54yyyu/zotero-mcp v0.1.5 on GitHub

What's New in v0.1.5

Write operations — 10+ new tools: create_item, update_item, create_note, add_tags, batch_update_tags, create_collection, add_to_collection, remove_from_collection, add_by_doi, add_by_url, add_from_file (PR #165)
BetterBibTeX citation key lookup — zotero_search_by_citation_key searches both BetterBibTeX JSON-RPC and the Extra field
PDF outline extraction — zotero_get_pdf_outline returns table of contents from PDFs
Annotation page labels — zotero_get_annotations now includes annotationPageLabel and annotationPosition data
PDF timeout — configurable pdf_timeout (default 30s) skips slow PDFs during fulltext extraction
Semantic search quality — combined field+fulltext embeddings, Gemini retrieval_query/document fix, model-aware tokenizer, optional cross-encoder re-ranking
Abstracts in collection items — zotero_get_collection_items now includes abstracts (#143)

search_notes — fixed qmode and client-side filter to actually find notes (#137)
batch_update_tags — fixed stale tag set, response type check, hybrid local+web mode (#162)
get_tags pagination — uses zot.everything() for reliable tag retrieval (#70)
Fulltext truncation — removed hardcoded 10k/5k char caps, model-aware truncation via embedding_max_tokens (#153, #134)
Local mode file:// paths — resolves file://, absolute paths, and attachments: prefixes (#116)
Local-first fulltext extraction — prefers local DB/storage before remote dump() for file-backed attachments (PR #166)
Child notes — create_note properly attaches as child via web API in local mode (#133)
ChromaDB embedding conflict — auto-detects and resets collection on model change (#109)
FastMCP compatibility — removed deprecated dependencies parameter (#117, #61)
PDF outline import order — defers PyMuPDF import until after attachment check
--fulltext guard — aborts with clear error when used without ZOTERO_LOCAL (PR #156)
Update interval display — fixed misleading display for daily schedule (PR #144)