Implement visual provenance (🦞) for fallback responses and update config.example.json with NVIDIA/FreeRide examples.

config/config.example.json

@@ -80,6 +80,15 @@
       "api_key": "sk-key1",
       "api_base": "https://api1.example.com/v1"
     },
+    {
+      "model_name": "nemotron-120b",
+      "model": "nvidia/nemotron-3-super-120b-a12b",
+      "protocol": "nvidia",
+      "api_base": "https://integrate.api.nvidia.com/v1",
+      "api_keys": [
+        "env://NVIDIA_API_KEY"
+      ]
+    },
     {
       "model_name": "loadbalanced-gpt-5.4",
       "model": "openai/gpt-5.4",
@@ -429,6 +438,9 @@
     "list_dir": {
       "enabled": true
     },
+    "freeride": {
+      "enabled": true
+    },
     "message": {
       "enabled": true
     },

pkg/agent/loop_turn.go

@@ -1475,6 +1475,10 @@ turnLoop:
 	}
 
 	ts.setPhase(TurnPhaseFinalizing)
+	// pico: freeride provenance marker
+	if used, model := ts.GetFallbackInfo(); used && model != "" {
+		finalContent += "\n\n🦞 provenance: fallback via " + model
+	}
 	ts.setFinalContent(finalContent)
 	if !ts.opts.NoHistory {
 		finalMsg := providers.Message{Role: "assistant", Content: finalContent}