54yyyu/zotero-mcp v0.6.0

on GitHub

Highlights

Semantic search + research

• Passage-level chunking (opt-in) — index each item as overlapping passages with char/page provenance; search returns a grounded snippet and long PDFs stay searchable past the single-vector limit (#350).
• Agentic research tools — zotero_find_related_papers (OpenAlex citation graph), zotero_library_coverage, zotero_synthesize_annotations, zotero_export_bibliography, plus MCP prompts and resources (#350).
• Ollama embedding backend (#349) and OpenAI Batch API indexing for cheaper large-library embeds (#346).
• Embedding sub-batching + rate-limit throttle for provider input caps and 429 safety (#356).

Library management

• zotero_batch_update_extra for batch Extra-field edits (#334).
• Collection names/paths/keys resolved across all add paths; bad specs fail early instead of misfiling (#340).
• Idempotent adds via if_exists=duplicate|file|skip (#341).
• zotero-cli add isbn|bibtex|csl-json subcommands (#342).
• citation_key on zotero_update_item (#321); ZOTERO_WEBDAV_TIMEOUT env var (#345).

Fixes

• Incremental sync no longer skips newly-added items when the sqlite snapshot lags the API (#292/#333).
• update-db --fulltext no longer truncates to one vector per item (#290).
• zotero-cli add file TypeError fixed (#339); read_pdf WebDAV support (#351); scite /papers + DOI casing (#331); deterministic Gemini/OpenRouter embeddings and non-blocking db-status (#348).

Full details in CHANGELOG.md.