Cleans up the crawl-generated facility table that surfaced garbage on /Facilities
(«بیمارستان هستم», «... از مدجابز», bare «کلینیک», «سازمان برنامه جنوبی» x3):
- FacilityMatcher.IsJunkName: shared detector for non-names — bare type words, cores
made only of filler/verb tokens, and leaked crawl-source/placeholder text. Added
داروخانه/آسایشگاه to the generic type words so bare ones are caught and dedupe better.
- HeuristicListingParser.ExtractFacilityName now rejects junk candidates (and emoji), so
new ingests fall back to the shared placeholder instead of forging a fake facility.
- IngestionService.MergeAndCleanFacilitiesAsync (+ admin button): folds junk facilities
into the placeholder and merges Persian-fuzzy duplicates into one keeper, repointing
their shifts/jobs first. Hard guard: only purely crawl-generated, unmanaged facilities
are removed — employer-owned and verified facilities are never touched.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Per the project archive-not-delete convention, the in-place purge now sets out-of-scope
and duplicate aggregated jobs/shifts to ShiftStatus.Archived instead of hard-deleting:
- The row is retained for analysis and the change is reversible.
- The listing drops out of every public screen and the sitemap (which filter Status == Open).
- Its detail page now returns 410 Gone (the standard permanent-removal signal) so search
engines deindex it cleanly, instead of leaving the off-topic page live at 200 or hard-404ing.
Dedupe of job reposts archives the older copies the same way. Coordinate backfill now also
skips non-Open rows. Valid listings are untouched, so IDs/URLs stay stable.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Jobs now keep the AI-extracted salary (d.PayAmount ?? parsed.PayAmount); they
previously used only the parser figure, so every aggregated opening showed «توافقی».
- Geocoder also scans the ad body, so Tehran ads that name a neighbourhood only in
free text («… در سهروردی») get an approximate map point.
- New BackfillCoordsAsync (+ admin button): fills missing coords on existing aggregated
listings from their stored text, in place — no ID/URL churn, SEO-safe.
- New PurgeInvalidAggregatedAsync + DedupeJobsAsync (+ admin button): in-place removal of
out-of-scope (domestic/promo/spam) aggregated jobs/shifts and duplicate job reposts,
keeping valid listings' IDs.
- Jobs detail page always renders the location card (matches Shifts) instead of hiding it
when coords are missing.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
When a listing's facility is the unknown placeholder, don't show «مرکز درمانی
(نامشخص)» anywhere — just leave the location out. Gated on HasRealEmployer:
- cards (shift/job/recommendation): the 🏥 facility line is omitted
- shift detail: H1 drops the «— نامشخص» suffix; title/description use city only;
«شیفتهای دیگر این مرکز» hidden; report label generic
- job detail: subtitle drops 🏥, keeps 📍 city; title/description city-only
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Re-check of live applicants found two gaps:
- «کمک بهیار آماده به کار» — the availability phrase glued onto the role. StripRoleModifiers
now removes «آماده به کار / آماده همکاری / جویای کار / جهت همکاری» phrases before
token-stripping, so the role collapses to «کمک بهیار».
- «خانم امورسبک منزل» — light-housework domestic helpers (not کادر درمان). Validator
now discards ads with «امور منزل / نظافت منزل / خدمتکار / مستخدم …» markers.
Both take effect for existing data on the next applicant reprocess.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Shift/Job 426-style pages showed 09910540686 — the «نامشخص / ثبت نشده» placeholder
facility's phone, set once and shown on every unnamed-facility listing (and in the
contact modal), even though it isn't that ad's number. Now the facility phone/Bale
is only used as a fallback when the facility is a REAL named employer
(SeoJsonLd.HasRealEmployer); otherwise fall back to the Divar source link (if any)
or «شماره ثبت نشده». Fixed in the /contact modal endpoint and both detail-page
inline reveals.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Cards had a 📞 contact-trigger that opened the call modal straight from the list.
