Merge pull request #2208 from TriliumNext/fix/llm-chat-save-bug

fix(llm): save to the same note that the chat request was sent from

apps/client/src/widgets/llm_chat/llm_chat_panel.ts

@@ -350,6 +350,115 @@ export default class LlmChatPanel extends BasicWidget {
         }
     }
 
+    /**
+     * Save current chat data to a specific note ID
+     */
+    async saveCurrentDataToSpecificNote(targetNoteId: string | null) {
+        if (!this.onSaveData || !targetNoteId) {
+            console.warn('Cannot save chat data: no saveData callback or no targetNoteId available');
+            return;
+        }
+
+        try {
+            // Extract current tool execution steps if any exist
+            const toolSteps = extractInChatToolSteps(this.noteContextChatMessages);
+
+            // Get tool executions from both UI and any cached executions in metadata
+            let toolExecutions: Array<{
+                id: string;
+                name: string;
+                arguments: any;
+                result: any;
+                error?: string;
+                timestamp: string;
+            }> = [];
+
+            // First include any tool executions already in metadata (from streaming events)
+            if (this.metadata?.toolExecutions && Array.isArray(this.metadata.toolExecutions)) {
+                toolExecutions = [...this.metadata.toolExecutions];
+                console.log(`Including ${toolExecutions.length} tool executions from metadata`);
+            }
+
+            // Also extract any visible tool steps from the UI
+            const extractedExecutions = toolSteps.map(step => {
+                // Parse tool execution information
+                if (step.type === 'tool-execution') {
+                    try {
+                        const content = JSON.parse(step.content);
+                        return {
+                            id: content.toolCallId || `tool-${Date.now()}-${Math.random().toString(36).substring(2, 7)}`,
+                            name: content.tool || 'unknown',
+                            arguments: content.args || {},
+                            result: content.result || {},
+                            error: content.error,
+                            timestamp: new Date().toISOString()
+                        };
+                    } catch (e) {
+                        // If we can't parse it, create a basic record
+                        return {
+                            id: `tool-${Date.now()}-${Math.random().toString(36).substring(2, 7)}`,
+                            name: 'unknown',
+                            arguments: {},
+                            result: step.content,
+                            timestamp: new Date().toISOString()
+                        };
+                    }
+                } else if (step.type === 'result' && step.name) {
+                    // Handle result steps with a name
+                    return {
+                        id: `tool-${Date.now()}-${Math.random().toString(36).substring(2, 7)}`,
+                        name: step.name,
+                        arguments: {},
+                        result: step.content,
+                        timestamp: new Date().toISOString()
+                    };
+                }
+                return {
+                    id: `tool-${Date.now()}-${Math.random().toString(36).substring(2, 7)}`,
+                    name: 'unknown',
+                    arguments: {},
+                    result: 'Unrecognized tool step',
+                    timestamp: new Date().toISOString()
+                };
+            });
+
+            // Merge the tool executions, keeping only unique IDs
+            const existingIds = new Set(toolExecutions.map((t: {id: string}) => t.id));
+            for (const exec of extractedExecutions) {
+                if (!existingIds.has(exec.id)) {
+                    toolExecutions.push(exec);
+                    existingIds.add(exec.id);
+                }
+            }
+
+            const dataToSave = {
+                messages: this.messages,
+                noteId: targetNoteId,
+                chatNoteId: targetNoteId, // For backward compatibility
+                toolSteps: toolSteps,
+                // Add sources if we have them
+                sources: this.sources || [],
+                // Add metadata
+                metadata: {
+                    model: this.metadata?.model || undefined,
+                    provider: this.metadata?.provider || undefined,
+                    temperature: this.metadata?.temperature || 0.7,
+                    lastUpdated: new Date().toISOString(),
+                    // Add tool executions
+                    toolExecutions: toolExecutions
+                }
+            };
+
+            console.log(`Saving chat data to specific note ${targetNoteId}, ${toolSteps.length} tool steps, ${this.sources?.length || 0} sources, ${toolExecutions.length} tool executions`);
+
+            // Save the data to the note attribute via the callback
+            // This is the ONLY place we should save data, letting the container widget handle persistence
+            await this.onSaveData(dataToSave);
+        } catch (error) {
+            console.error('Error saving chat data to specific note:', error);
+        }
+    }
+
     /**
      * Load saved chat data from the note attribute
      */
@@ -867,8 +976,8 @@ export default class LlmChatPanel extends BasicWidget {
                     this.showSources(postResponse.sources);
                 }
 
