pgx cannot implicitly encode int arg into text for the `$1 || ' month'`
concatenation pattern (error: "unable to encode 12 into text format for
text (OID 25): cannot find encode plan"). Multiplication with a known
interval works directly with the int parameter and is semantically
equivalent.
Discovered during the T19 smoke test — the tick endpoint returned 500
on every call before this fix.
Adds a batch/v1 CronJob that POSTs to /api/v1/admin/discovery/tick on a
configurable schedule (default every 4h). Wires DISCOVERY_TOKEN into the
ci-secrets Secret and projects discovery/AI env vars into the backend
Deployment.
Construct discoveryRepo, discoveryAgent, discoveryService, and
discoveryHandler in registerRoutes(); register all 4 discovery routes
on /api/v1 with bearer-token guard on /tick and admin-session guard on
queue management endpoints.
Normalizes market names for dedup matching: lowercase, umlaut expansion,
punctuation stripping, whitespace collapse, and leading/trailing filler
word removal. Guards stripping so edge fillers are preserved when the
remaining content is purely numeric (e.g. 'Markt 2026' stays 'markt 2026').
The home page dropped `plz` server-side, so /markets was called with
radius but no center (unfiltered) and the PLZ input rendered empty
after reload. +page.server.ts now reads plz, geocodes via /geocode,
and echoes plz back in searchParams for form rehydration.
Relaxes /geocode DTO + guard to accept PLZ without city — Nominatim
already supports postal-only lookups. URL lat/lon (GPS flow) take
priority over plz on tie-break; geocode failures fall through to no
geo-filter so the page always renders.
- Set resources req=limit (100m/128Mi) for Guaranteed QoS class
- Add ConfigMap checksum annotation to trigger rollouts on config changes
- Add retry limit (60 attempts) to migration init container
- Use TARGETARCH in Dockerfile for multi-arch build support
- Set GOMAXPROCS and GOMEMLIMIT from cgroup limits to prevent
thread oversubscription and unbounded GC memory growth
- Add startup probe (60s budget) to gate liveness/readiness during
connection pool initialization
- Increase liveness failureThreshold to 5 to avoid restarts on
transient issues
- Remove initialDelaySeconds (startup probe replaces this)
- Upgrade CI from alpine/helm:3.17 to alpine/helm:4.1
- Replace deprecated --atomic with --rollback-on-failure + --wait=watcher
BusyBox 1.37 nc -z is broken (outputs "punt!" and never exits),
causing the wait-for-cache init container to loop indefinitely.
The cache is healthy — the backend should handle reconnects itself.
Prevents the backend from starting before the DragonflyDB operator
has the cache pod ready and reachable. Mirrors the existing
wait-for-postgres pattern in the migration job.
Replace manual Valkey Deployment+Service with DragonflyDB operator CRD.
Add sectionName to HTTPRoute for HTTPS listener pinning and a separate
HTTP→HTTPS 301 redirect route. Update resources from req=limit to
request/limit separation for pay-as-you-go billing. Fix NetworkPolicy
cache pod selector to match operator-managed labels.
Mistral rejects json response format when tools (web search) are
active. Override to text since the prompt already requests JSON
output in the instructions.
Use the Conversations API (POST /v1/conversations) via the SDK for
Pass 1 agent calls instead of /agents/completions which doesn't
support built-in web search connectors. Pass 2 uses the SDK's
ChatComplete method.
Add Woodpecker secrets for AI_API_KEY, AI_AGENT_SIMPLE, and
TURNSTILE_SECRET_KEY. Create ci-secrets.yaml template and wire
them through the deploy step alongside existing SMTP secrets.
Set CPU and memory requests equal to limits (100m/100Mi) for backend,
cache, and web. Switch rolling update strategy to maxSurge=1,
maxUnavailable=0 so new pods start before old ones terminate.
Add readiness probe to cache deployment.
- Public search dedup: CTE with ROW_NUMBER picks best edition per series
(nearest future, fallback most recent past), COUNT(*) OVER for edition_count
- Add EditionCount to Market model and MarketSummary DTO
- GetBySlug returns nearest-future edition by default, supports ?year= override
- GetBySlug returns sibling edition briefs for edition switcher UI
- New GET /admin/markets/grouped endpoint returns AdminSeriesGroup[] with
series-level pagination and all editions per group
- Add AdminSearchGrouped to repository, service, handler, and routes
Admin endpoint POST /admin/series/:id/editions creates a new yearly
edition for an existing series, inheriting series-level defaults.
Public market submissions now auto-match against existing series using
trigram similarity (>0.6) on name + exact city match. Matched
submissions link to the existing series instead of creating duplicates.
Fix Gin route tree conflict by nesting markets and series groups under
a shared /admin parent group.
Split markets into market_series (persistent identity) + market_editions
(yearly instances). Migration preserves edition UUIDs from old market IDs
for URL backward compatibility. Edition statuses: rumored, confirmed,
active, completed, cancelled, archived.
AI research now uses two-pass pipeline: Pass 1 via pre-created Mistral
agent with web search for structured extraction, Pass 2 via
mistral-large-latest for description writing and field retry. Confidence
scoring is source-derived (source count + extraction type) instead of
LLM self-reported.
- AI research via Mistral API for admin market editing
- Auto-geocoding via Nominatim OSM with rate limiting
- Public geocode endpoint (POST /api/v1/geocode)
- Duplicate detection using pg_trgm trigram similarity
- Extended SubmitMarketRequest with street, opening_hours, admission_info
- pg_trgm migration for fuzzy name matching
- Add PUT /auth/password for setting/changing passwords (handles both
first-time set for magic link/OAuth users and change for password users)
- Generate random medieval display names (e.g. Gaukler1025) for new
magic link and OAuth users instead of leaving display_name empty
- Add has_password field to ProfileData response
vikingowl