hypomnema 1.1.0 → 1.2.0

package/hooks/hypo-lookup.mjs

@@ -14,8 +14,8 @@ import { join, basename } from 'path';
 import { HYPO_DIR, buildOutput, loadHypoIgnore, isIgnored } from './hypo-shared.mjs';
 
 const INDEX_PATH = join(HYPO_DIR, 'index.md');
-const MAX_HITS   = 3;
-const MAX_CHARS  = 2000;
+const MAX_HITS = 3;
+const MAX_CHARS = 2000;
 
 // ── helpers ─────────────────────────────────────────────────────────────────
 
@@ -38,51 +38,120 @@ function buildPageMap(dir, root = dir, map = {}, ignorePatterns = [], hypoDir =
 
 function extractKeywords(prompt) {
   const stop = new Set([
-    'the','and','for','with','this','that','have','from','are','was','were',
-    'what','when','where','how','why','who','can','could','should','would',
-    'does','did','will','not','but','its','also','just','more','any','all',
-    '어떤','어떻게','무엇','이런','그런','하는','하고','해서','있어','없어',
-    '되는','이거','저거','그거','이건','그건','저건','같은','하면','되면',
-    '인지','에서','으로','까지','부터','에게','한테','에도','에만','에는',
+    'the',
+    'and',
+    'for',
+    'with',
+    'this',
+    'that',
+    'have',
+    'from',
+    'are',
+    'was',
+    'were',
+    'what',
+    'when',
+    'where',
+    'how',
+    'why',
+    'who',
+    'can',
+    'could',
+    'should',
+    'would',
+    'does',
+    'did',
+    'will',
+    'not',
+    'but',
+    'its',
+    'also',
+    'just',
+    'more',
+    'any',
+    'all',
+    '어떤',
+    '어떻게',
+    '무엇',
+    '이런',
+    '그런',
+    '하는',
+    '하고',
+    '해서',
+    '있어',
+    '없어',
+    '되는',
+    '이거',
+    '저거',
+    '그거',
+    '이건',
+    '그건',
+    '저건',
+    '같은',
+    '하면',
+    '되면',
+    '인지',
+    '에서',
+    '으로',
+    '까지',
+    '부터',
+    '에게',
+    '한테',
+    '에도',
+    '에만',
+    '에는',
   ]);
-  return [...new Set(
-    prompt.toLowerCase()
-      .split(/[\s,，.。?？!！()\[\]{}'"\/\\:;=+*&%$#@~`|<>]+/)
-      .filter(w => w.length >= 3 && !stop.has(w))
-  )];
+  return [
+    ...new Set(
+      prompt
+        .toLowerCase()
+        .split(/[\s,，.。?？!！()\[\]{}'"\/\\:;=+*&%$#@~`|<>]+/)
+        .filter((w) => w.length >= 3 && !stop.has(w)),
+    ),
+  ];
 }
 
 function tokenize(text) {
-  return text.toLowerCase()
+  return text
+    .toLowerCase()
     .split(/[\s\-_/.,，。?？!！()\[\]{}'"\\:;=+*&%$#@~`|<>]+/)
-    .filter(w => w.length >= 2);
+    .filter((w) => w.length >= 2);
 }
 
