feat(manifold): audit matching quality with ManifoldMatchResult and manifold_match_audit table

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

api/main.py

@@ -209,6 +209,28 @@ async def get_attribution():
     return {"attribution": attribution, "total_attributed_trades": total}
 
 
+@app.get("/api/metrics/manifold-matches")
+async def get_manifold_matches():
+    """Manifold match audit — summary stats and recent match attempts.
+
+    summary:
+      total_accepted           — matches accepted (score >= 0.40, inversion unambiguous)
+      total_rejected           — matches rejected (low score or ambiguous inversion)
+      total_no_results         — no Manifold market found or API error
+      avg_match_score          — mean Jaccard score for accepted matches
+      trades_dominated_by_mfld — open trades where feat_mfld_lo is the largest signal
+
+    recent_matches: last 50 rows from manifold_match_audit, newest first.
+      used_in_trade=True only when status='accepted' AND a trade was actually executed.
+    """
+    data = await db.get_manifold_matches(limit=50)
+    for match in data["recent_matches"]:
+        ts = match.get("timestamp")
+        if ts is not None and hasattr(ts, "isoformat"):
+            match["timestamp"] = ts.isoformat()
+    return data
+
+
 @app.get("/api/summary")
 async def get_summary():
     """Dashboard summary card data.