Every page refresh feeds the learning loop — contracts logged as playbook entries

Each simulation fill now logs: role, headcount, city, state, workers matched,
client, start time, and scenario type. One page refresh = ~20 playbook entries.
4 refreshes = 28 entries with patterns already forming.

Fixed activity counters: shows Contract Fills, Searches, and Patterns.
Activity feed now shows the actual fill data with worker names and scenarios.

This is the PRD's learning loop in action — the system records every
successful match so future queries can learn from past decisions.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

mcp-server/index.ts

@@ -1047,23 +1047,25 @@ tr:hover{background:#111827}
         // Intelligence: Activity feed — what the system has learned
         if (url.pathname === "/intelligence/activity" && req.method === "POST") {
           const start = Date.now();
-          const [playbooksR, searchCountR, simCountR] = await Promise.all([
+          const [playbooksR, searchCountR, fillCountR, totalR] = await Promise.all([
             api("POST", "/query/sql", { sql: "SELECT * FROM successful_playbooks ORDER BY timestamp DESC LIMIT 20" }).catch(() => ({ rows: [] })),
             api("POST", "/query/sql", { sql: "SELECT COUNT(*) cnt FROM successful_playbooks WHERE operation LIKE 'search:%'" }).catch(() => ({ rows: [{ cnt: 0 }] })),
-            api("POST", "/query/sql", { sql: "SELECT COUNT(*) cnt FROM successful_playbooks WHERE operation LIKE 'week_simulation%'" }).catch(() => ({ rows: [{ cnt: 0 }] })),
+            api("POST", "/query/sql", { sql: "SELECT COUNT(*) cnt FROM successful_playbooks WHERE operation LIKE 'fill:%'" }).catch(() => ({ rows: [{ cnt: 0 }] })),
+            api("POST", "/query/sql", { sql: "SELECT COUNT(*) cnt FROM successful_playbooks" }).catch(() => ({ rows: [{ cnt: 0 }] })),
           ]);
-          // Extract learned patterns — which workers get picked for which queries
+          // Extract learned patterns — which roles+cities get filled most
           const patterns: Record<string, number> = {};
           for (const p of (playbooksR.rows || [])) {
-            if (p.operation?.startsWith("search:")) {
+            if (p.operation?.startsWith("fill:") || p.operation?.startsWith("search:")) {
-              const key = p.operation.replace("search: ", "").trim();
+              const key = p.operation.replace(/^(fill|search): ?/, "").trim();
               patterns[key] = (patterns[key] || 0) + 1;
             }
           }
           return ok({
             playbooks: playbooksR.rows || [],
             search_count: searchCountR.rows?.[0]?.cnt || 0,
-            sim_count: simCountR.rows?.[0]?.cnt || 0,
+            fill_count: fillCountR.rows?.[0]?.cnt || 0,
+            total_operations: totalR.rows?.[0]?.cnt || 0,
             learned_patterns: Object.entries(patterns).map(([q, c]) => ({ query: q, times: c })).sort((a, b) => b.times - a.times),
             duration_ms: Date.now() - start,
           });
@@ -1452,12 +1454,28 @@ async function runWeekSimulation() {
     playbook_entries: playbookEntries,
   };
 
-  // Log the week to playbooks
+  // Log every filled contract as a playbook entry — this is the training data
   try {
-    const form = new FormData();
+    const ts = new Date().toISOString();
-    const csv = `timestamp,operation,approach,result,context\n"${new Date().toISOString()}","week_simulation: ${summary.total_contracts} contracts over 5 days","hybrid SQL+vector with multi-model routing","${summary.total_filled}/${summary.total_needed} filled (${summary.fill_pct}%)","${summary.emergencies} emergencies, ${summary.handoffs} handoffs"`;
+    const rows: string[] = [];
-    form.append("file", new Blob([csv], { type: "text/csv" }), "playbook.csv");
+    for (const day of results) {
-    await fetch(`${BASE}/ingest/file?name=successful_playbooks`, { method: "POST", body: form });
+      for (const c of day.contracts) {
+        if (c.matches && c.matches.length > 0) {
+          const workerNames = c.matches.slice(0, 3).map((m: any) => m.name || m.doc_id).join(", ");
+          const op = `fill: ${c.role} x${c.headcount} in ${c.city}, ${c.state}`;
+          const approach = `${c.situation} (${c.priority}) → hybrid search`;
+          const result = `${c.filled}/${c.headcount} filled → ${workerNames}`;
+          const context = `client=${c.client} start=${c.start} scenario=${c.situation}`;
+          rows.push(`"${ts}","${op.replace(/"/g,'""')}","${approach}","${result.replace(/"/g,'""')}","${context.replace(/"/g,'""')}"`);
+        }
+      }
+    }
+    if (rows.length) {
+      const csv = `timestamp,operation,approach,result,context\n${rows.join("\n")}`;
+      const form = new FormData();
+      form.append("file", new Blob([csv], { type: "text/csv" }), "playbook.csv");
+      await fetch(`${BASE}/ingest/file?name=successful_playbooks`, { method: "POST", body: form });
+    }
   } catch {}
 
   return { days: results, summary };

mcp-server/search.html

@@ -614,7 +614,7 @@ function loadLearning(){
   api('/intelligence/activity',{}).then(function(d){
     var el=document.getElementById('learning');
     el.textContent='';
-    var total=(d.search_count||0)+(d.sim_count||0);
+    var total=d.total_operations||0;
     if(total===0&&(!d.playbooks||!d.playbooks.length))return; // nothing to show yet
 
     var card=document.createElement('div');card.className='insight info';
@@ -626,9 +626,9 @@ function loadLearning(){
 
     // Stats row
     var stats=document.createElement('div');stats.style.cssText='display:flex;gap:16px;margin-bottom:12px';
-    addLearnStat(stats,d.search_count||0,'Searches Logged','#58a6ff');
+    addLearnStat(stats,d.fill_count||0,'Contract Fills','#3fb950');
-    addLearnStat(stats,d.sim_count||0,'Simulations Run','#3fb950');
+    addLearnStat(stats,d.search_count||0,'Searches','#58a6ff');
-    addLearnStat(stats,(d.learned_patterns||[]).length,'Patterns Detected','#bc8cff');
+    addLearnStat(stats,(d.learned_patterns||[]).length,'Patterns','#bc8cff');
     card.appendChild(stats);
 
     // Learned patterns