-                // Process the assistant response
+                // Process the assistant response with original chat note ID
-                this.processAssistantResponse(postResponse.content, postResponse);
+                this.processAssistantResponse(postResponse.content, postResponse, this.noteId);
 
                 hideLoadingIndicator(this.loadingIndicator);
                 return true;
@@ -884,7 +993,7 @@ export default class LlmChatPanel extends BasicWidget {
     /**
      * Process an assistant response - add to UI and save
      */
-    private async processAssistantResponse(content: string, fullResponse?: any) {
+    private async processAssistantResponse(content: string, fullResponse?: any, originalChatNoteId?: string | null) {
         // Add the response to the chat UI
         this.addMessageToChat('assistant', content);
 
@@ -910,8 +1019,8 @@ export default class LlmChatPanel extends BasicWidget {
             ];
         }
 
-        // Save to note
+        // Save to note - use original chat note ID if provided
-        this.saveCurrentData().catch(err => {
+        this.saveCurrentDataToSpecificNote(originalChatNoteId || this.noteId).catch(err => {
             console.error("Failed to save assistant response to note:", err);
         });
     }
@@ -936,12 +1045,15 @@ export default class LlmChatPanel extends BasicWidget {
             timestamp: string;
         }> = [];
 
+        // Store the original chat note ID to ensure we save to the correct note even if user switches
+        const originalChatNoteId = this.noteId;
+
         return setupStreamingResponse(
             this.noteId,
             messageParams,
             // Content update handler
             (content: string, isDone: boolean = false) => {
-                this.updateStreamingUI(content, isDone);
+                this.updateStreamingUI(content, isDone, originalChatNoteId);
 
                 // Update session data with additional metadata when streaming is complete
                 if (isDone) {
@@ -1067,13 +1179,13 @@ export default class LlmChatPanel extends BasicWidget {
     /**
      * Update the UI with streaming content
      */
-    private updateStreamingUI(assistantResponse: string, isDone: boolean = false) {
+    private updateStreamingUI(assistantResponse: string, isDone: boolean = false, originalChatNoteId?: string | null) {
         // Track if we have a streaming message in progress
         const hasStreamingMessage = !!this.noteContextChatMessages.querySelector('.assistant-message.streaming');
-        
+
         // Create a new message element or use the existing streaming one
         let assistantMessageEl: HTMLElement;
-        
+
         if (hasStreamingMessage) {
             // Use the existing streaming message
             assistantMessageEl = this.noteContextChatMessages.querySelector('.assistant-message.streaming')!;
@@ -1103,7 +1215,7 @@ export default class LlmChatPanel extends BasicWidget {
         if (isDone) {
             // Remove the streaming class to mark this message as complete
             assistantMessageEl.classList.remove('streaming');
-            
+
             // Apply syntax highlighting
             formatCodeBlocks($(assistantMessageEl as HTMLElement));
 
@@ -1118,8 +1230,8 @@ export default class LlmChatPanel extends BasicWidget {
                 timestamp: new Date()
             });
 
-            // Save the updated message list
+            // Save the updated message list to the original chat note
-            this.saveCurrentData();
+            this.saveCurrentDataToSpecificNote(originalChatNoteId || this.noteId);
         }
 
