Skip to content

v4.3.0: connector ingestion and managed assets

Choose a tag to compare

View all tags
@HabermannR HabermannR released this 12 Aug 19:45
· 11 commits to master since this release
v4.3.0
24f2276

Highlights

Nexidion 4.3.0 introduces the connector-based ingestion foundation, a complete PDF import workflow, managed image
assets, and clearer separation between human users, AI execution, and source provenance.

Connector ingestion

  • Added a connector/plugin registry and installation model.
  • Connectors can be configured for read, ingest, or both capabilities.
  • Added connector management through both the UI and REST API.
  • Added source artifacts, ingestion runs, execution identity, and provenance metadata.
  • Provides the foundation for future SharePoint, wiki, Abaqus HTML, and other readers.

PDF import and AI organization

PDFs now support three modes:

  1. Import — mechanically creates source-backed nodes.
  2. Import and organize — preserves imported nodes and adds AI-created synthesis nodes.
  3. Organize only — creates the AI-organized hierarchy without retaining mechanical page/chapter nodes.

Additional improvements:

  • Supports OpenAI and local OpenAI-compatible LLMs.
  • Improved hierarchy prompts and groups generated material below a single import root.
  • Extracted PDF images are stored and rendered as managed assets.
  • Added AI-generated backend summaries with explicit provenance.
  • Added language, tags, authority, content kind, source, and curation metadata.

Managed images and migration

  • Replaced the legacy shared image endpoint with vault-scoped managed assets.
  • Validates uploaded image formats and enforces vault permissions.
  • Added flask convert-legacy-images with dry-run and apply modes.
  • Legacy SVG/XHTML diagrams are sanitized and rasterized to PNG rather than served as active SVG.
  • Conversion preserves node history and is idempotent.

Security and identity

  • AI work can be attributed to a dedicated non-human assistant account.
  • Private-node filtering was tightened in agent context and full-text search.
  • Connector credentials are referenced rather than stored directly in connector configuration.
  • Docker build contexts now exclude secrets, backups, local environments, archives, and legacy image storage.
  • Local LLM access is opt-in in Docker instead of silently assuming a host endpoint.

Other changes

  • Demo mode and bundled demo data were removed.
  • Updated frontend dependencies and switched container builds to reproducible npm ci.
  • Added connector, PDF, curation, image asset, migration, provenance, and security tests.

Validation

  • 271 backend tests passed.
  • Frontend production build passed.
  • npm audit: 0 vulnerabilities.
  • Multi-architecture Docker build: AMD64 and ARM64.
  • Database migrations and legacy-image conversion were tested against a restored Raspberry Pi production backup.

Upgrade notes

  1. Back up PostgreSQL, legacy secure_images, managed asset storage, configuration, and MCP state.
  2. Run the database migrations with flask db upgrade.
  3. Mount persistent storage at ASSET_STORAGE_FOLDER.
  4. Dry-run image migration:
    flask convert-legacy-images /path/to/secure_images
  5. If the report has no missing references, apply it:
    flask convert-legacy-images /path/to/secure_images --apply
  6. Keep the legacy image backup until all converted images have been visually verified.

Demo mode is no longer available in this release.

Assets 2
Loading