Prevent geographic false positives in cache lookups. Queries like
"musea in Amsterdam" vs "musea in Noord-Holland" have ~93%
embedding similarity but completely different answers.
Changes:
- Add ExtractedEntities interface for structured cache keys
- Implement fast entity extraction (<5ms, no LLM) with regex patterns
- Extract institution types (GLAMORCUBESFIXPHDNT), locations, and intent
- Generate structured cache keys (e.g., "count:M:amsterdam")
- Raise similarity threshold from 0.85 to 0.97 to match backend DSPy
- Add 'structured' match method to CacheLookupResult
The entity extractor recognizes:
- 19 institution types (Dutch + English patterns)
- 12 Dutch provinces with ISO 3166-2:NL codes
- Major Dutch cities with settlement codes
- Query intents (count, list, info)
This ensures geographic queries get different cache entries even when
embeddings are highly similar.
- auth.setup.ts: require env vars for test credentials (no hardcoded defaults)
- manifest.json: update schema manifest
- full_evaluation_results.json: add RAG evaluation results
- petra-links.json: update birth date from web claim
- Add chat.spec.ts for RAG query testing
- Add count-queries.spec.ts for aggregation validation
- Add map-panel.spec.ts for geographic feature testing
- Add cache.spec.ts for response caching verification
- Add auth.setup.ts for authentication handling
- Configure playwright.config.ts for multi-browser testing
- Tests run against production archief.support
- Add sparqlQuery field to CachedResponse interface
- Extract SPARQL query before cache storage (not after)
- Include sparqlQuery in cache HIT message objects
- Handle both snake_case (server) and camelCase field names
SPARQL queries are now displayed for both fresh API responses
and cached responses, improving debugging and transparency.
