semantic-search-setup

npx skills add majiayu000/claude-skill-registry --skill semantic-search-setup --agent claude-code

Same command for any agent — swap --agent for codex, cursor, copilot.

# Semantic Search Setup Skill ## Overview This skill sets up vector embedding infrastructure for semantic search. Unlike keyword search (FTS5), semantic search finds conceptually similar content even without exact word matches. ## Quick Start ```python from sentence_transformers import SentenceTransformer import numpy as np model = SentenceTransformer('all-MiniLM-L6-v2') # Generate embeddings texts = ["How to fix a bug", "Debugging software issues"] embeddings = model.encode(texts, normalize_embeddings=True) # Compute similarity similarity = np.dot(embeddings[0], embeddings[1]) print(f"Similarity: {similarity:.3f}") # ~0.85 ``` ## When to Use - Adding AI-powered search to document collections - Finding conceptually related documents - Preparing knowledge bases for RAG Q&A systems - Building recommendation systems - Enabling "more like this" functionality ## How Semantic Search Works ``` Text Chunk Query | | v v +---------+ +---------+ | Embed | | Embed | | Model | | Model | +----+----+ +----+----+ | | v v [0.12, -0.34, ...] [0.15, -0.31, ...] | | +------------+------------+ | v Cosine Similarity | v 0.847 (similar!) ``` ## Model Selection | Model | Dimensions | Speed | Quality | Us

pip install sentence-transformers numpy
or
uv pip install sentence-transformers numpy
embed-service.sh - Background embedding service
if [ -f "$PID_FILE" ] && kill -0 $(cat "$PID_FILE") 2>/dev/null; then
echo "Already running (PID: $(cat $PID_FILE))"
return
fi
export CUDA_VISIBLE_DEVICES=""
nohup python3 embed.py --db "$DB_PATH" --batch "$BATCH_SIZE" \