Per request, calling should happen only on the post's detail page. Reverted each
card's CTA to a plain «جزئیات»/«مشاهده و تماس» button that just navigates to the
detail page (the whole card is already a link to it); the contact modal/trigger
now lives only on the shift/job/talent detail pages.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
New /Admin/Roles screen lists every role with its shift/job/talent usage and lets
an admin merge a duplicate role into another — reassigns all listings (the Restrict
FKs) plus preferences/alerts/profiles to the target, then deletes the source — or
toggle a role's visibility. Linked from the admin panel nav (🏷️ نقشها). Lets you
clean up dynamic-ingestion sprawl («کمکیار»→«کمک بهیار») without DB surgery.
Improvement 7 of the backlog (data).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Mark up the result list as a schema.org ItemList (ordered listing URLs) so Google
reads the landing/list pages as a curated collection. Emitted alongside the
breadcrumb JSON-LD when there are results.
Improvement 6 of the backlog (SEO).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add SeoJsonLd.Breadcrumb + Crumb record + _Breadcrumbs partial, and wire a trail
into the Jobs/Shifts list (landing) and detail pages: خانه › استخدام/شیفت › {نقش}
› {شهر|عنوان}. The role crumb links to the role landing page (more internal
links), and Google can show the breadcrumb path in results. Detail pages emit it
alongside the existing JobPosting JSON-LD.
Improvement 5 of the backlog (SEO).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Role/city landing pages were heading + list only — thin-content risk that hurts
ranking. Add a short, unique-per-page intro (built from the dynamic heading) on
the Jobs/Shifts landing pages, with internal-link guidance. Generic /Jobs and
/Shifts stay as-is.
Improvement 4 of the backlog (SEO).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
JobPosting requires a valid hiringOrganization; emitting «نامشخص / ثبت نشده» (the
placeholder for aggregated ads with no named center) makes Google reject the
posting and can flag invalid structured data across the site. Add
SeoJsonLd.HasRealEmployer and gate the JobPosting/ShiftPosting <script> on it, so
only listings with a genuine employer get marked up (those are the Jobs-eligible
ones anyway).
Improvement 3 of the backlog (SEO).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The /استخدام/{role}/{city} and /شیفت/{role} landing pages were only reachable via
the sitemap — no internal links, which is weak for ranking. Add a role quick-link
chip strip to the Jobs and Shifts list pages linking to the per-role landing URLs.
Since those list pages ARE the landing pages, this also cross-links every landing
page to all the others, building an internal-link mesh that passes authority and
aids crawl far more than the sitemap alone.
Improvement 2 of the backlog (SEO).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Exact ContentHash dedup misses the same ad reposted with slightly different text
(e.g. the ~18 repeated «کمکیار آقا»). DedupeTalentAsync collapses open aggregated
applicants by two high-precision signals — identical phone, or identical
(role, city, normalized description core with digits/«… پیش» time-phrases
stripped) — keeping the newest of each group. Runs at the end of both RunAsync
and ReprocessAsync; removed count surfaces in the run log.
Improvement 1 of the data-quality/SEO backlog.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The server clock is correct (UTC); the app rendered UTC wall-clock directly, so
the run log showed ~3.5h behind Tehran. Add JalaliDate.ToTehran (flat UTC+3:30 —
Iran dropped DST in 2022) + DateTimeLabel, and convert the UTC-stored timestamp
displays (ingestion run log, RawListing FetchedAt, report CreatedAt). Shift
start/end inputs are TimeOnly, left as-is.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Reprocess deletes+rebuilds aggregated listings, which changes their IDs. Shift/Job
detail pages are indexed and in the sitemap, so churning them would 404 ranked
URLs. «آماده به کار» pages are NoIndex + Disallow, so rebuilding them has zero SEO
impact — and that's where all the duplicate/sprawl problems were.
ReprocessAsync(talentOnly: true) now only deletes/rebuilds TalentListings and
skips non-talent raws (leaving shift/job listings + their RawListing links
untouched). Admin button relabelled «پردازش مجددِ آماده به کارها (امن برای SEO)».
