Refactor docs from test-harness tour to system-architecture tour#7
Conversation
Reframe the HTML docs around how the index engine is *constructed* rather than around the test suite's file/step layout, using the test suite as ground truth for every claim. - Rename docs/test-harness-tour.html -> docs/architecture.html. - Organize by system architecture: the governing equation (index = projection of Markdown ∪ log), the three storage tiers, the crate/module layout, the substrate bet, the schema, then the data flows (ingest, notes/write discipline, chunks/links/graph, relation vocabulary, embedder seam, hybrid retrieval, suggestion lifecycle, event log/replay, and the Vault/CLI adapter). - Each layer cites the specific tests that pin it as ground truth, plus a single "Grounded in the tests" section that maps every architectural property to its covering test file. - Bring the doc current with the built system: 3 crates (adds b2-embed's real candle-backed LocalEmbedder), 73 passing tests, rustc 1.96.0, and the fail-fast model-mismatch / open()-never- mutates-vectors behavior. - Update the README link to point at the new doc. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01265nrsLFg8VDWMrBXuUoba
|
Caution Review failedThe pull request is closed. ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Free Run ID: 📒 Files selected for processing (3)
📝 WalkthroughWalkthroughReplaces ChangesArchitecture documentation swap
Estimated code review effort: 1 (Trivial) | ~5 minutes Sequence Diagram(s)Not applicable — this PR is a documentation-only change with no functional code flow. Poem
Note 🎁 Summarized by CodeRabbit FreeThe PR author is not assigned a seat. To perform a comprehensive line-by-line review, please assign a seat to the pull request author through the subscription management page by visiting https://app.coderabbit.ai/login. Comment |
… & graph) Four decisions locked ahead of their slices (research §9b #7–#10): - chunk/centroid DDL = one-of nullable FKs on the single chunks table + a sibling resource_centroids table (CASCADE intact; FTS/embeddings keep one rowid space) - similar/doc-ref anchors dispatch by extension (.md/b2id → note, else resource) — the §3 classification doctrine, one rule everywhere - "unlinked" for a resource anchor = the same 1-hop exclusion, direction-agnostic (inbound-only by construction) - façade grows list_resources; list_notes untouched Stage B: specs/resources-inventory-graph.md — the slice-1 build spec (v4 schema: resources table + edges dst_resource_path/embed/caption; generalized walk with (size,mtime)-gated blake3 hashing and pruning; parser forms with caption capture; doc_kind dispatch helper; explain/mv over resources; desktop tree + fallback card + opener; watcher allowlist→denylist inversion). tasks.md points at it. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Reframe the HTML docs around how the index engine is constructed
rather than around the test suite's file/step layout, using the test
suite as ground truth for every claim.
(index = projection of Markdown ∪ log), the three storage tiers,
the crate/module layout, the substrate bet, the schema, then the
data flows (ingest, notes/write discipline, chunks/links/graph,
relation vocabulary, embedder seam, hybrid retrieval, suggestion
lifecycle, event log/replay, and the Vault/CLI adapter).
plus a single "Grounded in the tests" section that maps every
architectural property to its covering test file.
b2-embed's real candle-backed LocalEmbedder), 73 passing tests,
rustc 1.96.0, and the fail-fast model-mismatch / open()-never-
mutates-vectors behavior.
Co-Authored-By: Claude Opus 4.8 noreply@anthropic.com
Claude-Session: https://claude.ai/code/session_01265nrsLFg8VDWMrBXuUoba
Summary by CodeRabbit