 function bm25Score(queryTerms, entries, k1 = 1.5, b = 0.75) {
   const N = entries.length;
   if (N === 0) return [];
-  const docTokens = entries.map(e => tokenize(e.slug + ' ' + e.desc));
+  const docTokens = entries.map((e) => tokenize(e.slug + ' ' + e.desc));
   const avgdl = docTokens.reduce((s, t) => s + t.length, 0) / N;
   const df = {};
   for (const tokens of docTokens) {
     for (const t of new Set(tokens)) df[t] = (df[t] || 0) + 1;
   }
-  const idf = t => Math.log(1 + (N - (df[t] || 0) + 0.5) / ((df[t] || 0) + 0.5));
-
-  return entries.map((e, i) => {
-    const tokens = docTokens[i];
-    const dl = tokens.length || 1;
-    const tf = {};
-    for (const t of tokens) tf[t] = (tf[t] || 0) + 1;
-    let score = 0;
-    for (const q of queryTerms) {
-      const f = tf[q] || 0;
-      if (f === 0) continue;
-      const norm = 1 - b + b * dl / avgdl;
-      score += idf(q) * (f * (k1 + 1)) / (f + k1 * norm);
-    }
-    return { ...e, score };
-  }).sort((a, c) => c.score - a.score);
+  const idf = (t) => Math.log(1 + (N - (df[t] || 0) + 0.5) / ((df[t] || 0) + 0.5));
+
+  return entries
+    .map((e, i) => {
+      const tokens = docTokens[i];
+      const dl = tokens.length || 1;
+      const tf = {};
+      for (const t of tokens) tf[t] = (tf[t] || 0) + 1;
+      let score = 0;
+      for (const q of queryTerms) {
+        const f = tf[q] || 0;
+        if (f === 0) continue;
+        const norm = 1 - b + (b * dl) / avgdl;
+        score += (idf(q) * (f * (k1 + 1))) / (f + k1 * norm);
+      }
+      return { ...e, score };
+    })
+    .sort((a, c) => c.score - a.score);
+}
+
+function typePrior(slug) {
+  if (/\/decisions\/|^decisions\//.test(slug)) return 1.5;
+  if (/\bprd\b|spec-v/.test(slug)) return 1.3;
+  if (/\/session-log\/|\/session-log$/.test(slug)) return 1.2;
+  return 1.0;
 }
 
 function parseIndexEntries(indexContent) {
@@ -92,7 +161,7 @@ function parseIndexEntries(indexContent) {
     if (line.trimStart().startsWith('>')) continue;
     const m = line.match(/\[\[([^\]]+)\]\]\s*[—\-]+\s*(.+)/);
     if (!m) continue;
-    const raw  = m[1].trim();
+    const raw = m[1].trim();
     const desc = m[2].trim();
     const slug = raw.includes('|') ? raw.split('|')[0].trim() : raw;
     entries.push({ slug, desc });
@@ -104,10 +173,10 @@ function parseIndexEntries(indexContent) {
 
 let input = '';
 process.stdin.setEncoding('utf-8');
-process.stdin.on('data', chunk => input += chunk);
+process.stdin.on('data', (chunk) => (input += chunk));
 process.stdin.on('end', () => {
   try {
-    const data   = JSON.parse(input);
+    const data = JSON.parse(input);
     const prompt = (data.prompt || '').trim();
 
     if (!prompt || !existsSync(INDEX_PATH)) {
@@ -121,58 +190,95 @@ process.stdin.on('end', () => {
       return;
     }
 
-    const entries  = parseIndexEntries(readFileSync(INDEX_PATH, 'utf-8'));
-    const scored   = bm25Score(keywords, entries).filter(e => e.score > 0);
+    const entries = parseIndexEntries(readFileSync(INDEX_PATH, 'utf-8'));
+    const scored = bm25Score(keywords, entries)
+      .map((e) => ({ ...e, score: e.score * typePrior(e.slug) }))
+      .sort((a, c) => c.score - a.score)
+      .filter((e) => e.score > 0);
     const topScore = scored[0]?.score ?? 0;
-    const matched  = scored.filter(e => e.score >= topScore * 0.5);
+    const matched = scored.filter((e) => e.score >= topScore * 0.5);
 
