fix: strip provider prefix from all AI provider models and instant URL refresh

Backend fixes:
- Strip provider prefix (anthropic:, openai:, deepseek:, ollama:) in all
  provider Chat methods and constructors for robust handling
- Models are now correctly parsed regardless of caller format

Frontend fixes:
- Tool cards now persist in AI chat after approval execution by adding
  to streamEvents array
- Dashboard now listens for pulse:metadata-changed custom event
- AI Chat emits this event when set_resource_url tool completes
- Guest URL icons now update instantly when AI sets them

frontend-modern/src/components/AI/AIChat.tsx

@@ -557,6 +557,14 @@ export const AIChat: Component<AIChatProps> = (props) => {
                   };
                   // Add to both toolCalls and streamEvents for chronological display
                   const events = msg.streamEvents || [];
+
+                  // Emit event for metadata refresh if AI set a resource URL
+                  if (data.name === 'set_resource_url' && data.success) {
+                    window.dispatchEvent(new CustomEvent('pulse:metadata-changed', {
+                      detail: { source: 'ai', tool: data.name }
+                    }));
+                  }
+
                   return {
                     ...msg,
                     pendingTools: updatedPending,
@@ -748,11 +756,14 @@ export const AIChat: Component<AIChatProps> = (props) => {
           };
 
           const remainingApprovals = m.pendingApprovals?.filter((a) => a.toolId !== approval.toolId) || [];
+          const existingEvents = m.streamEvents || [];
 
           return {
             ...m,
             pendingApprovals: remainingApprovals,
             toolCalls: [...(m.toolCalls || []), newToolCall],
+            // Also add to streamEvents so it appears in the chronological view
+            streamEvents: [...existingEvents, { type: 'tool' as const, tool: newToolCall }],
             // Clear the stale "I need approval" content after the last approval is processed
             // The tool output will show the result instead
             content: remainingApprovals.length === 0 ? '' : m.content,

frontend-modern/src/components/Dashboard/Dashboard.tsx

@@ -290,15 +290,31 @@ export function Dashboard(props: DashboardProps) {
   // Total columns for colspan calculations
   const totalColumns = createMemo(() => visibleColumns().length);
 
-  // Load all guest metadata on mount (single API call for all guests)
+  // Helper function to refresh guest metadata from server
-  onMount(async () => {
+  const refreshGuestMetadata = async () => {
     try {
       const metadata = await GuestMetadataAPI.getAllMetadata();
       updateGuestMetadataState(() => metadata || {});
+      logger.debug('Guest metadata refreshed');
     } catch (err) {
-      // Silently fail - metadata is optional for display
+      logger.debug('Failed to refresh guest metadata', err);
-      logger.debug('Failed to load guest metadata', err);
     }
+  };
+
+  // Load all guest metadata on mount (single API call for all guests)
+  onMount(async () => {
+    await refreshGuestMetadata();
+  });
+
+  // Listen for metadata changes from AI or other sources
+  createEffect(() => {
+    const handleMetadataChanged = () => {
+      logger.debug('Metadata changed event received, refreshing...');
+      refreshGuestMetadata();
+    };
+
+    window.addEventListener('pulse:metadata-changed', handleMetadataChanged);
+    return () => window.removeEventListener('pulse:metadata-changed', handleMetadataChanged);
   });
 
   // Callback to update a guest's custom URL in metadata

internal/ai/providers/anthropic.go

@@ -166,6 +166,10 @@ func (c *AnthropicClient) Chat(ctx context.Context, req ChatRequest) (*ChatRespo
 
 	// Use provided model or fall back to client default
 	model := req.Model
+	// Strip provider prefix if present - callers may pass the full "provider:model" string
+	if len(model) > 10 && model[:10] == "anthropic:" {
+		model = model[10:]
+	}
 	if model == "" {
 		model = c.model
 	}

internal/ai/providers/anthropic_oauth.go

@@ -431,6 +431,10 @@ func (c *AnthropicOAuthClient) Chat(ctx context.Context, req ChatRequest) (*Chat
 	}
 
 	model := req.Model
+	// Strip provider prefix if present - callers may pass the full "provider:model" string
+	if len(model) > 10 && model[:10] == "anthropic:" {
+		model = model[10:]
+	}
 	if model == "" {
 		model = c.model
 	}

internal/ai/providers/openai.go

@@ -28,11 +28,18 @@ type OpenAIClient struct {
 	client  *http.Client
 }
 
+
 // NewOpenAIClient creates a new OpenAI API client
 func NewOpenAIClient(apiKey, model, baseURL string) *OpenAIClient {
 	if baseURL == "" {
 		baseURL = openaiAPIURL
 	}
+	// Strip provider prefix if present - the model should be just the model name
+	if strings.HasPrefix(model, "openai:") {
+		model = strings.TrimPrefix(model, "openai:")
+	} else if strings.HasPrefix(model, "deepseek:") {
+		model = strings.TrimPrefix(model, "deepseek:")
+	}
 	return &OpenAIClient{
 		apiKey:  apiKey,
 		model:   model,
@@ -205,10 +212,19 @@ func (c *OpenAIClient) Chat(ctx context.Context, req ChatRequest) (*ChatResponse
 
 	// Use provided model or fall back to client default
 	model := req.Model
+	// Strip provider prefix if present - callers may pass the full "provider:model" string
+	if strings.HasPrefix(model, "openai:") {
+		model = strings.TrimPrefix(model, "openai:")
+	} else if strings.HasPrefix(model, "deepseek:") {
+		model = strings.TrimPrefix(model, "deepseek:")
+	}
 	if model == "" {
 		model = c.model
 	}
 
+	// Debug log to trace model issues
+	log.Debug().Str("model", model).Str("req_model", req.Model).Str("c_model", c.model).Str("base_url", c.baseURL).Msg("OpenAI/DeepSeek Chat request")
+
 	// Build request
 	openaiReq := openaiRequest{
 		Model:    model,
@@ -245,6 +261,9 @@ func (c *OpenAIClient) Chat(ctx context.Context, req ChatRequest) (*ChatResponse
 		}
 	}
 
+	// Log actual model being sent (INFO level for visibility)
+	log.Info().Str("model_in_request", openaiReq.Model).Str("base_url", c.baseURL).Msg("Sending OpenAI/DeepSeek request")
+
 	body, err := json.Marshal(openaiReq)
 	if err != nil {
 		return nil, fmt.Errorf("failed to marshal request: %w", err)