Skip to content

Understanding Tools

Understanding tools help your AI assistant comprehend what your code does. They provide semantic search across your codebase, precise symbol lookup by fully qualified name, AI-generated module summaries, and pattern detection for architectural and design patterns.

Search your codebase using natural language. Returns code snippets and symbols that are semantically relevant to your query, powered by vector embeddings.

Name Type Required Description
query string Yes Natural language search query
workspace_id string No Limit results to a specific workspace
repo_id string No Limit results to a specific repository
language string No Filter results by programming language
limit integer No Maximum number of results (default: 10)

“Find code that handles user authentication”

The assistant calls semantic_search with the query and returns relevant code snippets with file locations, helping you understand how authentication is implemented without manually searching the codebase.


Look up a symbol by its fully qualified name. Returns the symbol’s definition, documentation, location, visibility, and relationships to other symbols.

Name Type Required Description
fqn string Yes Fully qualified name of the symbol (e.g. com.example.UserService.findById)
workspace_id string No Limit results to a specific workspace
repo_id string No Limit results to a specific repository

“What does com.example.UserService.findById do?”

The assistant calls symbol_lookup with the FQN and returns the method’s signature, documentation, file location, and any metadata extracted during indexing.


Get an AI-generated summary of a module or package. Returns a natural language description of what the module does, its key exports, and how it fits into the broader codebase.

Name Type Required Description
module_path string Yes Path to the module or package (e.g. src/auth or pkg/handlers)
workspace_id string No Limit results to a specific workspace
repo_id string No Limit results to a specific repository

“Summarize the auth module”

The assistant calls module_summary and returns a concise description of the module’s purpose, its public API, internal structure, and dependencies. Summaries are generated by the enrichment pipeline after indexing.


Search for code patterns, design patterns, or anti-patterns in your codebase. Returns matches with file locations and context.

Name Type Required Description
pattern string Yes Pattern to search for (e.g. singleton, factory, god class)
workspace_id string No Limit results to a specific workspace
repo_id string No Limit results to a specific repository
language string No Filter results by programming language

“Find all singleton patterns in this repo”

The assistant calls pattern_search and returns locations where the specified pattern was detected, along with confidence scores and contextual information about each match.