fix: resolve Chronos API error and display advanced risk metrics

Fix critical issues preventing ML forecasts and hiding P1 metrics:

1. Chronos API call mismatch
- Changed predict() to use positional argument instead of keyword
- Fixed: ChronosBoltPipeline.predict() error with 'context' parameter
- Changed torch_dtype to dtype parameter to resolve deprecation warning

2. Missing advanced metrics display
- Extract Information Ratio, Calmar Ratio, Ulcer Index from risk analysis
- Display metrics conditionally in UI when available
- Metrics were calculated but never shown to users

Backend changes:
- backend/mcp_servers/ensemble_predictor_mcp.py (lines 178, 192)

Frontend changes:
- app.py format_analysis_result() (lines 300-321)

Test coverage: All 19 P1 tests passing

app.py

@@ -297,6 +297,11 @@ def format_analysis_result(final_state: AgentState, holdings: List[Dict[str, Any
     sharpe = float(risk_metrics.get("sharpe_ratio", 0))
     volatility = float(risk_metrics.get("volatility_annual", 0)) * 100
 
+    # Extract advanced metrics
+    information_ratio = risk_metrics.get("information_ratio")
+    calmar_ratio = risk_metrics.get("calmar_ratio")
+    ulcer_index = risk_metrics.get("ulcer_index")
+
     metrics_md = f"""
 ### Key Metrics
 
@@ -307,6 +312,14 @@ def format_analysis_result(final_state: AgentState, holdings: List[Dict[str, Any
 - **CVaR (95%)**: ${cvar_95:,.2f}
 """
 
+    # Add advanced metrics if available
+    if information_ratio is not None:
+        metrics_md += f"- **Information Ratio**: {float(information_ratio):.2f}\n"
+    if calmar_ratio is not None:
+        metrics_md += f"- **Calmar Ratio**: {float(calmar_ratio):.2f}\n"
+    if ulcer_index is not None:
+        metrics_md += f"- **Ulcer Index**: {float(ulcer_index):.2f}\n"
+
     # Format ML forecasts if available
     forecasts_md = ""
     ensemble_forecasts = final_state.get("ensemble_forecasts", {})

backend/mcp_servers/ensemble_predictor_mcp.py

@@ -175,7 +175,7 @@ def _forecast_chronos(
             _chronos_pipeline = BaseChronosPipeline.from_pretrained(
                 model_name,
                 device_map=device,
-                torch_dtype=dtype,
+                dtype=dtype,
             )
             logger.info(f"Chronos model loaded successfully on {device}")
         except Exception as e:
@@ -189,7 +189,7 @@ def _forecast_chronos(
     # Generate forecast
     try:
         forecast = _chronos_pipeline.predict(
-            context=context_tensor,
+            context_tensor,
             prediction_length=horizon,
         )