The flagship slice (docs/discovery/05): when a free-text query runs and embeddings
are available, fuse lexical top-50 (ts_rank_cd) with vector top-50 (pgvector cosine
over the structured-filtered set) using Reciprocal Rank Fusion. Placed BEFORE the
fuzzy fallback so queries with ZERO lexical hits still get semantic recall — found
and fixed during review (the fuzzy path would otherwise swallow pure-semantic
queries). Exact lexical hits keep winning (rank in both lists); deeper pages fall
through to lexical paging; ANY failure (Ollama down, nothing embedded) silently
degrades to the lexical/fuzzy path — AI can never break search.
LiveDb test proves the recall win: a 'banana' query surfaces an email with no
keyword overlap, ranked first, against real pgvector. 5 LiveDb + full suite green;
format clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Phase 5 re-validation caught a functional regression: the V-10 clamp in
SearchService.SearchAsync also capped CleanupService's internal target resolution
(pageSize 10000 -> 200), silently limiting bulk cleanup-by-query to 200 emails.
The clamp belongs at the user-facing trust boundary, not the shared service: move
MaxPageSize (200) enforcement into SearchController (both the POST body and GET query
paths). Internal callers of ISearchService now request large pages unhindered, while
user requests are still bounded. Adds a regression test proving SearchService returns
a 250-row page uncapped. No security regressions per Phase 5. All 39 tests green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Defense-in-depth tenant isolation: every user-owned entity (Email, Sender,
MailThread, MailDomain, Attachment, Label, SyncState, AnalyticsAggregate,
WidgetLayout, UnsubscribeItem) gets a global query filter restricting reads to the
authenticated user. AppDbContext takes an optional ICurrentUser; CurrentUserId is
Guid.Empty for background workers / design-time, which DISABLES the filter so sync
and tooling (which already scope by an explicit userId) are unaffected. On the HTTP
attack surface a forgotten manual `WHERE UserId ==` can no longer leak another
tenant's rows.
Phase 1 confirmed no active IDOR; this is preventive, and prioritised now because the
upcoming automation engine will add many new queries.
Also: moved the Npgsql-only tsvector FTS mapping out of EmailConfiguration into
AppDbContext.OnModelCreating, guarded by Database.IsRelational() (Ignored otherwise),
so non-relational test providers work — honouring the existing Email.SearchVector
comment. Production (Npgsql) model is unchanged; no migration needed.
Adds 3 cross-user tenant-isolation integration tests. All 38 tests green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
cesnimda
cesnimda