         // Scroll to bottom

apps/client/src/widgets/type_widgets/ai_chat.ts

@@ -182,17 +182,30 @@ export default class AiChatTypeWidget extends TypeWidget {
 
     // Save chat data to the note
     async saveData(data: any) {
-        if (!this.note) {
+        // If we have a noteId in the data, that's the AI Chat note we should save to
+        // This happens when the chat panel is saving its conversation
+        const targetNoteId = data.noteId;
+        
+        // If no noteId in data, use the current note (for new chats)
+        const noteIdToUse = targetNoteId || this.note?.noteId;
+        
+        if (!noteIdToUse) {
+            console.warn("Cannot save AI Chat data: no note ID available");
             return;
         }
 
         try {
-            console.log(`AiChatTypeWidget: Saving data for note ${this.note.noteId}`);
+            console.log(`AiChatTypeWidget: Saving data for note ${noteIdToUse} (current note: ${this.note?.noteId}, data.noteId: ${data.noteId})`);
+
+            // Safety check: if we have both IDs and they don't match, warn about it
+            if (targetNoteId && this.note?.noteId && targetNoteId !== this.note.noteId) {
+                console.warn(`Note ID mismatch: saving to ${targetNoteId} but current note is ${this.note.noteId}`);
+            }
 
             // Format the data properly - this is the canonical format of the data
             const formattedData = {
                 messages: data.messages || [],
-                noteId: this.note.noteId, // Always use the note's own ID
+                noteId: noteIdToUse, // Always preserve the correct note ID
                 toolSteps: data.toolSteps || [],
                 sources: data.sources || [],
                 metadata: {
@@ -201,8 +214,8 @@ export default class AiChatTypeWidget extends TypeWidget {
                 }
             };
 
-            // Save the data to the note
+            // Save the data to the correct note
-            await server.put(`notes/${this.note.noteId}/data`, {
+            await server.put(`notes/${noteIdToUse}/data`, {
                 content: JSON.stringify(formattedData, null, 2)
             });
         } catch (e) {

apps/client/src/widgets/type_widgets/options/ai_settings/ai_settings_widget.ts

@@ -48,7 +48,7 @@ export default class AiSettingsWidget extends OptionsWidget {
             if (optionName === 'aiEnabled') {
                 try {
                     const isEnabled = value === 'true';
-                    
+
                     if (isEnabled) {
                         toastService.showMessage(t("ai_llm.ai_enabled") || "AI features enabled");
                     } else {

apps/server/src/services/llm/ai_service_manager.ts

@@ -40,8 +40,8 @@ interface NoteContext {
 }
 
 export class AIServiceManager implements IAIServiceManager {
-    private services: Partial<Record<ServiceProviders, AIService>> = {};
+    private currentService: AIService | null = null;
-
+    private currentProvider: ServiceProviders | null = null;
     private initialized = false;
 
     constructor() {
@@ -50,9 +50,8 @@ export class AIServiceManager implements IAIServiceManager {
             log.error(`Error initializing LLM tools during AIServiceManager construction: ${error.message || String(error)}`);
         });
 
-        // Set up event listener for provider changes
+        // Removed complex provider change listener - we'll read options fresh each time
-        this.setupProviderChangeListener();
+
-        
         this.initialized = true;
     }
 
