get-repository-overview
This tool is the heart of the extension. It allows the agent to get a high-level overview of the codebase and its architecture.
It provides the LLM with a list of symbols and files, merging directory structure with symbol runtime connections. Additionally, the included metadata helps the agent understand the relative importance of symbols and their relationships.
- Ranks symbols using a weighted centrality score.
- Filters noise to fit a configurable token budget.
- Provides an architectural top-down view of the whole codebase.
get-symbol-context
This tool acts as a query engine for our language-server-derived symbol graph. It lets the LLM explore the surroundings of a symbol to understand cause and effect.
Critically, it resolves cross-file references so the agent can follow real edges and trace execution paths across the workspace without having to read those adjacent files in full.
- Returns the actual code of the symbol, not just a summary.
- Can be seen as an upgraded and more targeted version of common read-file tools.
- Provides a more detailed view of the symbol's surroundings, including its callers and callees, children and ancestors.
read-file & read-many-files
Sometimes agents still want to read a whole file. These two tools return the full contents plus metadata that keeps the agent oriented.
This is more of a fallback solution for edge cases when the agent still thinks, he is missing some important piece of context which he didn't get when using the Symbol Context tool.
- Returns the full contents of a single or multiple files.
- Provides metadata about the files, including the symbols defined in them and their references.