scip-atlas-web

Read-only single-file SPA plus a Node pack CLI.

Tree explorer with search (name, path, symbols), file panel (owners, defined symbols, deps/rdeps), and GitHub/GitLab deep links when atlas stores remote metadata. Pack builds one slim DB (SCIP graph + atlas overlay, no source).

Canonical contract: docs/spec.md.

Artifact (SPA)

Build emits one index.html: TypeScript, CSS, and SQLite WASM inlined. GitHub Pages or Download → file://. Zero outbound HTTP after load. User DB stays in memory.

Dev	src/**/*.ts, CSS, @sqlite.org/sqlite-wasm
Ship	one HTML
Pack	bin/pack.ts — Node, not in the HTML

Prepare a repo (3 tools)

From the repo root, one step per tool. Artifacts land in ~/.cache/scip-cli/projects/<slug>/ (index.db, atlas.db, then <repo>.db when atlas meta has github_repo, else explorer.db).

# 1. scip-cli — index the code (symbols, deps graph, source chunks)
scip-cli reindex

# 2. scip-atlas — git overlay + path columns (no LLM required for browse)
scip-atlas sync

# 3. scip-atlas-web — slim SCIP + merge atlas → packed DB (no source in output)
npm run pack -- --gzip

Optional: scip-atlas summarize after sync for LLM blurbs in the tree.

Load in the SPA: grab the packed .db or .db.gz from the Output: line pack prints, then drop or open it in the app. --output accepts a file path or a directory (uses <repo>.db, or explorer.db when atlas has no github_repo).

Development

npm install
npm test
npm run pack -- --index path/to/index.db --atlas path/to/atlas.db
npm run build

Testing

Command	What
npm test	Vitest (pack, search, inspect, paths, …)
npm run dump-ui -- path/to/packed.db	Text report of tree/overlay data the SPA would show
npm run test:ui	Playwright browser e2e (optional; run npx playwright install once)

GitHub Pages

Pushes to main build dist/index.html and deploy via .github/workflows/pages.yml.

Live site: https://flesler.github.io/scip-atlas-web/