Spaces:
Sleeping
Sleeping
Commit History
-
deploy: alert at 40/50/60/70/80/90/95/100% of each key's daily budget
3443a75 -
deploy: rotate across all Groq keys, alert admins at 80/100%
edc0702 -
deploy: claim/evidence scoring, CV integrity, upload safety, leak fixes
d209291
-
deploy: users admin tab, signup dates, campaign vs direct source
fe607c0 -
deploy: fill the profile from the CV, and rank courses on measured skill gaps
c3b8f02
-
Skill matching that survives plurals, tracker sync, and the apply funnel
e8b9cf0
-
Validate dates, split applicant counts
5df09a9 -
Stop counting a career break as work
fbeb1f9 -
Carry the role through the email links
b697865 -
Refuse a link that is not a job advert
e74475d -
Read the advert, score the location
a632649 -
Automatic feedback, readiness, calibration
a3fd220 -
Finish the rating plan
82bfe82 -
Rate a CV on what the role actually asked for
6e74a1f
-
Score the CV, show the CV, show the interview
5104934 -
Restore AI: Groq retired the Llama models
2f666f7 -
Design the emails
089da8b -
Repair the campaign flow end to end
3f14d37 -
Decide what a live campaign may still change
7ad99cd -
deploy: structured country targeting for campaign matching
da3259f -
deploy: third-party hosting/proxy flags need corroboration to block
e4646fc
-
deploy: operator role endpoint + verification rule tuning
4f1b2ed -
deploy: campaign flow + location verification check
Ports the location-poc verification rules verbatim (27/27 VPN exits caught, 0/18 false positives, re-verified through the new endpoint) and adds the campaign flow on top: public apply page, shortlisting, invite tokens stored hashed, and the admin surface. Client IP is read via a trust-proxy HOP COUNT (TRUSTED_PROXY_HOPS, default 1) rather than trusting the whole chain — the leftmost X-Forwarded-For value is client-supplied and would defeat every network rule. If the Space turns out to sit behind more than one proxy, set TRUSTED_PROXY_HOPS accordingly: that value being wrong misjudges every candidate. New tables are created by TypeORM synchronize on boot (campaigns, campaign_candidates, campaign_application_drafts, verification_sessions), plus nullable/defaulted columns on user and interview_sessions. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>86063f3
-
deploy: correct salary parsing + backfill endpoint for existing rows
parseSalaryValue no longer mines digits out of unparseable strings ("$30 - $130/hour" was becoming 30130). Adds admin POST /scraper/maintenance/backfill-salaries to fill the numeric mirrors on jobs written before those columns existed — without it, salary filtering returns nothing for the existing feed until expiry churns it.88800df -
deploy: activity-gated scraping, rescore on profile change, job-query cache
Nightly refresh now picks users by users.lastActiveAt instead of by rows in scrape_runs, which the cron itself writes — the old gate re-qualified users forever. Login refreshes a stale feed so quiet users cost nothing while away. Profile edits rescore existing jobs and courses. Job-board results cache for 12h. Salary filters run in SQL.a0eafe5
-
fix(normalize): stop labelling US-only remote jobs as open to anywhere
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>d42c579 -
fix(scraper): remote jobs classify as Anywhere; cap query hints at one
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>c3a9e33 -
fix(scraper): apply country filter on every scrape trigger
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>44c5c7a -
feat(scraper): filter-driven job search — push UI filters into Adzuna/JSearch APIs
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>a030856
-
deploy: scraper pipeline (Postgres storage, ranking, expiry, tracker, hardened LLM plumbing)
6fb7573
-
feat(resume): add targeted analysis and editing workflow
mahrooshishaqa1af4d9