Overview first turn

Overview (first-turn project summary)

One bounded call that answers what an agent asks first: how fresh is the index, how big is the project, how is it layered, what is referenced most, where does it start, which forms exist. Six sections in a fixed order, every one capped.

Running it from the CLI

drag-lint overview --db <file.sqlite> [--max-rows N] [--json]

--max-rows caps the rows of every section (default 20). A section that was cut says so: -- 1578 omitted by cap 20; rerun with --max-rows 1598. With --json the top-level keys are freshness, counts, layers, fanin, entrypoints, forms, and each holds total, cap, omitted and rows. The exit code does not change when a section is capped.

The sections

• freshness -- project tag, scan type, schema version, when the index was built, file and symbol counts.
• counts -- symbols per kind, largest first.
• layers -- units by dependency layer: layer 0 uses no other project unit, layer n uses only lower layers. A unit in or above a circular uses group is listed in the final cyclic row. Each row shows the first 8 unit names and +N more.
• fanin -- the most-referenced names (the same aggregation as top).
• entrypoints -- program files, initialization / finalization sections and .dfm event handlers.
• forms -- symbols of kind form.

What it needs

An index IS required. The project's DB lives at <project folder>\_D-RAG\<project file>.sqlite; run drag-lint resolve-dbs --project <X.dproj> if unsure. Reads only; never writes the DB.

Example

Illustrative only:

drag-lint overview --db C:\Projects\MyApp\_D-RAG\MyApp.sqlite --max-rows 10

This would print the six sections with at most 10 rows each.