search engines (web, academic, images)
   - Processes results into clean, structured format
   - Extracts titles, URLs, and content snippets
 
3. Basic semantic reranking that:
   - Converts query and results to embeddings
   - Calculates cosine similarity scores
   - Filters results below 0.3 relevance threshold
   - Returns top 10 most relevant results
 
IMPLEMENTATION FILES NEEDED:
- /lib/search.ts (main search orchestration)
- /lib/searxng.ts (SearXNG API client)
- /lib/rerank.ts (embedding and similarity logic)
- /app/api/search/route.ts (API endpoint)
 
OUTPUT: Complete TypeScript implementations with proper error handling and types."

"You are AnswerStreamGPT. Implement streaming AI answer generation with citations.
 
CONTEXT: Search results are ready, now need to generate streaming answers with source citations.
 
REQUIREMENTS:
1. LangChain streaming setup:
   - Create chat prompt template with system instructions
   - Include search results as numbered context
   - Stream responses token by token to frontend
   - Handle streaming events and error states
 
2. Citation system:
   - Format search results as numbered sources [1], [2], [3]
   - Instruct AI to cite every statement with numbers
   - Track which sources are used in the response
   - Display clickable source links below answers
 
3. Response quality controls:
   - Professional, journalistic tone
   - Clear structure with headings
   - Comprehensive coverage of the topic
   - Conclusion that synthesizes information
 
IMPLEMENTATION FILES:
- /lib/generate.ts (streaming answer generation)
- /app/api/chat/route.ts (streaming endpoint)
- /components/StreamingAnswer.tsx (real-time UI)
- /components/SourceCitations.tsx (citation display)
 
PROMPT TEMPLATE REQUIREMENTS:
- Role: Expert research assistant
- Format: Markdown with headings
- Citations: [number] notation for every fact
- Length: Comprehensive but not excessive
- Sources: Always show numbered source list
 
OUTPUT: Complete streaming implementation with citation management."

"You are FeatureExpansionGPT. Add advanced features to make this AI search engine production-ready.
 
CURRENT STATE: Basic search with streaming answers and citations working.
 
ADVANCED FEATURES TO ADD:
 
1. FOCUS MODES SYSTEM:
   - Web Search: General queries with multiple engines
   - Academic Search: ArXiv, Google Scholar, PubMed only
   - YouTube Search: Video content with engagement filtering
   - Reddit Search: Community discussions with upvote weighting
 
2. FILE UPLOAD INTEGRATION:
   - PDF and DOCX document processing
   - Text extraction and intelligent chunking
   - Embedding generation for uploaded content
   - Hybrid search (web + documents) with unified ranking
 
3. CONVERSATION PERSISTENCE:
   - SQLite database for chat history
   - Conversation-aware query processing
   - Smart follow-up question suggestions
   - File association with conversations
 
4. OPTIMIZATION MODES:
   - Speed: Skip reranking for fast responses
   - Balanced: Full semantic reranking (default)
   - Quality: Enhanced processing with multiple signals
 
IMPLEMENTATION PRIORITIES:
1. Focus modes configuration system
2. File upload API with embedding processing
3. Database schema and conversation management
4. UI components for file upload and mode selection
 
OUTPUT: Implementation plan with specific file modifications and new components needed."

"Help me create a specialized AI search engine for [YOUR DOMAIN]:
 
DOMAIN FOCUS: [Legal Research / Medical Information / Financial Analysis / Academic Research]
 
CUSTOMIZATIONS NEEDED:
1. Domain-specific search sources and databases
2. Specialized query processing for domain terminology
3. Custom relevance ranking for domain-specific quality signals
4. Professional citation formats for the domain
5. Domain-specific UI elements and branding
 
EXAMPLE DOMAINS:
- Legal: Westlaw, Google Scholar, court databases
- Medical: PubMed, clinical guidelines, drug databases  
- Financial: SEC filings, financial news, analyst reports
- Academic: ArXiv, Google Scholar, institutional repositories
 
OUTPUT: Complete customization plan with specific implementations."

"Create an enterprise AI search engine that combines:
 
1. Internal company documents and wikis
2. External web search for industry information
3. Team collaboration and shared knowledge
4. Access controls and permission management
5. Audit trails and compliance features
 
ENTERPRISE FEATURES:
- Single sign-on (SSO) integration
- Role-based access to different document sets
- Compliance logging for sensitive searches
- Team workspaces with shared conversations
- Analytics and usage reporting
 
OUTPUT: Enterprise-grade architecture with security and compliance considerations."

"Deploy a production-ready AI search engine for enterprise use:
 
INFRASTRUCTURE REQUIREMENTS:
- High-availability setup with load balancers
- Secure API key management with vault systems
- Monitoring and alerting for all components
- Backup and disaster recovery procedures
- Compliance logging and audit trails
 
SCALING ARCHITECTURE:
- Microservices separation (search, embedding, generation)
- Horizontal scaling for high query volumes
- Database optimization for large document collections
- CDN integration for global performance
- Auto-scaling based on usage patterns
 
OUTPUT: Complete enterprise deployment guide with security and scaling best practices."