fh-library 0.2.19 → 0.2.21

package/CHANGELOG.md

@@ -1,5 +1,14 @@
 # FH Library — Change Log
 
+### v0.2.21 — 2026-11-09
+
+* **Cooldowns count from the moment:** a failed page, a not-found ID and a site's wall are remembered with the time, not the date — a bare date counted from midnight UTC, so a page that failed at 23:53 UTC was fetched up the whole ladder again 13 minutes later.
+
+### v0.2.20 — 2026-11-09
+
+* **`library_verify` without the reference list:** the decision model no longer gets the paper's References / Bibliography section — other papers' titles, which state nothing this paper found and could pass a claim on a cited title; 22.7% less text per full text (96 full texts measured). The stored text and `library_read` keep it.
+* **Less noise to the decision model:** a paper's citation line (authors, journal, PMID, DOI) is sent as its year only — 9.8% of an abstract.
+
 ### v0.2.19 — 2026-11-09
 
 * **Faster full texts:** a browser rung (or Bright Data) that met the same site's challenge twice in a row is skipped for that site for `FH_LIBRARY_RETRY_AFTER_DAYS` — the other rungs still run, a success clears it; the open-access lookups (Europe PMC, Crossref, Unpaywall, OpenAlex) start together instead of one after the other, the downloads stay in tier order.

package/README.md

@@ -1,4 +1,4 @@
-![Version 0.1.3](https://img.shields.io/badge/Version-0.2.19-green.svg)
+![Version 0.1.3](https://img.shields.io/badge/Version-0.2.21-green.svg)
 ![License MIT](https://img.shields.io/badge/License-MIT-yellow.svg)
 [![Forever Healthy](https://img.shields.io/badge/(c)_2026-Forever_Healthy-573D7D.svg)](https://forever-healthy.org)
 [![npm](https://img.shields.io/badge/npm-fh--library-cb3837.svg)](https://www.npmjs.com/package/fh-library)

package/dist/index.js

@@ -1815,6 +1815,7 @@ function emptyStatus(requested, type, id, libraryDir) {
   };
 }
 var today = () => (/* @__PURE__ */ new Date()).toISOString().slice(0, 10);
+var now = () => (/* @__PURE__ */ new Date()).toISOString().slice(0, 19) + "Z";
 function headLine(r) {
   return [
     r.authors.join(", ") || null,
@@ -2143,9 +2144,9 @@ function refreshRecord(p, m, r, xml) {
   writeMeta(p, next);
   return next;
 }
-function daysSince(date, now = /* @__PURE__ */ new Date()) {
+function daysSince(date, now2 = /* @__PURE__ */ new Date()) {
   const t = new Date(date).getTime();
-  return Number.isFinite(t) ? (now.getTime() - t) / 864e5 : Infinity;
+  return Number.isFinite(t) ? (now2.getTime() - t) / 864e5 : Infinity;
 }
 var FUNDING_RE = /^[*_\s]*(funding|financial support|sources? of (financial )?support|support(ed by)?|grant support|this (work|study|research) was (supported|funded))\b[^:]*:?[*_\s]*(.*)$/i;
 var COI_RE = /^[*_\s]*(conflicts? of interest|competing interests?|disclosures?|financial disclosures?|declarations? of (competing |conflicting )?interests?|conflict[- ]of[- ]interest statement)\b[^:]*:?[*_\s]*(.*)$/i;
@@ -2348,14 +2349,14 @@ function recordRung(url, rung2, ok, reason = "", dir = FH_LIBRARY_DIR) {
   if (ok) {
     if (!all[rung2]) return;
     delete all[rung2];
-  } else all[rung2] = { streak: (all[rung2]?.streak ?? 0) + 1, date: today(), reason: why };
+  } else all[rung2] = { streak: (all[rung2]?.streak ?? 0) + 1, date: now(), reason: why };
   try {
     mkdirSync3(join5(dir, "walls"), { recursive: true });
     writeAtomic(file(dir, h), JSON.stringify(all) + "\n");
   } catch {
   }
 }
-var wallNote = (r) => `${r.reason} ${r.streak}\xD7 on this host, last ${r.date}`;
+var wallNote = (r) => `${r.reason} ${r.streak}\xD7 on this host, last ${r.date.slice(0, 10)}`;
 