Shifts/jobs self-clean via normal ingestion turnover.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Where deterministic geocoding gives up (neighborhood not in the TehranGeo table),
fall back to the registered AI model: the auditor now also returns approximate
lat/lng for a recognized Tehran neighborhood (folded into the existing single
audit call — no extra requests), and Publish uses it only after the source ad and
the local table, and only when it falls inside greater Tehran (InTehran bbox
guard rejects hallucinated points). Coords order: Divar point → TehranGeo → AI.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Re-checked live data and found cases the first pass missed:
- Gender baked into roles («پرستار آقا», «کمک بهیار آقا») → StripRoleModifiers
removes آقا/خانم/مرد/زن/کارآموز/ارشد… from role names (none of the real roles
contain these), collapsing the sprawl; gender still lives on the Gender field.
- «کمکیار» vs «کمک بهیار» forking → alias maps them to one role.
- Personality words («خوشاخلاق», «دلسوز», «منظم»…) added to the tag stop-list.
- Prompt: gender goes to the gender field, not the role.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Many Medjobs/Telegram ads name a Tehran neighborhood («ونک», «تهرانپارس»…) but
carry no coordinates. New TehranGeo geocoder maps ~45 neighborhood names to a
rough center; Publish falls back to it (from the resolved district / AI district
/ area note) when the source ad has no point. Shown via the existing «محدودهٔ
تقریبی» circle + disclaimer — never a precise pin. Tehran-only; extends the
existing approx-coords feature so non-Divar listings can show a map too.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
We captured Divar's privacy-fuzzed coords on RawListing but discarded them for
the listings that need them: unnamed-facility shifts/jobs dropped them (to avoid
piling on the shared placeholder) and applicants had no coordinate field at all.
- Add Lat/Lng to Shift, JobOpening, TalentListing (migration ListingApproxCoords).
- Publish stores the source ad's approx coords on each aggregated listing.
- Detail pages render the map from the listing's own coords (fallback: facility),
and aggregated coords show as a shaded «محدودهٔ تقریبی» circle (not a precise
pin) via _NeshanMap data-approx, with a disclaimer. Applicants get a map card
(they had none) + the page now loads the Neshan key.
Only Divar provides coords; the map needs NeshanMapKey set in admin settings.
Existing rows get coords once reprocessed (RawListing already has them).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The /search/suggest endpoint now returns { items, total } — each filtered query
is reused for both the Take(5) preview and a CountAsync — and the dropdown's
footer link reads «مشاهده همه N نتیجه برای «q»» (Persian digits) instead of a
bare «همه نتایج». The /Search page already showed counts.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Qualified live applicants and found three problems, all fixed:
- Duplicate cards: one ad fanned out into «پرستار» + «پرستار کودک» (same person).
Applicants now publish ONE listing (no role fan-out); secondary roles → tags.
- Role sprawl: modifiers became roles. Prompt now returns the BASE profession
and pushes age-group/ward/seniority to tags; new roles only for a genuinely
new base profession (تکنسین داروخانه ✓, پرستار کودک ✗).
- Tag/category noise: categories pinned to the 5 fixed groups (+سایر, never
invented); BuildTags drops pay/contact/location/fragment words.
Reprocess action: IngestionService.ReprocessAsync re-runs the current pipeline
over every stored RawListing WITHOUT re-fetching (keeps the raw text, so nothing
is lost to sources only exposing recent posts), deleting the old aggregated
posts and republishing cleanly. Admin dashboard button «پردازش مجددِ آیتمهای
ذخیرهشده» runs it on a background scope; result lands in the run-log.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds a lazy-loaded contact modal. Any element with data-contact-type +
data-contact-id (the «📞 تماس» button on shift/job/talent/recommendation cards,
and the contact CTA on the three detail pages) opens a modal that fetches the
listing's numbers from a new GET /contact endpoint and renders them with click-
to-call links. Numbers are loaded only on click, so they never sit in list-page
HTML (privacy / anti-scrape). The endpoint logs the same Apply interest signal
for shift/job that the old inline-reveal POST did, and falls back to the
facility phone (or Divar source link for talent) when an ad has no own contacts.
Verified locally: GET /contact?type=shift&id=1 → {title, contacts:[{value:
'021-82032000', href:'tel:...'}]}, and the modal opens and renders on the shift
detail page.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Google Search Console shows all top queries are «استخدام [نقش] [شهر]», but the
filtered index pages all shared the generic title «موقعیتهای استخدامی» and
weren't in the sitemap, so nothing ranked for those exact searches.