@@ -140,15 +139,15 @@ export class AIServiceManager implements IAIServiceManager {
      */
     async getOrCreateAnyService(): Promise<AIService> {
         this.ensureInitialized();
-        
+
         // Get the selected provider using the new configuration system
         const selectedProvider = await this.getSelectedProviderAsync();
-        
+
-        
+
         if (!selectedProvider) {
             throw new Error('No AI provider is selected. Please select a provider (OpenAI, Anthropic, or Ollama) in your AI settings.');
         }
-        
+
         try {
             const service = await this.getOrCreateChatProvider(selectedProvider);
             if (service) {
@@ -166,7 +165,7 @@ export class AIServiceManager implements IAIServiceManager {
      */
     isAnyServiceAvailable(): boolean {
         this.ensureInitialized();
-        
+
         // Check if we have the selected provider available
         return this.getAvailableProviders().length > 0;
     }
@@ -174,43 +173,37 @@ export class AIServiceManager implements IAIServiceManager {
     /**
      * Get list of available providers
      */
-    getAvailableProviders(): ServiceProviders[] {
+        getAvailableProviders(): ServiceProviders[] {
         this.ensureInitialized();
-        
+
         const allProviders: ServiceProviders[] = ['openai', 'anthropic', 'ollama'];
         const availableProviders: ServiceProviders[] = [];
-        
+
         for (const providerName of allProviders) {
-            // Use a sync approach - check if we can create the provider
+            // Check configuration to see if provider would be available
-            const service = this.services[providerName];
+            try {
-            if (service && service.isAvailable()) {
+                switch (providerName) {
-                availableProviders.push(providerName);
+                    case 'openai':
-            } else {
+                        if (options.getOption('openaiApiKey') || options.getOption('openaiBaseUrl')) {
-                // For providers not yet created, check configuration to see if they would be available
+                            availableProviders.push(providerName);
-                try {
+                        }
-                    switch (providerName) {
+                        break;
-                        case 'openai':
+                    case 'anthropic':
-                            if (options.getOption('openaiApiKey')) {
+                        if (options.getOption('anthropicApiKey')) {
-                                availableProviders.push(providerName);
+                            availableProviders.push(providerName);
-                            }
+                        }
-                            break;
+                        break;
-                        case 'anthropic':
+                    case 'ollama':
-                            if (options.getOption('anthropicApiKey')) {
+                        if (options.getOption('ollamaBaseUrl')) {
-                                availableProviders.push(providerName);
+                            availableProviders.push(providerName);
-                            }
+                        }
-                            break;
+                        break;
-                        case 'ollama':
-                            if (options.getOption('ollamaBaseUrl')) {
-                                availableProviders.push(providerName);
-                            }
-                            break;
-                    }
-                } catch (error) {
-                    // Ignore configuration errors, provider just won't be available
                 }
+            } catch (error) {
+                // Ignore configuration errors, provider just won't be available
             }
         }
-        
+
         return availableProviders;
     }
 
@@ -234,11 +227,11 @@ export class AIServiceManager implements IAIServiceManager {
 
         // Get the selected provider
         const selectedProvider = await this.getSelectedProviderAsync();
-        
+
         if (!selectedProvider) {
             throw new Error('No AI provider is selected. Please select a provider in your AI settings.');
         }
-        
+
         // Check if the selected provider is available
         const availableProviders = this.getAvailableProviders();
         if (!availableProviders.includes(selectedProvider)) {
@@ -379,47 +372,68 @@ export class AIServiceManager implements IAIServiceManager {
     }
 
     /**
-     * Get or create a chat provider on-demand with inline validation
+     * Clear the current provider (forces recreation on next access)
+     */
+    public clearCurrentProvider(): void {
+        this.currentService = null;
+        this.currentProvider = null;
+        log.info('Cleared current provider - will be recreated on next access');
+    }
+
+    /**
+     * Get or create the current provider instance - only one instance total
      */
     private async getOrCreateChatProvider(providerName: ServiceProviders): Promise<AIService | null> {
-        // Return existing provider if already created
+        // If provider type changed, clear the old one
-        if (this.services[providerName]) {
+        if (this.currentProvider && this.currentProvider !== providerName) {
-            return this.services[providerName];
+            log.info(`Provider changed from ${this.currentProvider} to ${providerName}, clearing old service`);
+            this.currentService = null;
+            this.currentProvider = null;
         }
 
