From 3ae2f8e43704deab8071ff37f73ceb1430c413db Mon Sep 17 00:00:00 2001
From: gelu <mihesgelu@easysoft.ro>
Date: Fri, 22 May 2026 19:01:05 +0300
Subject: [PATCH] Add Features/Internet-Job-Search wiki page

---
 Features-Internet-Job-Search.md | 81 +++++++++++++++++++++++++++++++++
 1 file changed, 81 insertions(+)
 create mode 100644 Features-Internet-Job-Search.md

diff --git a/Features-Internet-Job-Search.md b/Features-Internet-Job-Search.md
new file mode 100644
index 0000000..d1d2082
--- /dev/null
+++ b/Features-Internet-Job-Search.md
@@ -0,0 +1,81 @@
+# Internet Job Search from CV Match Email
+
+## Summary
+
+After a CV match with email, the result email contains a one-time job search link. Clicking it triggers an automated internet job search: the system scrapes configured job boards, pre-filters using CV keywords, runs the existing LLM CV-to-job match on the top candidates, and emails a ranked list of real-world job matches back to the user.
+
+## Status
+
+Deployed to staging 2026-05-22. End-to-end testing pending.
+
+## Commits
+
+- `6293fa8` -- Add internet job search feature (cv-search-job)
+- `a4c128f` -- Fix cv-matcher-api Dockerfile: add cv-search-models to build context
+- `cf06453` -- Refactor docker-compose: single deployable file + local override
+
+## Related Issues
+
+- #2 cv-search-models shared library
+- #3 cv-search-job background worker
+- #4 cv-matcher-api job search token + session endpoints
+- #5 api job-search proxy route and email link
+- #6 fix: cv-matcher-api Dockerfile missing cv-search-models COPY
+- #7 chore: docker-compose refactor
+
+## Architecture
+
+### Flow
+
+1. User does CV match and provides email
+2. cv-matcher-api saves a JobSearchToken (UUID, 7-day expiry) in the cvSearch DB schema
+3. Match result email includes link: https://myai.ro/api/cv-matcher/job-search/start?t=tokenId
+4. User clicks link -- api validates token, creates JobSearchSession (Pending), returns styled HTML confirmation
+5. cv-search-job polls every 30s, picks up Pending sessions:
+   - Extracts CV keywords (heuristic, no LLM cost)
+   - Scrapes each enabled provider via HtmlJobSearcher
+   - Runs existing ScorePairAsync LLM match on top MaxJobsToMatch URLs
+   - Emails ranked results to user + Contact:ToEmail with CV PDF attached
+
+### New Projects
+
+| Project | Path | Role |
+|---------|------|------|
+| cv-search-models | Apis/cv-search-models/ | Shared EF entities + CvSearchDbContext (schema cvSearch) |
+| cv-search-job | Jobs/cv-search-job/ | Worker: polls Pending sessions, scrapes, scores, emails |
+
+### Database Schema (cvSearch)
+
+| Table | Purpose |
+|-------|---------|
+| JobSearchTokens | One-time tokens created after each CV match with email |
+| JobSearchSessions | Lifecycle tracking: Pending to Processing to Done / Failed |
+| JobSearchResults | Per-job score, title, full extracted text, LLM result JSON |
+
+### Provider Configuration
+
+Fully configurable in appsettings under JobSearch:Providers. Each entry has SearchUrlTemplate (with {keywords} placeholder), JobLinkContains, InitialKeywords merged with CV keywords, and MaxResults. All providers disabled by default. Configured providers: ejobs.ro, bestjobs.eu, linkedin.com.
+
+## Docker Compose
+
+docker-compose.yml is the single file to paste into Portainer for both staging and production.
+
+| Variable | Default (Portainer) | Local dev (.env) |
+|----------|--------------------|--------------------|
+| IMAGE_TAG | staging / production | local |
+| LOGS_PATH | /opt/myai/logs | ./logs |
+| FILES_PATH | /opt/myai/files | ../Apis/api/Files |
+
+docker-compose.override.yml adds build, ports, and env_file for local dev (auto-merged by docker compose up).
+
+## EF Migrations
+
+    dotnet ef migrations add <Name> --context CvSearchDbContext --project Apis/cv-search-models --startup-project Apis/cv-matcher-api
+
+## Key Implementation Notes
+
+- cv-matcher-api/Dockerfile MUST COPY Apis/cv-search-models/ -- omitting this causes CI build failure
+- Use $$""" raw string literals for HTML generation -- CSS braces cause CS9006 with $"""
+- Orphaned Processing sessions (from container crash) are reset to Pending on cv-search-job startup
+- cv-search-job and api share the same bind-mount volume for CV PDF email attachment
+- No captcha on the job-search start link -- the one-time UUID token is the credential