Skip to main content
Ask questions about your codebase in natural language or search for specific patterns. Semantic search uses AI embeddings to understand the meaning of your queries.

​
Overview

The wit search command provides both AI-powered semantic search and traditional text search. 
wit search [query] [options]

​
Quick Start

# Ask a question about your code
wit search "where do we handle authentication"

# Find specific patterns
wit search "find all API endpoints"

# Interactive mode
wit search -i

​
Commands

​
Search Query

Search your codebase with a natural language query. 
wit search "query"

​
Examples

# Semantic questions
wit search "how does the database connection work"
wit search "where are errors handled"
wit search "find the main entry point"

# Specific patterns
wit search "find all TODO comments"
wit search "functions that use async/await"

​
index

Index the repository for semantic search. 
wit search index [options]

​
Options

OptionDescription
--force, -fForce reindex all files
Indexing analyzes your code files and creates embeddings for semantic search. 
# Index repository
wit search index

# Force full reindex
wit search index --force

​
Example Output

wit search index · Indexing repository for semantic search

Scanning files...

✓ Indexing complete

    Files indexed:  156
    Files skipped:  23
    Code chunks:    1,234
    Duration:       12.34s

​
status

Check the health of the search index. 
wit search status

​
Example Output

wit search status · Index health

✓ API key configured
✓ Index is ready

    Vectors:     1,234
    Files:       156
    Dimensions:  1536
    Last update: 2 hours ago

​
files

Search for files by glob pattern. 
wit search files "<pattern>"

# Find all TypeScript files
wit search files "*.ts"

# Find test files
wit search files "**/*.test.ts"

# Find config files
wit search files "*config*"

​
Search Options

​
Text Search

Force text search instead of semantic search. 
wit search --text "pattern"
wit search -t "pattern"

​
Files Only

Search only file names (not contents). 
wit search --files "pattern"

​
Content Only

Search only file contents (not names). 
wit search --content "pattern"
wit search -c "pattern"

​
Filter by File Type

Search within specific file types. 
wit search "query" --in "*.ts"
wit search "query" --file "*.py"
wit search "query" -f "src/**/*.js"

​
Interactive Mode

Launch an interactive search session. 
wit search -i
wit search --interactive

wit search · Interactive mode
Ask questions about your codebase. Press Ctrl+C to exit.

→ where do we handle user auth
  ...results...

→ how does caching work
  ...results...

→ exit

​
Search Results

​
Semantic Search Results

wit search · "where do we handle authentication"
Using AI-powered semantic search

  ● src/auth/handler.ts:45-89 (92% match)
    function: handleAuthentication
    │  45 │ export async function handleAuthentication(req: Request) {
    │  46 │   const token = extractToken(req);
    │  47 │   if (!token) {
    │  48 │     throw new AuthError('No token provided');
    │  49 │   }
    │  50 │   const user = await validateToken(token);
    │ ... │ (38 more lines)

  ● src/middleware/auth.ts:12-35 (85% match)
    function: authMiddleware
    │  12 │ export function authMiddleware() {
    │  13 │   return async (req, res, next) => {
    ...

──────────────────────────────────────────────────
Found 5 results in 0.34s · Index: 156 files

​
Text Search Results

wit search · "TODO"
Using text search

  Files
    ● src/index.ts
    ● src/utils/helpers.ts
    ● tests/setup.ts

  Content
  ● src/index.ts:42
    │ // TODO: Add error handling
    │ const result = processData(input);

  ● src/utils/helpers.ts:15
    │ // TODO: Optimize this loop
    │ for (const item of items) {

──────────────────────────────────────────────────
Found 8 results in 0.02s

​
Configuration

​
API Key

Semantic search requires an OpenAI API key for generating embeddings. 
# Set environment variable
export OPENAI_API_KEY=sk-...

# Or use Anthropic
export ANTHROPIC_API_KEY=...
Without an API key, wit falls back to text search.

​
Index Location

The search index is stored in .wit/search/ and includes:
  • Vector embeddings
  • File metadata
  • Index statistics

​
Best Practices

  • Understanding how something works
  • Finding related code across the codebase
  • Exploring unfamiliar code
  • Finding implementations of concepts

​
When to Use Text Search

  • Finding exact strings or patterns
  • Looking for specific identifiers
  • Finding TODO comments
  • Searching for error messages

​
Index Maintenance

# Reindex after major changes
wit search index

# Check index health
wit search status

# Force full reindex if issues occur
wit search index --force

​
Workflow Example

# 1. First time: index the repository
wit search index

# 2. Search for concepts
wit search "how does error handling work"

# 3. Search for specific patterns
wit search --text "catch (error)"

# 4. Find related files
wit search files "**/*error*"

# 5. Search in specific files
wit search "throw" --in "src/**/*.ts"

# 6. Interactive exploration
wit search -i