Skip to content

Searching

How to get the most out of giddyanne's semantic search.

How It Works

Giddyanne uses embedding models to understand the meaning of code, not just match keywords:

  • "auth logic" finds login.py, session_manager.py, permissions.go
  • "error handling" finds try/catch blocks, error types, recovery code
  • "database queries" finds SQL, ORM calls, connection pooling

The search combines two approaches: - Semantic search: Finds code with similar meaning to your query - Full-text search: Finds exact matches for specific terms

By default, results use both (hybrid search). Use --semantic or --full-text to use only one.

Writing Good Queries

What Works Well

Conceptual queries - describe what the code does: 

giddy find "user authentication"
giddy find "parse command line arguments"
giddy find "send email notifications"

Problem-oriented queries - describe what you're trying to solve: 

giddy find "handle network timeouts"
giddy find "validate user input"
giddy find "cache expensive computations"

Behavior descriptions - describe what happens: 

giddy find "retry failed requests"
giddy find "log errors to file"
giddy find "convert dates to timestamps"

What Works Less Well

Variable names - use grep instead: 

# Less effective:
giddy find "userAuthToken"

# Better:
grep -r "userAuthToken" src/

Very short queries - add context: 

# Too vague:
giddy find "config"

# Better:
giddy find "load configuration from yaml file"

Tips

  1. Use natural language - write queries like you'd describe the code to someone
  2. Be specific - "database connection pooling" beats "database stuff"
  3. Include verbs - "validate", "parse", "send", "handle" help semantic matching
  4. Add context - "http error handling" is better than just "errors"

Search Modes

giddy find "auth"              # Hybrid (default) - combines both approaches
giddy find "auth" --semantic   # Semantic only - meaning-based matching
giddy find "auth" --full-text  # Full-text only - keyword matching

When to use each: - Hybrid: Best for most queries - Semantic: When you're describing functionality conceptually - Full-text: When you know specific terms that should appear in the code