Eliminates one Anthropic API call entirely by using pattern matching:
- Add 120+ known single-company names (Ernst & Young, M&S, law firms, etc.)
- Detect "/" separator as clear indicator of multiple companies
- Use company suffixes (Ltd, PLC) to identify when "&" means two companies
- Conservative approach: don't split ambiguous cases
Added 40 unit tests for compound name detection covering:
- Known single companies with & and "and"
- Slash-separated company names
- Ambiguous cases
- Edge cases (empty, null, short names)
Estimated savings: ~$0.01 per CV check, 100% elimination of this API call
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Major improvements to company name matching accuracy:
- Add well-known brands dictionary with correct Companies House numbers
for fast-track verification (Boots, Legal & General, EY, etc.)
- Add safe expansion words (UK, LIMITED, GROUP, PLC) that don't change
company identity
- Fix core word validation to require original company's core words
- Remove overly aggressive skip words that removed meaningful identifiers
(industries, technology, consulting, services, etc.)
- Add industry context hints for AI matching
- Fix CVBatchTester JSON deserialization for test files
Before: 98% verified but with false positives like:
- Boots → BOOTS AND BEARDS (wrong)
- Legal & General → LEGAL LIMITED (wrong)
After: 97% verified with correct matches:
- Boots → BOOTS UK LIMITED (correct)
- Legal & General → fast-tracked to correct company
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Uses Claude Haiku to intelligently detect when a company name contains
multiple companies (e.g., "ASDA/WALMART", "Corus & Laura Ashley Hotels")
vs single companies with similar patterns (e.g., "Ernst & Young").
- Adds ExtractCompanyNamesAsync to ICompanyNameMatcherService
- Only triggers for names with potential separators (/, &, "and")
- Verifies each extracted part individually, returns first match
- Uses fast Haiku model to minimize cost
Results:
- ASDA/WALMART → verified via 'ASDA' → ASDA GROUP LIMITED
- Corus & Laura Ashley Hotels → verified via 'Corus' → Tata Steel UK
- Employers: 104/120 verified (86%)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Rename all TrueCV references to RealCV across the codebase
- Add new transparent RealCV logo
- Switch from JetBrains Mono to Inter font for better number clarity
- Update solution, project files, and namespaces
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Update flag assertions to filter by specific flag types
(job now creates additional informational flags)
- Update overlap tests: now "Concurrent Employment" with Info severity
- Update overlap score tests: no penalty for overlaps
(legitimate for part-time, consulting, job transitions)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Replace fuzzy string matching with semantic AI matching to fix false
positives where similar-sounding but different companies were matched
(e.g., "Families First CiC" incorrectly matching "FAMILIES AGAINST
CONFORMITY LTD").
Changes:
- Add ICompanyNameMatcherService interface and AICompanyNameMatcherService
implementation using Claude Sonnet 4 for semantic company name comparison
- Add SemanticMatchResult and related models for AI match results
- Update CompanyVerifierService to use AI matching with fuzzy fallback
- Add detection for public sector employers, charities, and self-employed
entries that cannot be verified via Companies House
- Update tests to work with new AI matcher integration
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add audit logging system for tracking CV uploads, processing, deletion,
report views, and PDF exports for billing/reference purposes
- Add processing stage display on dashboard instead of generic "Processing"
- Add delete button for CV checks on dashboard
- Fix duplicate primary key error in CompanyCache (race condition)
- Fix DbContext concurrency in Dashboard (concurrent delete/load operations)
- Fix ProcessCVCheckJob to handle deleted records gracefully
- Fix duplicate flags in verification report by deduplicating on Title+Description
- Remove internal cache notes from verification results
- Add EF migrations for ProcessingStage and AuditLog table
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add dashboard auto-refresh polling to update when processing completes
- Skip verification for freelance employers (but not contractors)
- Add incorporation date check (flags employment before company existed)
- Add dissolution date check (flags employment at dissolved companies)
- Add dormant company check (flags non-director roles at dormant companies)
- Add company size vs role check (flags senior roles at micro-entities)
- Add SIC code mismatch check (flags tech roles at non-tech companies)
- Add director verification against Companies House officers
- Add rapid career progression detection (3+ seniority jumps in <2 years)
- Add early career senior role detection (<2 years after education)
- Extend CompanyVerificationResult with flags and company data
- Add officers endpoint to Companies House client
- Fix null reference warning in Report.razor
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Use IDbContextFactory pattern to create isolated DbContext instances
for each cache operation, making parallel verification thread-safe.
Changes:
- Add IDbContextFactory<ApplicationDbContext> registration
- Update CompanyVerifierService to use factory for cache operations
- Update tests with InMemoryDatabaseRoot for shared test data
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Features:
- Add UK institution recognition (170+ universities)
- Add diploma mill detection (100+ blacklisted institutions)
- Add education verification service with date plausibility checks
- Add local file storage option (no Azure required)
- Add default admin user seeding on startup
- Enhance Serilog logging with file output
Security fixes:
- Fix path traversal vulnerability in LocalFileStorageService
- Fix open redirect in login endpoint (use LocalRedirect)
- Fix password validation message (12 chars, not 6)
- Fix login to use HTTP POST endpoint (avoid Blazor cookie issues)
Code improvements:
- Add CancellationToken propagation to CV parser
- Add shared helpers (JsonDefaults, DateHelpers, ScoreThresholds)
- Add IUserContextService for user ID extraction
- Parallelized company verification in ProcessCVCheckJob
- Add 28 unit tests for education verification
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
