kushi-agents 6.1.0 → 6.1.1

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "kushi-agents",
-  "version": "6.1.0",
+  "version": "6.1.1",
   "description": "Install Kushi — multi-source project evidence agent with Comprehensive Structured Capture (CSC) into weekly-only files across Email, Teams, OneNote, Loop, SharePoint, Meetings, CRM, ADO. Meetings retain a sibling verbatim/ audit folder. WorkIQ-only for M365 sources (Graph / m365_* FORBIDDEN as fallbacks; user-paste is first-class). Host-agnostic.",
   "type": "module",
   "bin": {

package/plugin/runners/discover.mjs

@@ -162,7 +162,7 @@ function isPlaceholder(v) {
   if (!s) return true;
   if (/^<.*>$/.test(s)) return true;              // <value>, <chat_id>, etc.
   if (/^turn\d+search\d+$/i.test(s)) return true; // WorkIQ web citation tokens
-  if (/^(unknown|n\/a|none|null|tbd|todo)$/i.test(s)) return true;
+  if (/^(unknown|n\/a|none|null|tbd|todo|not\s+explicitly\s+(available|provided|specified)|not\s+(available|provided|specified|applicable)|undisclosed)$/i.test(s)) return true;
   return false;
 }
 
@@ -178,19 +178,26 @@ function isValidValueFor(source, field, raw) {
     return true;
   }
   if (source === 'teams' && field === 'chat_id') {
-    // Graph chat IDs are long strings ending in '@thread.v2' or similar.
-    return v.includes('@thread') || v.length > 30;
+    // Accept Graph chat IDs ('@thread' / long opaque) OR human-readable
+    // chat topics — WorkIQ frequently returns the topic when the Graph ID
+    // is unavailable. Refresh stage will resolve to a real chat.
+    return v.length >= 2 && v.length <= 200;
   }
   if (source === 'meetings' && field === 'join_url') {
-    return /^https:\/\/teams\.microsoft\.com\/.+meetup-join/i.test(v);
+    // Prefer real Teams meetup URLs but also accept any https URL or
+    // a meeting subject — refresh stage will resolve.
+    if (v.startsWith('http')) return true;
+    return v.length >= 3 && v.length <= 200;
   }
   if (source === 'onenote' && field === 'section_file_id') {
-    // OneNote section file IDs are hex strings, typically prefixed `0-` and
-    // dozens of chars long. Reject short citation tokens.
-    return /^[0-9a-f][0-9a-f\-]{20,}$/i.test(v) || v.startsWith('0-');
+    // Accept hex section IDs OR section/page names. WorkIQ often only has
+    // the human-readable name; refresh resolves to the file ID.
+    return v.length >= 2 && v.length <= 200;
   }
   if (source === 'sharepoint' && field === 'site_url') {
-    return v.startsWith('https://') && v.includes('.sharepoint.com');
+    // Prefer real SharePoint URLs; also accept site names/relative paths.
+    if (v.startsWith('http')) return v.includes('.sharepoint.com') || v.includes('/sites/');
+    return v.length >= 2 && v.length <= 300;
   }
   if (source === 'crm' && (field === 'request_id' || field === 'incident_number')) {
     // CRM IDs vary: FE-2026-001458, REQ-12345, INC-9, plain numerics. Reject
@@ -228,7 +235,16 @@ function applyRows(source, rows, currentBounds, currentInteg) {
   }
   if (source === 'teams') {
     const existing = currentBounds.teams?.chats || [];
-    const incoming = rows.map(r => r.chat_id).filter(v => isValidValueFor('teams', 'chat_id', v));
+    // WorkIQ rarely has Graph chat IDs — citation tokens like "turn1search1"
+    // are common. Prefer chat_id; fall back to topic so refresh has a usable
+    // boundary descriptor instead of nothing.
+    const incoming = rows.map(r => {
+      const id = r.chat_id;
+      if (id && !isPlaceholder(id) && isValidValueFor('teams', 'chat_id', id)) return id;
+      const topic = r.topic;
+      if (topic && !isPlaceholder(topic)) return topic;
+      return null;
+    }).filter(Boolean);
     const merged = dedup([...existing, ...incoming]);
     const added = merged.filter(v => !existing.includes(v));
     if (added.length) accepted.push(...added);
@@ -236,7 +252,13 @@ function applyRows(source, rows, currentBounds, currentInteg) {
   }
   if (source === 'meetings') {
     const existing = currentBounds.meetings?.joinUrls || [];
-    const incoming = rows.map(r => r.join_url).filter(v => isValidValueFor('meetings', 'join_url', v));
+    const incoming = rows.map(r => {
+      const url = r.join_url;
+      if (url && !isPlaceholder(url) && isValidValueFor('meetings', 'join_url', url) && url.startsWith('http')) return url;
+      const subj = r.subject;
+      if (subj && !isPlaceholder(subj)) return subj;
+      return null;
+    }).filter(Boolean);
     const merged = dedup([...existing, ...incoming]);
     const added = merged.filter(v => !existing.includes(v));
     if (added.length) accepted.push(...added);
@@ -244,7 +266,16 @@ function applyRows(source, rows, currentBounds, currentInteg) {
   }
   if (source === 'onenote') {
     const existing = currentBounds.onenote?.section_file_ids || [];
-    const incoming = rows.map(r => r.section_file_id).filter(v => isValidValueFor('onenote', 'section_file_id', v));
+    // WorkIQ rarely has section_file_id (Graph property); usually returns
+    // citation tokens. Prefer real hex IDs; fall back to section_name so
+    // refresh has a usable boundary descriptor.
+    const incoming = rows.map(r => {
+      const id = r.section_file_id;
+      if (id && !isPlaceholder(id) && /^[0-9a-f][0-9a-f\-]{20,}$/i.test(String(id).trim())) return id;
+      const name = r.section_name;
+      if (name && !isPlaceholder(name)) return name;
+      return null;
+    }).filter(Boolean);
     const merged = dedup([...existing, ...incoming]);
     const added = merged.filter(v => !existing.includes(v));
     if (added.length) accepted.push(...added);
@@ -391,6 +422,16 @@ async function main() {
       rows = rowsFromBlocks(blocks, source);
       const elapsed = Date.now() - t0;
       const bytes = Buffer.byteLength(workiqStdout || '', 'utf8');
+      // Persist raw WorkIQ output for diagnosis. Lets users inspect why a
+      // source returned 0 accepted rows without re-running the query.
+      if (!args.dryRun) {
+        try {
+          const discoveryDir = path.join(aliasRoot(args.project, args.alias), '_discovery');
+          await fs.mkdir(discoveryDir, { recursive: true });
+          const header = `# discover ${source} @ ${new Date().toISOString()}\n# elapsed=${elapsed}ms bytes=${bytes} blocks=${blocks.length} rows=${rows.length}\n# prompt:\n${prompt.split('\n').map(l => '# ' + l).join('\n')}\n# --- workiq stdout ---\n`;
+          await fs.writeFile(path.join(discoveryDir, `${source}-raw.txt`), header + (workiqStdout || ''), 'utf8');
+        } catch { /* best-effort */ }
+      }
       if (rows.length === 0 && bytes < 8) {
         // Distinguish "workiq returned empty" from "timeout" — both used to look the same.
         skipReason = 'workiq-empty-response';

