getDocument preferred the on-disk file via readFileSync (raw), returning
the full frontmatter block as visible text on doc pages and MCP resources.
ReactMarkdown rendered the '---' delimiters as <hr/>, exposing YAML keys
(id/title/type/etc.) as plain paragraphs.
Root cause: getDocument bypassed parseFile's gray-matter stripping.
The indexer correctly stripped frontmatter (DB body was clean), but
getDocument read raw from disk for the 'source of truth' path.
Fix: use parseFile(abs, relPath).body (same as the indexer) so the
on-disk path returns frontmatter-free markdown. DB fallback unchanged.
Affects web doc pages + MCP mcpedia://docs/{+slug} and /chunks resources.
Verified: frontmatter leakage 0/9 doc pages, headings render as <h2>,
browser snapshot clean.
Added a real test suite (32 tests, 0 external services) using bun:test with
in-process module mocking for @mcpedia/db, @mcpedia/queue, @mcpedia/core.
Enablers:
- apps/api: extracted createApp(deps?) factory + dashboard.ts module from
index.ts so the HTTP surface is unit-testable (real queue is lazy-imported).
- packages/core: exported shouldCreateRevision pure predicate; restoreRevision
gained an opts.reindex seam for the chunk-rebuild contract.
- apps/mcp: renamed smoke.test.ts -> smoke.ts (bun test now owns .test.ts),
updated stale assertions (10 tools, 4 docs in docs section).
- infra: turbo test task (cache:false), test scripts across packages,
@types/bun + tsconfig base types, CI 'Test' step after Build.
Packages with tests: embeddings(5), parser(5), search(8), core(4),
mcp(6 auth-gates), api(8 contracts).
All green: typecheck(4/4), test(6/6 pkgs), build(web). Live API verified
/health, /metrics, /dashboard, /hooks/* auth gate on temp port.
Constructor threw if EMBED_BASE_URL/API_KEY/MODEL unset, which broke next
build SSG collection (imports search pkg before .env exists). Validation
now happens in embed() so the package is build-safe; a missing config still
errors clearly when a semantic/hybrid search actually runs.
With the DATABASE_URL import-time throw also removed, the whole monorepo
now builds in CI without any .env present.
The import-time guard broke 'next build' (SSG data collection imports
@mcpedia/config before any .env exists) and any runtime-injected env.
postgres.js connects lazily on first query, so a missing DATABASE_URL now
surfaces as a clear connect error at runtime instead of a cryptic build
failure. Lets CI build without .env present.
Reinterpreted from the plan's YAGNI 'Scale-out' (OpenSearch/object-storage
/multi-tenant deferred at KB scale). Phase 4 = make the Phase 3 async +
revision system correct, secure, observable, deployable.
- T1 (correctness bug): restoreRevision now rebuilds semantic chunks via new
@mcpedia/core reindexChunks(slug) so semantic/hybrid search stay consistent
after a restore (previously document_chunks held the NEW body while
documents.body held the restored OLD body -> stale search).
- T2 (security): /hooks/* git-sync webhooks now require x-webhook-secret header
matching WEBHOOK_SECRET (401 otherwise); API fails fast at startup if unset.
Added WEBHOOK_SECRET to @mcpedia/config + .env.example; set real secret in .env.
- T3 (UX): web doc page shows a History panel (revision no/reason/date/length)
with per-revision Restore; app/api/revisions/restore/route.ts calls
restoreRevision + revalidatePath (server-component only, no client JS).
- T4: listRevisions gains offset paging; summary never includes body.
- T5 (ops): deploy/mcpedia-api.service + deploy/mcpedia-worker.service systemd
units (Restart=on-failure, EnvironmentFile=.env). Not auto-enabled on host.
Verified against live imrnes Redis + Postgres: turbo typecheck+build green;
restore-rebuilds-chunks (marker present -> gone after restore); webhook 401/200;
web restore route redirects to doc + reverts body; revisions API returns summary
(no body); systemd-analyze verify passes.
asepharyana
asepharyana