docs: update changelog for v0.12.0-library-desk-enhancements

🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-10 21:44:20 +01:00
parent 4b9092d631
commit 40aafb0331
1 changed files with 62 additions and 0 deletions
@@ -12,6 +12,68 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Authentik SSO Milestones 4-5: Protect remaining services (deferred)
 - Disaster recovery and offsite backup strategy

+## [0.12.0-library-desk-enhancements] - 2025-12-10
+
+### Added
+- **Wiki Change Detection** - Real-time page change notifications via PostgreSQL LISTEN/NOTIFY
+  - WikiChangeListener service connects to Wiki.js database
+  - Automatic re-ingestion when pages are created/updated/deleted
+  - Debouncing and loop prevention for automated updates
+  - Webhook router as HTTP fallback mechanism
+  - Setup scripts and documentation for PostgreSQL triggers
+
+- **Taxonomy-Aware Classification** - LLM uses existing wiki structure when classifying content
+  - `get_taxonomy_structure()` extracts category/subcategory paths from wiki
+  - Consolidation prompts include existing paths to prefer over creating new ones
+  - Prevents duplicate category creation (e.g., reuses `reference/political-entities/` instead of creating `reference/military-alliances/`)
+
+- **WikiJS Client Methods**
+  - `list_all_pages()` - Fetch all pages with path prefix filter (replaces stale search index)
+  - `get_taxonomy_structure()` - Extract category hierarchy for a user namespace
+
+- **Test Coverage** - New test files for recent features
+  - `test_graph_service.py` - Document node tags, entity-stub skipping
+  - `test_ingestion.py` - list_all_pages usage, batch ingestion
+  - `test_wiki_change_listener.py` - PostgreSQL notification handling
+  - Updated `test_consolidation.py` with taxonomy and ingestion_service tests
+  - Updated `test_integration.py` with list_all_pages and taxonomy tests
+
+### Fixed
+- **Ingestion Pipeline** - Pages created during consolidation now properly indexed
+  - `ingestion_service` was not passed to ConsolidationService in router
+  - Added ingestion call to `_update_page_with_facts()` (was only in create path)
+
+- **Bulk Re-index** - `/ingest/all` endpoint now works reliably
+  - Changed from `search_pages("")` (stale index) to `list_all_pages()`
+  - Successfully re-indexed 95 pages
+
+- **Document Node Tags** - Neo4j Document nodes now include `tags` property
+  - Prevents warnings in related documents query
+  - Set via `d.tags = $tags` in MERGE query
+
+- **WikiJS Integration Script** - Page ID fetched via GraphQL
+  - Uses `pages.singleByPath(path, locale)` query during initialization
+  - Replaces unreliable page list search method
+
+### Changed
+- **Entity Linking** - Improved matching with fuzzy logic
+  - Confidence scoring for containment and token overlap matches
+  - Self-referential link filtering (entities don't link to current page)
+  - Path preservation (keeps full `/users/username/path` format)
+  - Longest-first matching to prevent partial matches
+
+- **Consolidation Processing** - Searches marked processed even when skipped/errored
+  - Prevents unprocessed searches from accumulating indefinitely
+
+- **Stack Configuration** - Updated library-desk environment variables
+  - Replaced `WIKIJS_API_KEY` with `WIKIJS_USERNAME`/`WIKIJS_PASSWORD`
+  - Added `WIKIJS_DB_PASSWORD` for PostgreSQL connection
+
+### Technical Details
+- **Dependencies:** Added `asyncpg~=0.29.0` for PostgreSQL async driver
+- **New Files:** 10 files added (services, routers, tests, documentation)
+- **Commits:** 10 logical commits covering all changes
+
 ## [0.11.0-pydantic-ai-cleanup] - 2025-12-03

 ### Major Changes