-        // Create and validate provider on-demand
+        // Return existing service if it matches and is available
+        if (this.currentService && this.currentProvider === providerName && this.currentService.isAvailable()) {
+            return this.currentService;
+        }
+
+        // Clear invalid service
+        if (this.currentService) {
+            this.currentService = null;
+            this.currentProvider = null;
+        }
+
+        // Create new service for the requested provider
         try {
             let service: AIService | null = null;
-            
+
             switch (providerName) {
                 case 'openai': {
                     const apiKey = options.getOption('openaiApiKey');
                     const baseUrl = options.getOption('openaiBaseUrl');
                     if (!apiKey && !baseUrl) return null;
-                    
+
                     service = new OpenAIService();
-                    // Validate by checking if it's available
                     if (!service.isAvailable()) {
                         throw new Error('OpenAI service not available');
                     }
                     break;
                 }
-                
+
                 case 'anthropic': {
                     const apiKey = options.getOption('anthropicApiKey');
                     if (!apiKey) return null;
-                    
+
                     service = new AnthropicService();
                     if (!service.isAvailable()) {
                         throw new Error('Anthropic service not available');
                     }
                     break;
                 }
-                
+
                 case 'ollama': {
                     const baseUrl = options.getOption('ollamaBaseUrl');
                     if (!baseUrl) return null;
-                    
+
                     service = new OllamaService();
                     if (!service.isAvailable()) {
                         throw new Error('Ollama service not available');
@@ -427,9 +441,12 @@ export class AIServiceManager implements IAIServiceManager {
                     break;
                 }
             }
-            
+
             if (service) {
-                this.services[providerName] = service;
+                // Cache the new service
+                this.currentService = service;
+                this.currentProvider = providerName;
+                log.info(`Created and cached new ${providerName} service`);
                 return service;
             }
         } catch (error: any) {
@@ -630,28 +647,47 @@ export class AIServiceManager implements IAIServiceManager {
      * Check if a specific provider is available
      */
     isProviderAvailable(provider: string): boolean {
-        return this.services[provider as ServiceProviders]?.isAvailable() ?? false;
+        // Check if this is the current provider and if it's available
+        if (this.currentProvider === provider && this.currentService) {
+            return this.currentService.isAvailable();
+        }
+
+        // For other providers, check configuration
+        try {
+            switch (provider) {
+                case 'openai':
+                    return !!(options.getOption('openaiApiKey') || options.getOption('openaiBaseUrl'));
+                case 'anthropic':
+                    return !!options.getOption('anthropicApiKey');
+                case 'ollama':
+                    return !!options.getOption('ollamaBaseUrl');
+                default:
+                    return false;
+            }
+        } catch {
+            return false;
+        }
     }
 
     /**
      * Get metadata about a provider
      */
     getProviderMetadata(provider: string): ProviderMetadata | null {
-        const service = this.services[provider as ServiceProviders];
+        // Only return metadata if this is the current active provider
-        if (!service) {
+        if (this.currentProvider === provider && this.currentService) {
-            return null;
+            return {
+                name: provider,
+                capabilities: {
+                    chat: true,
+                    streaming: true,
+                    functionCalling: provider === 'openai' // Only OpenAI has function calling
+                },
+                models: ['default'], // Placeholder, could be populated from the service
+                defaultModel: 'default'
+            };
         }
 
-        return {
+        return null;
-            name: provider,
-            capabilities: {
-                chat: true,
-                streaming: true,
-                functionCalling: provider === 'openai' // Only OpenAI has function calling
-            },
-            models: ['default'], // Placeholder, could be populated from the service
-            defaultModel: 'default'
-        };
     }
 
 
@@ -665,67 +701,8 @@ export class AIServiceManager implements IAIServiceManager {
         return String(error);
     }
 
-    /**
+    // Removed complex event listener and cache invalidation logic
-     * Set up event listener for provider changes
+    // Services will be created fresh when needed by reading current options
-     */
-    private setupProviderChangeListener(): void {
-        // List of AI-related options that should trigger service recreation
-        const aiRelatedOptions = [
-            'aiEnabled',
-            'aiSelectedProvider',
-            'openaiApiKey',
-            'openaiBaseUrl', 
-            'openaiDefaultModel',
-            'anthropicApiKey',
-            'anthropicBaseUrl',
-            'anthropicDefaultModel',
-            'ollamaBaseUrl',
-            'ollamaDefaultModel'
-        ];
-
-        eventService.subscribe(['entityChanged'], async ({ entityName, entity }) => {
-            if (entityName === 'options' && entity && aiRelatedOptions.includes(entity.name)) {
-                log.info(`AI-related option '${entity.name}' changed, recreating LLM services`);
-                
-                // Special handling for aiEnabled toggle
-                if (entity.name === 'aiEnabled') {
-                    const isEnabled = entity.value === 'true';
-                    
-                    if (isEnabled) {
-                        log.info('AI features enabled, initializing AI service');
-                        // Initialize the AI service
-                        await this.initialize();
-                    } else {
-                        log.info('AI features disabled, clearing providers');
-                        // Clear chat providers
-                        this.services = {};
-                    }
-                } else {
-                    // For other AI-related options, recreate services on-demand
-                    await this.recreateServices();
-                }
-            }
-        });
-    }
-
-    /**
-     * Recreate LLM services when provider settings change
-     */
-    private async recreateServices(): Promise<void> {
-        try {
-            log.info('Recreating LLM services due to configuration change');
-
-            // Clear configuration cache first
-            clearConfigurationCache();
-
-            // Clear existing chat providers (they will be recreated on-demand)
-            this.services = {};
-
-            log.info('LLM services recreated successfully');
-        } catch (error) {
-            log.error(`Error recreating LLM services: ${this.handleError(error)}`);
-        }
-    }
 
