jmiba/zotero-redisearch-rag: 1.0.10

• Adopt Obsidian's declarative settings API:
  • expose the settings as six native sub-pages with searchable control names and descriptions,
  • keep dynamic provider, model, annotation-map, and maintenance actions inside searchable declarative rows,
  • require Obsidian 1.13.0 or newer and compile against the released 1.13.1 API typings.
• Keep Redis identity stable when a vault moves:
  • persist the legacy vault-derived namespace in plugin settings,
  • add a Maintenance → Redis indexing namespace control that can reconnect a moved vault to existing indexes and keys without renaming or deleting Redis data.
• Restore automatic missing-index creation across Redis Search error variants:
  • centralize detection for SEARCH_INDEX_NOT_FOUND, Index not found, No such index, and the legacy Unknown index name,
  • use the shared detector for both index creation and idempotent drop/rebuild,
  • add regression tests proving FT.CREATE runs for every supported missing-index response while unrelated Redis errors remain fatal.
• Harden Paddle API OCR failures:
  • retry temporary PaddleOCR-VL queue saturation (errorCode 10010 / HTTP 503) twice with bounded backoff,
  • keep explicitly selected Paddle API engines strict instead of falling through to native local Paddle,
  • report native worker exits such as SIGSEGV ahead of generic Paddle error headers and retain non-JSON stdout diagnostics in the import log.

Full Changelog: https://github.com/jmiba/zotero-redisearch-rag/compare/1.0.9...1.0.10