Fuse
Scenarios

Survey a codebase cheaply

Build a first-pass map of an unfamiliar .NET solution with a table of contents or a signature-only skeleton, for a fraction of the token cost.

Goal: understand the shape of an unfamiliar solution, its types, endpoints, and layering, without reading the source.

The cheapest map: a table of contents

fuse dotnet --directory ./src --toc

--toc emits a directory tree with a per-file symbol outline and token cost, instead of file bodies. It is the cheapest first call: on the sample fixture the table of contents is 221 tokens against 624 to read every listed file, and the saving grows with file count because each file contributes one line plus an outline regardless of its size.

The architecture map: a skeleton

fuse dotnet --directory ./src --skeleton --all --semantic-markers

A skeleton keeps class, interface, and method signatures and drops bodies, so the type and member surface of the whole solution fits in a fraction of the tokens. --semantic-markers annotates each type with its kind, the interfaces it implements, and the types it depends on.

Add structural maps

These prepend to the output, so combine them for a layered overview:

fuse dotnet --directory ./src --route-map --project-graph --skeleton
  • --route-map prepends a table of HTTP verb, path, and handler from controllers and minimal API endpoints.
  • --project-graph prepends the solution and project reference structure.
  • --public-api keeps only public and protected member skeletons, the contract a project presents to consumers.

What you get

A small, structural payload that opens with the project graph and route map, then the skeleton of every type: enough to orient a reader or an agent before any scoped drill-in.

When to use it

Use a survey at the start of work on a codebase you do not know, or to produce a high-level map for review. Follow it with a scoped fusion to bring detail to the area that matters.

Skeletons sit at the deep end of Reduction levels; the detected conventions in a pattern summary are in Pattern detectors.

On this page