 }
 

apps/server/src/services/llm/config/configuration_helpers.ts

@@ -1,4 +1,3 @@
-import configurationManager from './configuration_manager.js';
 import optionService from '../../options.js';
 import log from '../../log.js';
 import type {
@@ -13,7 +12,7 @@ import type {
  */
 
 /**
- * Get the selected AI provider
+ * Get the selected AI provider - always fresh from options
  */
 export async function getSelectedProvider(): Promise<ProviderType | null> {
     const providerOption = optionService.getOption('aiSelectedProvider');
@@ -25,38 +24,100 @@ export async function getSelectedProvider(): Promise<ProviderType | null> {
  * Parse a model identifier (handles "provider:model" format)
  */
 export function parseModelIdentifier(modelString: string): ModelIdentifier {
-    return configurationManager.parseModelIdentifier(modelString);
+    if (!modelString) {
+        return {
+            modelId: '',
+            fullIdentifier: ''
+        };
+    }
+
+    const parts = modelString.split(':');
+
+    if (parts.length === 1) {
+        // No provider prefix, just model name
+        return {
+            modelId: modelString,
+            fullIdentifier: modelString
+        };
+    }
+
+    // Check if first part is a known provider
+    const potentialProvider = parts[0].toLowerCase();
+    const knownProviders: ProviderType[] = ['openai', 'anthropic', 'ollama'];
+
+    if (knownProviders.includes(potentialProvider as ProviderType)) {
+        // Provider prefix format
+        const provider = potentialProvider as ProviderType;
+        const modelId = parts.slice(1).join(':'); // Rejoin in case model has colons
+
+        return {
+            provider,
+            modelId,
+            fullIdentifier: modelString
+        };
+    }
+
+    // Not a provider prefix, treat whole string as model name
+    return {
+        modelId: modelString,
+        fullIdentifier: modelString
+    };
 }
 
 /**
  * Create a model configuration from a model string
  */
 export function createModelConfig(modelString: string, defaultProvider?: ProviderType): ModelConfig {
-    return configurationManager.createModelConfig(modelString, defaultProvider);
+    const identifier = parseModelIdentifier(modelString);
+    const provider = identifier.provider || defaultProvider || 'openai'; // fallback to openai if no provider specified
+
+    return {
+        provider,
+        modelId: identifier.modelId,
+        displayName: identifier.fullIdentifier
+    };
 }
 
 /**
- * Get the default model for a specific provider
+ * Get the default model for a specific provider - always fresh from options
  */
 export async function getDefaultModelForProvider(provider: ProviderType): Promise<string | undefined> {
-    const config = await configurationManager.getAIConfig();
+    const optionKey = `${provider}DefaultModel` as const;
-    return config.defaultModels[provider]; // This can now be undefined
+    return optionService.getOption(optionKey) || undefined;
 }
 
 /**
- * Get provider settings for a specific provider
+ * Get provider settings for a specific provider - always fresh from options
  */
 export async function getProviderSettings(provider: ProviderType) {
-    const config = await configurationManager.getAIConfig();
+    switch (provider) {
-    return config.providerSettings[provider];
+        case 'openai':
+            return {
+                apiKey: optionService.getOption('openaiApiKey'),
+                baseUrl: optionService.getOption('openaiBaseUrl'),
+                defaultModel: optionService.getOption('openaiDefaultModel')
+            };
+        case 'anthropic':
+            return {
+                apiKey: optionService.getOption('anthropicApiKey'),
+                baseUrl: optionService.getOption('anthropicBaseUrl'),
+                defaultModel: optionService.getOption('anthropicDefaultModel')
+            };
+        case 'ollama':
+            return {
+                baseUrl: optionService.getOption('ollamaBaseUrl'),
+                defaultModel: optionService.getOption('ollamaDefaultModel')
+            };
+        default:
+            return {};
+    }
 }
 
