Semantic Search
Find what you mean, not just what you say. Search by concept, not keywords.
Why Semantic Search Matters
Traditional keyword search is like looking for exact matches in a dictionary. Semantic search understands what you're actually looking for.
Keyword Search
- ✗ Only finds exact word matches
- ✗ Misses synonyms and related concepts
- ✗ Fails with natural questions
Semantic Search
- ✓ Understands meaning and context
- ✓ Finds conceptually similar content
- ✓ Works with natural language
Example
Understanding Search Scores
Each result has a score showing how well it matches your query. Focus on results above 0.7 - those are almost always relevant.
0.85+Excellent - Use theseDirectly relevant. These are what you're looking for.
0.70-0.84Good - Worth checkingRelated content. May have useful context.
Below 0.70Weak - Usually skipProbably not what you need.
Search Examples
See how semantic search finds related concepts, not just exact matches:
"How do I handle user authentication?"Will also find conversations about:
- • OAuth 2.0 implementation
- • Login/signup flows
- • JWT token management
- • Session handling
- • Password reset flows
"database performance issues"Will also find conversations about:
- • Slow query optimization
- • Index creation
- • N+1 query problems
- • Connection pooling
- • Database scaling
Query vs Browse
Spaces has two ways to find your snapshots:
query-workspace
Best when searching for something specific
query-workspace({
query: "authentication decisions"
})- ✓ Semantic similarity ranking
- ✓ Natural language queries
- ✓ Finds related concepts
list-snapshots
Best for browsing or filtering
list-snapshots({
category: "implementation",
limit: 20
})- ✓ Chronological ordering
- ✓ Category/scope filtering
- ✓ Pagination support
Writing Better Queries
Good Queries
- "How did we implement authentication?"
- "database performance optimization approaches"
- "What framework did we choose for the frontend?"
- "discussions about API versioning"
Why they work: Natural language, specific topics, complete thoughts
Less Effective Queries
- "auth" - Too broad, lacks context
- "fix bug yesterday" - Temporal references don't help
- "that thing we talked about" - Too vague
Tip: Be specific about technical topics, not time or vague references