Knowledge graph (graphify)

This section describes the semantic knowledge graph of the TMC Suite package, built with the graphify tool. The graph is a machine- and human-readable map of the relationships between the files, classes, functions and concepts of the project. It is intended for online publication: a visitor can visually see the semantic relationships inside Tamic, search through them, and reuse that semantics in their own project.

Interactive graph: open graph.html (a self-contained page — the nodes are clickable and there is a search). Graph data: graph.json. In the compiled Sphinx site the graph is embedded directly into this page (see the RST version).

1. Why this belongs in the architecture

Ordinary documentation (API, build, user manual) describes the components individually. The knowledge graph shows the package as a single network of relationships: which modules reference each other, which functions are called together, which concepts belong to a single semantic group. This gives three practical benefits:

• Navigation by meaning. You can start from any node (for example, "Reading S-matrices") and walk the relationships to adjacent parts without knowing the file names in advance.
• Internal search. The graph includes a search index — a word finds all related nodes and communities.
• Reuse. The graph data is exported in an open format (graph.json), suitable for GraphRAG pipelines and third-party tools. Anyone building their own project around Tamic can rely on ready-made semantics instead of studying the sources from scratch.

2. What the graph contains

The graph is built over the entire repository (C++/MFC/OpenGL sources, headers, documentation). Current build figures:

The nodes are grouped into communities — automatically identified semantic clusters (for example: "TMCGROUT: graph coordinates", "Reading S-matrices (SFILE95)", "TMCIndan: output (IndanOutput)", "FieldView: rotations/shifts", "Radiation-pattern computation (DiaNapGr)"). Within the communities there are hub nodes (god nodes) — points through which the most relationships pass; they serve as "entry points" during navigation.

3. Artifacts (what is published)

All graph files live in the graphify-out/ directory at the project root:

To host it online it is enough to publish graph.html (a self-contained page) and, optionally, graph.json next to it — so that those interested can download the graph data.

4. How the graph is kept up to date

The graph is built from the code without calling paid services during updates — an update runs over the syntactic parse (AST). After changes to the code, the graph is rebuilt with the command:

graphify update .

A full rebuild (re-parsing all files) and export of the interactive HTML:

graphify .            # build/update the graph
graphify export html  # rebuild graph.html

Queries against an already-built graph (without a rebuild) — for search and tracing relationships:

graphify query "How does kernel data reach the viewer?"   # overview answer over the graph
graphify path "PlanarRT_H" "FieldView"                    # shortest path between nodes
graphify explain "Reading S-matrices"                      # explanation of a single node/concept

5. Status and purpose

The knowledge graph is an additional, reference documentation layer. It is not part of the compiled programs and does not affect the computations; it is a tool for navigation and reuse. It is intended for open publication together with the rest of the architecture documentation, so that external developers and researchers can study the semantics of Tamic and build on it in their own projects.