Semantic Search Engine
Search through the demo knowledge base using natural language. The search uses OpenAI embeddings and pgvector for semantic similarity matching.
Demo Knowledge Base
This demo is pre-populated with sample documents about semantic search, vector databases, embeddings, and evaluation metrics. Try queries like "how do embeddings work" or "what metrics measure search quality".
Features Demonstrated
Vector Search
Documents are chunked and embedded using OpenAI's text-embedding-3-small model, then stored in PostgreSQL with pgvector.
Relevance Scoring
Each result shows a similarity percentage calculated using cosine distance between the query and document chunk embeddings.
Filtering
Filter results by source type (internal, technical, external, research) and by tags to narrow down your search.
Evaluation
Label results as relevant or not to build a quality evaluation dataset. This feedback is stored for measuring search precision.