 // src/tiers/tier.ts
 var looked = (t, p, ctx) => ctx.lookups?.get(t.name) ?? t.lookup(p);
@@ -2537,9 +2538,9 @@ var MAX_WAIT_MS = 65e3;
 var TIMEOUT_MS3 = 1e4;
 var calls = [];
 async function pace() {
-  const now = Date.now();
-  while (calls.length && calls[0] <= now - WINDOW_MS) calls.shift();
-  if (calls.length >= PER_WINDOW) await new Promise((r) => setTimeout(r, calls[0] + WINDOW_MS - now + 200));
+  const now2 = Date.now();
+  while (calls.length && calls[0] <= now2 - WINDOW_MS) calls.shift();
+  if (calls.length >= PER_WINDOW) await new Promise((r) => setTimeout(r, calls[0] + WINDOW_MS - now2 + 200));
   calls.push(Date.now());
 }
 async function coreFetch(url, key) {
@@ -3011,13 +3012,13 @@ function parseAttempt(a) {
   return m ? { tier: m[1], reason: m[2], date: m[3] ?? null } : null;
 }
 var attemptLabel = (m, date = (/* @__PURE__ */ new Date()).toISOString().slice(0, 10)) => `${m.tier} (${m.reason}) ${date}`;
-function shouldSkip(attempts, tier, retryAfterDays, now = /* @__PURE__ */ new Date(), retryBlocked = false) {
+function shouldSkip(attempts, tier, retryAfterDays, now2 = /* @__PURE__ */ new Date(), retryBlocked = false) {
   for (const a of attempts) {
     const p = parseAttempt(a);
     if (!p || p.tier !== tier) continue;
     if (isPermanent(p.reason, retryBlocked)) return true;
     if (!p.date) return false;
-    const age = (now.getTime() - new Date(p.date).getTime()) / 864e5;
+    const age = (now2.getTime() - new Date(p.date).getTime()) / 864e5;
     return age < retryAfterDays;
   }
   return false;
@@ -3113,7 +3114,26 @@ function isLocal(baseUrl) {
   }
 }
 function cleanMarkdown(md) {
-  return md.replace(/!\[[^\]]*\]\([^)]*\)/g, "").replace(/\]\([^)]*\)/g, "]").replace(/^\s*#+\s*(see more about|more for you|back to top)\b.*$/gim, "").replace(/\n{3,}/g, "\n\n");
+  return paperYearOnly(dropReferences(md)).replace(/!\[[^\]]*\]\([^)]*\)/g, "").replace(/\]\([^)]*\)/g, "]").replace(/^\s*#+\s*(see more about|more for you|back to top)\b.*$/gim, "").replace(/\n{3,}/g, "\n\n");
+}
+var REF_HEAD = /^(#{1,6})\s*\**\s*(references|bibliography|literature cited|reference list|works cited|cited literature)\b[^\n]*$|^\**(references|bibliography|literature cited)\**:?[ \t]*$/gim;
+function dropReferences(md) {
+  let out = md;
+  for (; ; ) {
+    REF_HEAD.lastIndex = 0;
+    const m = REF_HEAD.exec(out);
+    if (!m) return out;
+    const level = m[1]?.length ?? 6;
+    const rest = out.slice(m.index + m[0].length);
+    const next = new RegExp(`^#{1,${level}}\\s`, "m").exec(rest);
+    out = out.slice(0, m.index) + (next ? rest.slice(next.index) : "");
+  }
+}
+function paperYearOnly(md) {
+  return md.replace(/^(#[^\n]*\n\n)([^\n]*· (?:PMID \d+|DOI 10\.)[^\n]*)$/m, (_, title, line) => {
+    const year = / · (\d{4})(?: · |$)/.exec(line)?.[1];
+    return year ? `${title}${year}` : title.trimEnd();
+  });
 }
 function windows(md, max) {
   const out = [];
@@ -3982,7 +4002,7 @@ async function libraryFetch(libraryDir, requested, args, converter, logs = {}, f
   if (!run) {
     run = (async () => {
       const r = await fetchImpl(url, logs.onTry, logs.onAlert);
-      const base = { url, fetched: today(), converted_by: converter };
+      const base = { url, fetched: now(), converted_by: converter };
       if (!r.ok) {
         const error = r.first && /^HTTP \d{3}/.test(r.first) ? r.first : r.reason;
         writeUrlMeta(p, { ...base, final_url: r.url ?? url, title: null, format: "failed", content_type: "", level: "title", retrieved_by: r.via, error, files: [`${p.entry}-meta.md`] });
@@ -4475,14 +4495,14 @@ function freshMissing(p, retryDays) {
   return m && daysSince(m.checked) < retryDays ? m : null;
 }
 function rememberMissing(p, reason) {
-  const m = { checked: today(), reason };
+  const m = { checked: now(), reason };
   mkdirSync10(p.entryDir, { recursive: true });
   writeAtomic(missingFile(p), `checked: ${m.checked}
 reason: ${m.reason}
 `);
   return m;
 }
-var missingNote = (m, retryDays) => `${m.reason} (checked ${m.checked}; asked again after ${retryDays} day${retryDays === 1 ? "" : "s"})`;
+var missingNote = (m, retryDays) => `${m.reason} (checked ${m.checked.slice(0, 10)}; asked again after ${retryDays} day${retryDays === 1 ? "" : "s"})`;
 
 // src/pubmedbatch.ts
 var MAX_PMIDS = 200;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "fh-library",
-  "version": "0.2.19",
+  "version": "0.2.21",
   "mcpName": "io.github.forever-healthy/fh-library",
   "description": "MCP Server that builds a local Markdown library of the scientific papers cited by evidence reviews — fetches metadata, abstract and full text per PubMed ID through a configurable cascade of retrieval services",
   "type": "module",