Add AI reasoning history to CH member detail panel

- New ch_ai_decisions table stores LLM responses and parsed orders
- AI trader logs every decision: LLM response, fallback, invalid parse
- Member detail API returns last 10 AI decisions
- Detail panel shows AI Reasoning section with:
- LLM/Fallback/Invalid badges
- Parsed order (side, qty, symbol, price)
- Raw LLM response in scrollable monospace box

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

clearing_house/app.py

@@ -371,6 +371,16 @@ def api_member_detail(member_id):
     total_value = round(member["capital"] + total_holdings_value, 2)
     total_pnl = round(total_value - db.CH_STARTING_CAPITAL, 2)
 
+    ai_decisions = db.get_ai_decisions(member_id, limit=10)
+    # Parse the stored JSON strings for the frontend
+    import json as _json
+    for d in ai_decisions:
+        if d.get("parsed_order") and isinstance(d["parsed_order"], str):
+            try:
+                d["parsed_order"] = _json.loads(d["parsed_order"])
+            except Exception:
+                pass
+
     return jsonify({
         "member_id": member_id,
         "capital": round(member["capital"], 2),
@@ -380,6 +390,7 @@ def api_member_detail(member_id):
         "pnl": total_pnl,
         "daily": daily,
         "trades": trades,
+        "ai_decisions": ai_decisions,
         "is_human": ai_trader.is_human_active(member_id),
         "obligation": db.CH_DAILY_OBLIGATION,
     })

clearing_house/ch_ai_trader.py

@@ -268,9 +268,24 @@ def _decide_order_llm(
     if text:
         order = _parse_llm_order(text, bbos)
         if order and _validate_order(order, capital, holdings, bbos):
+            try:
+                db.record_ai_decision(member_id, text, order, source="llm")
+            except Exception as e:
+                print(f"[CH-AI] Failed to log decision: {e}")
             return order
+        # Log failed parse/validation too
+        try:
+            db.record_ai_decision(member_id, text, order, source="llm-invalid")
+        except Exception:
+            pass
     # Fallback: rule-based
-    return _fallback_order(capital, holdings, bbos)
+    fallback = _fallback_order(capital, holdings, bbos)
+    if fallback:
+        try:
+            db.record_ai_decision(member_id, "LLM unavailable, using rule-based fallback", fallback, source="fallback")
+        except Exception:
+            pass
+    return fallback
 
 
 def _build_prompt(member_id, capital, holdings, daily_trades, bbos, obligation_remaining):

clearing_house/ch_database.py

@@ -65,10 +65,22 @@ CREATE TABLE IF NOT EXISTS ch_settlements (
     settled_at      REAL NOT NULL
 );
 
+CREATE TABLE IF NOT EXISTS ch_ai_decisions (
+    id           INTEGER PRIMARY KEY AUTOINCREMENT,
+    member_id    TEXT NOT NULL,
+    llm_response TEXT NOT NULL,
+    parsed_order TEXT,
+    source       TEXT NOT NULL DEFAULT 'llm',
+    trading_date TEXT NOT NULL,
+    timestamp    REAL NOT NULL
+);
+
 CREATE INDEX IF NOT EXISTS idx_ch_trade_log_member
     ON ch_trade_log(member_id, trading_date);
 CREATE INDEX IF NOT EXISTS idx_ch_settlements_member
     ON ch_settlements(member_id, trading_date);
+CREATE INDEX IF NOT EXISTS idx_ch_ai_decisions_member
+    ON ch_ai_decisions(member_id, trading_date);
 """
 
 
@@ -264,6 +276,41 @@ def get_settlements(member_id: str, limit: int = 30) -> list[dict]:
     return [dict(r) for r in rows]
 
 
+# ── AI Decisions ──────────────────────────────────────────────────────────────
+
+def record_ai_decision(
+    member_id: str,
+    llm_response: str,
+    parsed_order: dict | None,
+    source: str = "llm",
+) -> None:
+    import json as _json
+    _conn().execute(
+        """INSERT INTO ch_ai_decisions
+           (member_id, llm_response, parsed_order, source, trading_date, timestamp)
+           VALUES (?,?,?,?,?,?)""",
+        (
+            member_id,
+            llm_response,
+            _json.dumps(parsed_order) if parsed_order else None,
+            source,
+            today_str(),
+            time.time(),
+        ),
+    )
+    _conn().commit()
+
+
+def get_ai_decisions(member_id: str, limit: int = 10) -> list[dict]:
+    rows = _conn().execute(
+        """SELECT llm_response, parsed_order, source, timestamp
+           FROM ch_ai_decisions WHERE member_id=?
+           ORDER BY timestamp DESC LIMIT ?""",
+        (member_id, limit),
+    ).fetchall()
+    return [dict(r) for r in rows]
+
+
 # ── Leaderboard ────────────────────────────────────────────────────────────────
 
 def get_leaderboard(date: str | None = None) -> list[dict]:

clearing_house/templates/dashboard.html

@@ -262,6 +262,34 @@ async function openDetail(memberId) {
       html += '<p style="color:var(--muted)">No trades today</p>';
     }
 
+    // AI Decision History
+    if (d.ai_decisions && d.ai_decisions.length > 0) {
+      html += `<h3 style="margin:20px 0 8px;">AI Reasoning (last ${d.ai_decisions.length})</h3>`;
+      d.ai_decisions.forEach(dec => {
+        const ts = new Date(dec.timestamp * 1000).toLocaleTimeString();
+        const srcBadge = dec.source === 'llm'
+          ? '<span style="background:#e8eaf6; color:#5c6bc0; padding:1px 6px; border-radius:8px; font-size:10px;">LLM</span>'
+          : dec.source === 'fallback'
+            ? '<span style="background:#fff3e0; color:#e65100; padding:1px 6px; border-radius:8px; font-size:10px;">Fallback</span>'
+            : `<span style="background:#ffebee; color:#c62828; padding:1px 6px; border-radius:8px; font-size:10px;">${dec.source}</span>`;
+        const order = dec.parsed_order;
+        const orderLine = order
+          ? `<span class="${order.side === 'BUY' ? 'positive' : 'negative'}" style="font-weight:bold;">${order.side}</span> ${order.quantity} <strong>${order.symbol}</strong> @ €${fmt(order.price)}`
+          : '<span style="color:var(--muted)">No valid order</span>';
+
+        html += `
+          <div style="border:1px solid var(--border, #333); border-radius:6px; padding:10px; margin-bottom:8px; font-size:12px;">
+            <div style="display:flex; justify-content:space-between; align-items:center; margin-bottom:6px;">
+              <div>${srcBadge} &nbsp; ${orderLine}</div>
+              <span style="color:var(--muted); font-size:11px;">${ts}</span>
+            </div>
+            <div style="color:var(--muted); white-space:pre-wrap; font-family:monospace; font-size:11px; max-height:80px; overflow-y:auto; background:rgba(0,0,0,0.15); padding:6px; border-radius:4px;">${dec.llm_response}</div>
+          </div>`;
+      });
+    } else if (!d.is_human) {
+      html += '<h3 style="margin:20px 0 8px;">AI Reasoning</h3><p style="color:var(--muted)">No AI decisions recorded yet</p>';
+    }
+
     detailContent.innerHTML = html;
   } catch(e) {
     detailContent.innerHTML = '<p class="negative">Error loading member data</p>';