- Jobs/Shifts/Talent index pages now set a dynamic <title>/<h1>/meta from the
active role+city (e.g. «استخدام پزشک عمومی در تهران»).
- Pretty SEO routes /استخدام/{role}/{city?} and /شیفت/{role}/{city?} (via
AddPageRoute) resolve slugs → filters; unknown slug → 404. The layout already
derives the canonical from the path, so each pretty URL is its own canonical
and the query-string forms canonicalize to /Jobs (no duplicate content).
- sitemap.xml now lists role-only and role×city landing URLs for every combo
with live listings (URL-encoded), so Google discovers them.
- New SeoSlug helper (Persian-tolerant: ي/ك, ZWNJ, hyphen/space).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Phone fix: shifts/jobs showed Facility.Phone, but unnamed ads all share one
placeholder facility, so every such listing displayed the same stale number
while the ad's real phone sat unused in the description. ContactMethod is now
attachable to a Shift/JobOpening (not just talent); ingestion stores the ad's
own number(s) on each listing and the detail pages render them (new
_ContactList partial), falling back to the facility phone only when the ad had
none. Migration ShiftJobContacts (nullable owner FKs) — auto-applies on deploy.
Stale applicants: skip «آماده به کار» posts older than 7 days at ingest, by the
source's real timestamp (Telegram <time>, Bale date) or a Persian time-ago
phrase in the text (Divar «۲ هفته پیش»). Recorded as Discarded; shifts/jobs
are not aged out.
Admin: Review page now shows a «مشاهده آگهی در منبع» link (RawListing.SourceUrl)
so the source post can be checked before publishing.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Same inversion as the shift card — the «پیشنهادهای ویژه شما» box headlined
Facility.Name («مرکز درمانی (نامشخص)»). Role is now the headline; facility
moves to the second line with 🏥 alongside the city.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The headline showed Facility.Name (often «مرکز درمانی (نامشخص)» for ingested
shifts) while the actual role was a tiny badge. Match _JobCard/_TalentCard:
role becomes the headline; facility moves to the second line with 🏥.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Group RawListings by SourceChannel, fold per-channel/per-host labels into
source families (تلگرام/x → تلگرام, وبسایت (host) → وبسایت), and show a
published-vs-total table so it's clear which sources are actually producing
(e.g. why everything is coming from دیوار when Telegram's proxy is down).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Unknown roles from the AI are now resolved-or-CREATED (Persian-normalized dedupe) instead of dropped/fallback; new role gets the AI's category, assigned to the applicant.
- AI output gains category + tags; AI-detected skills/requirements (ICU, MMT, پروانهدار…) now fold into the applicant's searchable Tags.
- System prompt is hardcoded in AppSetting.DefaultPrompt and used directly by the auditor; admin sees it read-only (cannot edit/break it).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The Test-AI button called AuditAsync, which caught every exception and returned
null, and used EnsureSuccessStatusCode() (discarding the response body). So a
failing AI service only ever produced a generic 'no response' message with no
detail — impossible to diagnose.
- Add IAiAuditor.TestAsync: runs the real call and returns a detailed Persian
diagnostic — HTTP status + response body on non-2xx, raw body when the shape
isn't OpenAI-compatible, and network/proxy/timeout specifics on exceptions.
- AuditAsync now logs the actual HTTP status + response body (and proxy state)
instead of a bare warning, so server logs show why a call failed.
- ExtractContent / ParseVerdict no longer throw on unexpected JSON; they return
null so the caller can show the raw body.
- Settings 'Test AI' button uses TestAsync; result box renders multi-line and
switches to alert-error styling when the test fails.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Submit button is now a 44x44 magnify icon inside the search pill on mobile instead of a full-width stacked button (desktop keeps the جستجو text).
