fetch
ChatGPTFetch the authenticated user's full library dashboard — loans, holds, and purchase requests. Inputs: none (identity resolved from auth token). Output — UI widget result (mcp_res_ui): - loans (list): each item has {title, author, dueDate, renewable, barcode, loanId}. - requests (list): each item has {title, status, pickupLocation, expiryDate, requestId}. - purchase_requests (list): each item has {title, status, createdAt, isbn}. - total_loans / total_requests / total_purchase_requests (int): real totals from ALMA (may exceed items returned). Tags: read, account, high-latency.
fetch_account_page
ChatGPTInternal app-only pagination helper for the dashboard widget. Not intended for model use; public account access should go through fetch().
get_paper_details
ChatGPTFetch details for one or more papers by DOI or arXiv ID. Inputs: - paper_ids: list of DOIs or arXiv IDs. Examples: ["10.18653/v1/2020.acl-main.463", "arXiv:2106.15928"] Tags: read, research.
get_researcher_papers
ChatGPTGet a NYCU researcher's full publication list. Use when: user wants more papers from a specific professor found via search_researchers, or wants to filter a researcher's publications by year. Inputs: - researcher_id (str, required): professor UUID from search_researchers results (the portal_url contains it) - year_min (int|null, optional): filter papers published from this year - year_max (int|null, optional): filter papers published up to this year - limit (int, optional): max papers to return (1-100, default 20) Output — list of papers: title, year, DOI, type, subjects. Tags: read, research.
journal_search
ChatGPTSearch journals or browse a journal's latest issue. Provide exactly ONE of query or issn — never both, never neither. Mode 1 — Search (query): Find journals by title keyword. Returns up to 10 matches with title, ISSN, SJR score. Supports OR operator (e.g. "nature OR cell"). Prefer full/partial names over acronyms. Mode 2 — Latest issue (issn): Get the latest issue articles for a known journal. Use an ISSN from a previous search result. Returns up to 10 articles per page; use offset to paginate (increment by 10). Inputs: - query (str|null): keyword(s) to search journal titles. Mutually exclusive with issn. - issn (str|null): ISSN from a previous journal_search result. Mutually exclusive with query. - offset (int): pagination for latest-issue mode only. Default 0. Tags: read, journal.
journal_subscribe
ChatGPTManage journal subscriptions. Use ids from journal_search results. Inputs: - journal_ids: list of numeric journal ids (e.g. [123, 456]) - action: "subscribe" (default), "unsubscribe", or "check" Returns: - subscribe: subscribed_count, journal details, and optionally not_found ids. - unsubscribe: removed_count. - check: subscribed (list of journal ids the student is subscribed to).
journal_subscriptions
ChatGPTList all journals the current student has subscribed to. Use when: user asks "what journals am I following?" or "what's new in my subscribed journals?". Renders an interactive UI card listing subscribed journals with their latest issue articles.
paper_deep_dive
ChatGPTDeep-dive analysis of a single paper: its metadata, upstream influences (co-cited references), and downstream impact (citing papers and surveys). Use when: user wants to understand a paper's context — what it builds on, who cites it, and how it fits into the field. Inputs: - paper_id (str, required): a DOI, Semantic Scholar corpus ID, or arXiv ID. Examples: "10.18653/v1/2020.acl-main.463", "268536830", "arXiv:2106.15928" Output — dict with: - seed: {corpusid, title, abstract, year, authors} - upstream: list of influential references with {corpusid, title, year, venue, influentialcitationcount, tldr, cocitation_count} - downstream: {papers: [...], surveys: [...]} — citing works with overlap_ref count - summary: {n_cites_kept, weak_coupling, regime} Tags: read, research, high-latency.
search
ChatGPTSearch NYCU/Taiwan UST library catalog by keyword and display results as UI card. Use when: user wants to check library holdings, find a specific book, or see what's available on a topic in the NYCU catalog. Do NOT use as primary source for book recommendations. When user asks "recommend books on X": recommend from your own knowledge or web search first, then use this tool to check NYCU availability. Inputs: - query: keyword or phrase (Chinese/English). Supports AND / OR (uppercase). Prefer one OR query over multiple calls. - scope: "nycu" | "ust". Default "nycu". Prefer nycu; fall back to ust only if nycu returns no results. If results are empty or poor, retry with variations before giving up: - Switch query language (English ↔ Chinese) - Change search_by (e.g. "title" → "any", or try author name) - Fall back to scope="ust" Tags: read, catalog, high-latency.
search_academic_papers
ChatGPTSearch academic articles via the global paper index or NYCU Primo. Use when: user wants research-focused article results instead of general library catalog books/videos. Do NOT use as primary source for paper recommendations or topic overviews. When user asks "recommend papers on X" or "what are the key papers on Y": give an overview based on general knowledge, then use this tool to find/verify specific papers in NYCU. Use short natural-language keywords; no Boolean/operator syntax. For overview searches, append "survey". Inputs: - query (str, required): keyword or phrase - year_start (int|null, optional): filter results published from this year. If provided without year_end, year_end defaults to current year. - year_end (int|null, optional): filter results published up to this year. - scope (enum, optional): scope="taiwan" → searches Primo (broader humanities/social sciences coverage). Use when: topic is humanities, history, arts, literature, or social sciences; user wants Taiwan institution resources or NYCU thesis/ETD. scope="global" → searches academic papers with semantic or keyword matching. Default; use for natural sciences, engineering, CS, medicine. - search_mode (str, optional): search mode for global scope — e.g. "semantic", "keyword". Default "keyword". Output — UI widget (viewType "academic") with article metadata: title, year, venue, doi, citationCount, primo_id, open_url when available. Fields may be absent or null depending on the source. Taiwan results omit author; global results retain author metadata when supplied by the source. Tags: read, research, high-latency. After getting results, if closely relevant papers have a DOI or arXiv ID, call get_paper_details(paper_ids=[...]) to read their abstracts before responding. Skip papers without a supported identifier; do not invent one.
search_researchers
ChatGPTSearch NYCU faculty/researchers by name, research topic, or both. Use when: user wants to find professors, explore who works on a topic, or look up a researcher's recent publications at NYCU. Inputs: - name (str|null, optional): professor name (EN or ZH), substring match (e.g. "簡仁宗", "Chen") - topic (str|null, optional): research topic keyword, fulltext search on interests and papers (e.g. "machine learning") - limit (int, optional): max number of researchers to return (1-10, default 5) At least one of name or topic is required. Output — list of researchers, each with: name (EN/ZH), title, ORCID, research interests, portal URL, total matching papers, and up to 5 recent papers (title, year, DOI, subjects). Tags: read, research.