Code Indexing

The Code Indexing feature enables AI agents to understand, search, and navigate your codebase using semantic search. Instead of simple text matching, your AI assistant can find code by meaning – search for “user authentication” and find the relevant login functions, session handlers, and security modules.

What is Code Indexing?

Code Indexing analyzes your source code using tree-sitter for accurate parsing across multiple programming languages. It extracts meaningful symbols (classes, functions, methods, interfaces) and creates vector embeddings for semantic search.

This means you can:

• Search code by meaning: Find “error handling for database operations” and get relevant try-catch blocks, error handlers, and logging code
• Navigate large codebases: Quickly find implementations, references, and related code
• Get intelligent context: Help AI agents understand your code structure and patterns

Supported Languages

Remembrances supports 14+ programming languages out of the box:

How to Use Code Indexing

1. Index a Project

Use the code_index_project tool to start indexing your codebase:

code_index_project({
  "project_path": "/path/to/your/project",
  "project_name": "My Project",
  "languages": ["go", "typescript", "python"]
})

The indexing runs in the background. For large projects, you can check progress with code_index_status.

2. Search for Code

Semantic Search – Find code by describing what you’re looking for:

code_semantic_search({
  "project_id": "my-project",
  "query": "user authentication and session management",
  "limit": 10
})

This returns relevant code snippets ranked by semantic similarity, even if they don’t contain the exact words in your query.

Find Symbols – Search by function, class, or method name:

code_find_symbol({
  "project_id": "my-project",
  "name_path_pattern": "UserService/authenticate",
  "include_body": true
})

3. Navigate Code Structure

Get File Overview – See all symbols in a file:

code_get_file_symbols({
  "project_id": "my-project",
  "relative_path": "src/services/auth.go"
})

Find References – Find where a symbol is used:

code_find_references({
  "project_id": "my-project",
  "symbol_name": "validateToken"
})

Get Call Hierarchy – See what calls a function and what it calls:

code_get_call_hierarchy({
  "project_id": "my-project",
  "symbol_name": "processPayment"
})

4. Manage Projects

List Projects – See all indexed projects:

code_list_projects()

Get Project Statistics – See indexing details:

code_get_project_stats({
  "project_id": "my-project"
})

Re-index a File – Update after changes:

code_reindex_file({
  "project_id": "my-project",
  "relative_path": "src/services/auth.go"
})

Code Manipulation

Beyond search and navigation, Remembrances provides tools for modifying code:

Get Symbol Body – Retrieve the full implementation:

code_get_symbol_body({
  "project_id": "my-project",
  "symbol_name": "validateToken"
})

Replace Symbol Body – Update an implementation:

code_replace_symbol_body({
  "project_id": "my-project",
  "symbol_name": "validateToken",
  "new_body": "func validateToken(token string) bool {\n  // New implementation\n}"
})

Insert Symbol – Add new code at a specific location:

code_insert_symbol({
  "project_id": "my-project",
  "file_path": "src/services/auth.go",
  "position": "after:validateToken",
  "code": "func refreshToken(token string) (string, error) {\n  // Implementation\n}"
})

Best Practices

Project Organization

• Index related codebases as separate projects
• Use descriptive project names for easy identification
• Include only the languages you actively work with

Efficient Searching

• Use semantic search for conceptual queries (“error handling”, “data validation”)
• Use symbol search for specific functions or classes you know by name
• Combine searches to narrow down results

Keeping Index Updated

• The indexer detects file changes and re-indexes automatically when watching is enabled
• For manual updates, use code_reindex_file after significant changes
• Delete and re-index a project if you’ve made major structural changes

Specialized Code Embedding Models

For optimal code search results, you can configure a dedicated embedding model specialized for code. See the Configuration page for details on setting up code-specific embedding models like CodeRankEmbed or Jina Code Embeddings.