AI4L Library

Build or complete the library entry for one ID and return its status JSON. A drug label (dailymed:<set ID>) is fetched from DailyMed as the current prescribing information, a trial record (nct:<8 digits>, the NCT number's digits; nct:NCT… and a bare NCT… are accepted) from ClinicalTrials.gov as the current record with its posted results; both stored as a full text. A paper is a PMID, or a DOI (doi:10.…): a DOI PubMed indexes is added as its PMID's entry plus a copy under doi: (refreshed on every add of the DOI), one PubMed does not index gets its metadata, funders and update notices from Crossref and its abstract from Crossref, else OpenAlex. For a paper, pubmed gives metadata, abstract and the correction notices (errata, retractions, expressions of concern, updates → citable, comment, corrections); the full-text cascade (pmc, tdm, unpaywall, openalex, user url tiers) is then tried. Idempotent: a full_text entry is returned as is; an abstract entry retries only tiers whose earlier failure was transient and past the cooldown. An entry whose PubMed record is older than refresh_after_days gets the record re-read for new notices; abstract and full text are never touched.

Read-only status of the library entry for one paper ID: the same status JSON as library_add, or level "missing" when there is no entry. Never fetches.

Read one small piece of a stored paper's Markdown — never the whole paper. Without section/query: the outline (headings with word counts; says whether the full text or only the abstract is stored). section: a heading from the outline (case-insensitive substring, e.g. "results") → that section's text, capped at max_chars (default 6000). query: search terms → paragraphs containing all terms, each with its section, best first, at most max_hits (default 5); with no hit, the headings. Full texts converted from JATS XML have a "Declarations" section — funding, conflicts of interest, author notes — so section "Declarations" answers who paid and who declares what in one call; a paper without that section states nothing. For abstract-only entries the status carries coi and funding from PubMed. Call without section/query first, then pick. Reads may run in parallel.

Is one finding of a review stated by one source it cites — a paper, a drug label (dailymed:…) or a trial record (nct:…)? claim: the finding alone, verbatim from the review, without its citations ("IPSS fell 2.3 points more than placebo in men with ED"); id: one of the papers the review cites for it, which must be in the library (library_add first). A finding cited to three papers is three calls with the same claim. Build the claim as a complete statement: subject, intervention, outcome, number — when splitting a sentence leaves a fragment, restore the subject from the sentence; cut what the review adds around the finding (where it was published, who funded it, "in two trials", "pooled data", the review's own inference or advice after "so"); one number set per claim, so split at semicolons too. Statements about the literature as a whole ("no trial has tested"), recommendations and history are not claims. A claim that names a source, joins two findings, or is a fragment is refused. The server reads the paper and asks a decision model; nothing of the paper enters your context. verdict: "pass" (the paper states it; score is the probability), "fail" (its full text was read and does not state it), "na" (only its abstract is in the library and that does not state it, or only its title), or "error" (claim empty, too short, a fragment, naming a source or compound; paper not in the library; no model configured; model failure). basis: full_text, abstract or title. glyphs_unresolved: symbols the paper's PDF font gave no character for (a minus, ±, ≤ …) — when above 0, a fail may be the font, not the paper. superseded: true when the paper carries a stored correction ("Corrected in PubMed abstract") and the claim uses a figure that correction replaced; the verdict is kept, the figure is out of date; null for a paper without a stored correction.

Library-wide summary — everything the TUI shows: the library directory and host; the decision model with its endpoint and whether it answers; the entry count and the size on disk (size, size_bytes); per entry type (types) the entries and their levels (full_text, abstract, title) and retractions; entry counts by type and level, counts of retracted and not-citable entries and of entries with errata, expressions of concern and updates, the retrieval tiers in run order and which are off (tiers), the configuration (config: settings, problems, and per variable the source of its value — environment, .env or default — never the value), the decision model behind library_verify (model, endpoint, whether it runs on this machine, and whether it answers right now: ok with the answer time, or the error), the library directory and host. ids=true also lists every entry ID.

Get the running ai4l-library MCP server's package name and version, and the resolved library directory. Useful to confirm which build is loaded.

Takes no parameters.