Survey a codebase cheaply
Build a first-pass map of an unfamiliar .NET solution from the warm index, its symbols, routes, and counts, for a fraction of the token cost of reading source.
Goal: understand the shape of an unfamiliar solution, its types, endpoints, and layering, without reading the source.
The Cheapest Map
fuse map ./src --detail allfuse map prints what the index holds: the indexed symbols, the routes, and the counts,
no file bodies. It is the cheapest first call, because it reads the warm semantic index
rather than collecting and reducing source. Narrow it with --detail symbols or
--detail routes, and cap the rows per section with --maxRows.
See the Index State
fuse diagnostics ./srcfuse diagnostics reports the index schema version, mode (semantic, partial, or
syntax), record counts, and the store location, so you can confirm the workspace loaded
through MSBuild and Roslyn before trusting resolution.
An Architecture View: A Skeleton
When you want signatures rather than counts, plan a context at the skeleton render tier, which keeps class, interface, and method signatures and drops bodies:
fuse context ./src --seed OrderService --max-tokens 50000A budgeted fuse context call chooses the render tier per file to fit, so a tight budget
yields a signature-heavy architecture view of the seed neighborhood. To reduce a known set
of files to signatures directly, use fuse reduce --level skeleton.
What You Get
A small, structural payload: from fuse map, the symbols and routes the index knows about;
from a budgeted fuse context, a signature-level view of an area. Either is enough to
orient a reader or an agent before any deeper drill-in.
Survey the Current Repository
Run fuse diagnostics ./src, then fuse map ./src --detail all. If the map names the
route, service, request, or config section you need, resolve it directly. Otherwise run
fuse localize ./src --task "<your task>" and use the strongest candidate as the next
context seed.
Run Fuse verbs in CI on a pull request
A recipe that runs fuse review and fuse test on a pull request from the command line, so the same verified-edit verbs an agent uses also gate the pipeline.
Cut tokens for a .NET project
Render a .NET solution, or a known set of files, into a token-efficient payload and cap output to a budget when needed.