package/plugin/runners/lib/workiq.mjs

@@ -89,13 +89,16 @@ export async function ask(prompt, { bin, timeoutMs = 120_000, env = process.env,
 export function parseCscBlocks(text) {
   if (!text || typeof text !== 'string') return [];
   const blocks = [];
-  const blockquoteRe = /(^|\n)>\s*\[block:\s*([a-zA-Z0-9_.-]+)\]\s*\n((?:>\s*[^\n]*\n?)+)/g;
+  // Accept either `>` (blockquote) or `|` (table-row style) line prefixes —
+  // WorkIQ emits both depending on rendering. Trailing two-space markdown line
+  // breaks are tolerated by the per-line strip.
+  const blockquoteRe = /(^|\n)[>|]\s*\[block:\s*([a-zA-Z0-9_.-]+)\]\s*\n((?:[>|]\s*[^\n]*\n?)+)/g;
   let m;
   while ((m = blockquoteRe.exec(text)) !== null) {
     const name = m[2];
     const body = m[3]
       .split('\n')
-      .map(l => l.replace(/^>\s?/, ''))
+      .map(l => l.replace(/^[>|]\s?/, '').replace(/\s+$/, ''))
       .filter(l => l.length > 0)
       .join('\n');
     blocks.push({ name, raw: body, fields: parseKvLines(body) });