Quickstart

codedb Skill

Use the codedb plugin for fast, structured code intelligence queries via the codedb CLI. It provides tree-sitter-based structural parsing, trigram-accelerated full-text search, O(1) inverted word index, and reverse dependency graphs.

Quick Start

1. Install

codedb --version

2. Explore a Codebase

Get the file tree:

dcli codedb tree run

Outline symbols in a specific file:

dcli codedb outline run --path src/main.zig

3. Find Symbols and Search

Find where a symbol is defined:

dcli codedb find run --name AgentRegistry

Full-text search (trigram-accelerated):

dcli codedb search run --query "handleAuth"

Regex search:

dcli codedb search regex --pattern "handle[A-Z]\w+"

Exact word lookup (O(1) inverted index):

dcli codedb word run --identifier Store

4. Recently Modified Files

dcli codedb hot run

5. Nuke Codedb Data

Remove all codedb data, snapshots, and kill processes:

dcli codedb nuke run

Agent Notes

• codedb CLI pays ~55ms process startup per invocation. For repeated queries, consider running codedb serve as a long-running HTTP server on :7719 instead.
• Results are structured (JSON-like) rather than raw line dumps — significantly fewer tokens for AI agents.
• Sensitive files (.env, credentials, keys) are auto-excluded from indexing.
• Telemetry is on by default. Disable with CODEDB_NO_TELEMETRY=1 or --no-telemetry.
• For passthrough of undocumented subcommands, use dcli codedb _ _ <args...> (e.g., dcli codedb _ _ -- --help).