     if (matched.length === 0) {
-      const topic   = keywords.slice(0, 5).join(', ');
-      const closest = bm25Score(keywords, entries).slice(0, 3).map(e => `[[${e.slug}]]`).join(', ');
-      console.log(JSON.stringify(
-        buildOutput(
-          `[WIKI LOOKUP: miss] "${topic}" — no match. Closest: ${closest || 'none'}`,
-          { continue: true, suppressOutput: true }
-        )
-      ));
+      const topic = keywords.slice(0, 5).join(', ');
+      const closest = bm25Score(keywords, entries)
+        .map((e) => ({ ...e, score: e.score * typePrior(e.slug) }))
+        .sort((a, c) => c.score - a.score)
+        .slice(0, 3)
+        .map((e) => `[[${e.slug}]]`)
+        .join(', ');
+      console.log(
+        JSON.stringify(
+          buildOutput(`[WIKI LOOKUP: miss] "${topic}" — no match. Closest: ${closest || 'none'}`, {
+            continue: true,
+            suppressOutput: true,
+          }),
+        ),
+      );
       return;
     }
 
     const ignorePatterns = loadHypoIgnore(HYPO_DIR);
     const pageMap = {
-      ...buildPageMap(join(HYPO_DIR, 'pages'), join(HYPO_DIR, 'pages'), {}, ignorePatterns, HYPO_DIR),
-      ...buildPageMap(join(HYPO_DIR, 'projects'), join(HYPO_DIR, 'projects'), {}, ignorePatterns, HYPO_DIR),
+      ...buildPageMap(
+        join(HYPO_DIR, 'pages'),
+        join(HYPO_DIR, 'pages'),
+        {},
+        ignorePatterns,
+        HYPO_DIR,
+      ),
+      ...buildPageMap(
+        join(HYPO_DIR, 'projects'),
+        join(HYPO_DIR, 'projects'),
+        {},
+        ignorePatterns,
+        HYPO_DIR,
+      ),
     };
 
     const injected = [];
     for (const { slug } of matched.slice(0, MAX_HITS)) {
-      const path = pageMap[slug]
-        ?? pageMap[slug.replace(/^(pages|projects)\//, '')]
-        ?? pageMap[basename(slug)];
+      const path =
+        pageMap[slug] ??
+        pageMap[slug.replace(/^(pages|projects)\//, '')] ??
+        pageMap[basename(slug)];
       if (path && existsSync(path)) {
         injected.push(`=== [[${slug}]] ===\n${readFileSync(path, 'utf-8').slice(0, MAX_CHARS)}`);
       }
     }
 
     if (injected.length === 0) {
-      const slugs = matched.slice(0, MAX_HITS).map(e => e.slug).join(', ');
-      console.log(JSON.stringify(
-        buildOutput(`[WIKI LOOKUP: index hit but files missing] ${slugs}`, { continue: true, suppressOutput: true })
-      ));
+      const slugs = matched
+        .slice(0, MAX_HITS)
+        .map((e) => e.slug)
+        .join(', ');
+      console.log(
+        JSON.stringify(
+          buildOutput(`[WIKI LOOKUP: index hit but files missing] ${slugs}`, {
+            continue: true,
+            suppressOutput: true,
+          }),
+        ),
+      );
       return;
     }
 
-    const overflow = matched.length > MAX_HITS
-      ? `\n(+${matched.length - MAX_HITS} more matches — search wiki index for more)` : '';
+    const overflow =
+      matched.length > MAX_HITS
+        ? `\n(+${matched.length - MAX_HITS} more matches — search wiki index for more)`
+        : '';
 
-    console.log(JSON.stringify(
-      buildOutput(
-        `[WIKI LOOKUP: ${injected.length} page(s) matched]\n\n` + injected.join('\n\n') + overflow,
-        { continue: true, suppressOutput: true }
-      )
-    ));
-
-  } catch {
+    console.log(
+      JSON.stringify(
+        buildOutput(
+          `[WIKI LOOKUP: ${injected.length} page(s) matched]\n\n` +
+            injected.join('\n\n') +
+            overflow,
+          { continue: true, suppressOutput: true },
+        ),
+      ),
+    );
+  } catch (err) {
+    process.stderr.write(`[hypo-lookup] error: ${err?.message ?? String(err)}\n`);
     console.log(JSON.stringify({ continue: true, suppressOutput: true }));
   }
 });