sinapse-ai 1.16.0 → 1.17.0

package/docs/framework/atlas/README.md

@@ -0,0 +1,34 @@
+# Framework Operating Atlas
+
+The living, generated map of **how the SINAPSE framework works** — routing,
+model selection, the constitution, the full workflow catalog, every agent and
+squad. One source of truth, two renderings (plus the raw data):
+
+| File | For | Format |
+|---|---|---|
+| [`OPERATING-ATLAS.md`](./OPERATING-ATLAS.md) | reading (human or LLM) | dense, navigable markdown with Mermaid diagrams |
+| [`atlas.html`](./atlas.html) | exploring visually | self-contained dashboard — searchable/filterable tables + flow diagram |
+| [`atlas-data.json`](./atlas-data.json) | tooling | structured data (one object per workflow / agent / squad / article / rule) |
+
+## Regenerate
+
+```bash
+sinapse atlas          # regenerate all three from disk
+sinapse atlas --open    # ...and open the visual dashboard
+```
+
+Counts are read directly from the repository every time, so the atlas is always
+exact (Constitution **Article VII — Ecosystem Metrics Accuracy**). After adding
+or changing a squad, agent, workflow, rule, or constitutional article, run
+`sinapse atlas` to refresh.
+
+## How it's built
+
+- `​.sinapse-ai/core/atlas/atlas-data.js` — scans the repo → the single data object.
+- `​.sinapse-ai/core/atlas/render-markdown.js` — the LLM-readable atlas.
+- `​.sinapse-ai/core/atlas/render-html.js` — the visual dashboard.
+- `​.sinapse-ai/core/atlas/index.js` — `generateAtlas()` (writes all three).
+
+The narrative "operating model" and "routing" sections are authored in the
+markdown renderer; every catalog (constitution, workflows, agents, squads,
+rules) is generated, so the docs can never drift from the code.