master_architect

The project map. It indexes a codebase once and lets the agent see files, functions and how they connect, before reading a single line.

Components

How it works

master_architect scans a project and stores it as a graph in SQLite: every file, module, function and import, plus the databases, programs and servers the code refers to. Each item gets a hierarchical address such as 13.2.4: file 13, module 2, method 4. Files it cannot parse still get an address (with letters), so they can be linked to.

Capabilities

• topology: the starting point. The project's files and how they import each other, plus databases and external programs.
• bones: one file's structure: its imports and its functions, with addresses.
• navigate: open a file, module or method by its address.
• connections: follow the links into and out of a file.
• ping: traces where a file leads, hop by hop through its imports (see below).
• context: a cross-reference summary for a file, by its path.
• missing: imports that point at nothing, found before anything runs.
• estimate and scan: check a project's size, then index it.
• crawl: follow links outward from a file to discover related files.
• languages and stats: supported languages and index statistics.

Tracing: where does a file lead?

ping works like a network trace-route, over the import graph. Starting from one file, it follows every internal import hop by hop, numbering the hops, until each branch ends. Each route is labelled by how it ends:

• external: an import of an outside library.
• blocked: an internal import that resolves to nothing: a broken link.
• cycle: an import of a file the trace has already reached.
• leaf: a file that imports nothing.
• endpoint: a server, program, model or port the code refers to.

Routes come back longest first, the best place to start looking for a bug that surfaces far from its cause. The walk stops after 25 hops unless told otherwise. Tracing is per file today; tracing individual function calls is planned.

Why it is built this way

• SQLite keeps the graph self-contained: no database server, and it handles graphs with hundreds of thousands of items.
• Hierarchical addresses are short enough to pass between tool calls and stay valid when lines move.
• Import resolution follows real project layouts: packages, plain sibling imports in script folders, and src/ layouts. An import rooted in the project that resolves to nothing is reported as broken, instead of being mistaken for an external library.

In the sandbox

Scanning is limited to your sandbox, and your project map is stored inside it, so it disappears with the sandbox.