- Anchor the typeahead dropdown to the search pill so results appear directly under the input rather than below the popular-search chips; full pill width.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The hero is now the primary search; the navbar just links to the search
page (cleaner header, less clutter on mobile). Typeahead remains on the
hero (form[data-suggest]).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Reduce hero h1/p, page/section headings, stat pills and the hero search
font sizes on phones (<=560px); tighter hero padding. Desktop unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
On small screens the pill now stacks cleanly: a bordered, padded input above
a full-width جستجو button; icon hidden; chips centered. Shorter placeholder so
it never overflows.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The hero is now a single big search box → /Search (the rich, ranked,
highlighted search across shifts/jobs/applicants), with popular-search
chips. Typeahead is generalized to any form[data-suggest], so the hero box
shows the same instant highlighted dropdown as the header pill.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The suggest endpoint only matched role/city/tags/facility, so a term that
lives only in the ad body (e.g. mmt) returned nothing and the dropdown
never opened — even though /Search found it. Now each type also ILIKEs the
description, and the dropdown's sub-line is a snippet windowed around the
match (client highlights it). Title is bold; body wraps to 2 lines.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
An ad can cover several roles (e.g. «پرستار سالمند و کودک و همراه بیمار»).
The role dropdown is now a checkbox multi-select; on publish we fan out and
create one Shift/Job/Talent per selected role (mirrors the auto-ingest
fan-out). Jobs get a per-role title when multiple are chosen; talent
listings each get their own contact rows; all created items notify matches.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- SearchHighlight.Snippet: extracts a ±70-char window around the first
matching term and marks it (with ellipses) — the ES "highlight" fragment.
- Result cards (shift/job/talent) now show that snippet from the matched
description/tags when a query is present, so you SEE where the term hit
(e.g. «…دارای مدرک <mark>mmt</mark>…») instead of just the role.
- Typeahead suggestions gain a highlighted "sub" line (talent→tags,
shift→city·specialty, job→facility·city) so matches show in the dropdown too.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
NuGet loads the service index of EVERY listed source, so a 500 from the
Liara fallback aborted the whole restore (NU1301) even though Nexus was
healthy. Mirror cert chain is fixed now, so use our Nexus mirror as the
single source of truth.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Both the CI restore (/tmp/nuget.ci.config) and the Docker image build
(nuget.docker.config) now use https://mirror.soroushasadi.com/repository/
nuget-group/ as the primary source with Liara as fallback, so a single
mirror returning 500 no longer breaks restore.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Moved overflow:hidden onto an inner .nav-search-pill so the rounded corners
still clip the input/button, but the absolutely-positioned suggestions box
(a child of the non-clipped .nav-search) is no longer hidden. Dropdown given
a readable min-width.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Header search restyled as one clean RTL pill (input + button flush).
- Google-style autocomplete: typing ≥2 chars fetches /search/suggest and
shows up to 5 live matches (round-robin across shifts/jobs/applicants)
with the query highlighted, plus a «همه نتایج» link. Debounced, closes on
outside-click/Escape.
- Search results page now RANKS by relevance (term hits in role/title/
facility/city/tags weighted ×3, description ×1) instead of date-only.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- /Search: searches shifts, hiring openings, and applicants together via
Postgres ILIKE (every term must match across role/city/facility/title/
description/tags/person). Results grouped per type.
- Keyword highlighting (<mark>) extended to shift & job cards (was talent-only),
so matches stand out everywhere.
- Persistent header search box (.nav-search) → /Search; big hero box on the
page itself.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replaced the plain "interest recorded" alert with a styled .contact-reveal
card that fades/slides in and lists each channel as its own row (icon +
label + value + action button). Shift/job show facility phone + Bale;
talent shows all its ContactMethods in the same table style.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Tags: parser extracts cert/skill keywords (mmt, ICU/CCU, دیالیز, اتاق عمل,
اورژانس, مسئول فنی, پروانهدار…) + role + city into TalentListing.Tags
(+ migration); shown as chips on cards.
- Deep search on /Talent: «جستجوی عمیق» box does Postgres ILIKE across
tags, description, person, area, role, city (every term must match);
matches are highlighted with <mark> via SearchHighlight.
- Never delete: ShiftStatus.Archived + the admin «بایگانی گروهی» action now
ARCHIVES aggregated posts (hidden from site, kept in DB) and leaves the
raw crawl rows intact — a permanent archive for future analytics.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
soroush.asadi