The structured /cv/* calls funnel through a provider router so production can
offload a weak local GPU (GTX 1060) to a cloud provider without any .NET change.
Default stays "ollama" (keyless/local) and /summarize remains local distilbart.
- AI_PROVIDER=ollama|gemini|groq dispatch inside _ollama_generate_json/_text
(entry-point names kept, so no call sites change; Ollama path is byte-identical).
- Gemini (x-goog-api-key header, not URL query) and Groq (OpenAI-compatible
chat/completions) added via stdlib urllib — zero new dependencies.
- /health reports ai_provider + ai_provider_configured.
- Keys read from env only; never logged/committed.
- Compose + .env.example pass AI_PROVIDER/GEMINI_*/GROQ_* through.
Tests: 11 passed (default Ollama unchanged, Gemini/Groq dispatch, missing-key 503,
health reports provider).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add composite indexes (OwnerUserId, IsDeleted) — for the tenant-scoped
list/board/stats/analytics queries that all filter !IsDeleted — and
(OwnerUserId, FollowUpAt) for the reminders surface. Every JobApplication
query is scoped by the OwnerUserId global filter first, so owner-prefixed
composites are the useful shape; the pre-existing single OwnerUserId index
is now a redundant prefix but kept to avoid churn.
Status is intentionally excluded: Pomelo maps the unbounded string column to
MariaDB longtext, which cannot be indexed without a prefix length.
Applied via the startup schema reconciler (StartupInitializationExtensions),
which is how this repo actually provisions schema/indexes on both providers
(SQLite: CREATE INDEX IF NOT EXISTS; MariaDB: MySqlIndexExists-guarded CREATE
INDEX) — NOT via EF migrations, whose committed ModelSnapshot is stale.
OnModelCreating also declares the indexes for model consistency.
Backend suite: 92/92 green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Both endpoints materialised full JobApplication rows (GetAnalyticsOverview also
Include-d full Company) purely to aggregate a few fields, dragging the large
Description/TranslatedDescription/TailoredCvText/Notes/CoverLetter blobs over
the wire on every dashboard load. Project to only the columns each aggregation
needs (mirrors the existing GetTagTrends pattern). Behaviour is identical;
aggregation stays in memory over a small per-tenant set.
Backend suite: 92/92 green.
Note: the planned hot-path *index* migration is deferred — the committed EF
ModelSnapshot is stale (21 lines, no entities), so `migrations add` cannot
produce a clean incremental diff. Resyncing the snapshot is a prerequisite and
is tracked as its own task.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
cesnimda
cesnimda