 /**
- * Check if AI is enabled
+ * Check if AI is enabled - always fresh from options
  */
 export async function isAIEnabled(): Promise<boolean> {
-    const config = await configurationManager.getAIConfig();
+    return optionService.getOptionBool('aiEnabled');
-    return config.enabled;
 }
 
 /**
@@ -82,7 +143,7 @@ export async function isProviderConfigured(provider: ProviderType): Promise<bool
  */
 export async function getAvailableSelectedProvider(): Promise<ProviderType | null> {
     const selectedProvider = await getSelectedProvider();
-    
+
     if (!selectedProvider) {
         return null; // No provider selected
     }
@@ -95,17 +156,51 @@ export async function getAvailableSelectedProvider(): Promise<ProviderType | nul
 }
 
 /**
- * Validate the current AI configuration
+ * Validate the current AI configuration - simplified validation
  */
 export async function validateConfiguration() {
-    return configurationManager.validateConfig();
+    const result = {
+        isValid: true,
+        errors: [] as string[],
+        warnings: [] as string[]
+    };
+
+    const aiEnabled = await isAIEnabled();
+    if (!aiEnabled) {
+        result.warnings.push('AI features are disabled');
+        return result;
+    }
+
+    const selectedProvider = await getSelectedProvider();
+    if (!selectedProvider) {
+        result.errors.push('No AI provider selected');
+        result.isValid = false;
+        return result;
+    }
+
+    // Validate provider-specific settings
+    const settings = await getProviderSettings(selectedProvider);
+
+    if (selectedProvider === 'openai' && !(settings as any)?.apiKey) {
+        result.warnings.push('OpenAI API key is not configured');
+    }
+
+    if (selectedProvider === 'anthropic' && !(settings as any)?.apiKey) {
+        result.warnings.push('Anthropic API key is not configured');
+    }
+
+    if (selectedProvider === 'ollama' && !(settings as any)?.baseUrl) {
+        result.warnings.push('Ollama base URL is not configured');
+    }
+
+    return result;
 }
 
 /**
- * Clear cached configuration (use when settings change)
+ * Clear cached configuration (no-op since we removed caching)
  */
 export function clearConfigurationCache(): void {
-    configurationManager.clearCache();
+    // No caching anymore, so nothing to clear
 }
 
 /**
@@ -136,7 +231,7 @@ export async function getValidModelConfig(provider: ProviderType): Promise<{ mod
  */
 export async function getSelectedModelConfig(): Promise<{ model: string; provider: ProviderType } | null> {
     const selectedProvider = await getSelectedProvider();
-    
+
     if (!selectedProvider) {
         return null; // No provider selected
     }

apps/server/src/services/llm/config/configuration_manager.ts

@@ -21,11 +21,6 @@ import type {
  */
 export class ConfigurationManager {
     private static instance: ConfigurationManager | null = null;
-    private cachedConfig: AIConfig | null = null;
-    private lastConfigUpdate: number = 0;
-
-    // Cache for 5 minutes to avoid excessive option reads
-    private static readonly CACHE_DURATION = 5 * 60 * 1000;
 
     private constructor() {}
 
@@ -37,14 +32,9 @@ export class ConfigurationManager {
     }
 
     /**
-     * Get the complete AI configuration
+     * Get the complete AI configuration - always fresh, no caching
      */
     public async getAIConfig(): Promise<AIConfig> {
-        const now = Date.now();
-        if (this.cachedConfig && (now - this.lastConfigUpdate) < ConfigurationManager.CACHE_DURATION) {
-            return this.cachedConfig;
-        }
-
         try {
             const config: AIConfig = {
                 enabled: await this.getAIEnabled(),
@@ -53,8 +43,6 @@ export class ConfigurationManager {
                 providerSettings: await this.getProviderSettings()
             };
 
-            this.cachedConfig = config;
-            this.lastConfigUpdate = now;
             return config;
         } catch (error) {
             log.error(`Error loading AI configuration: ${error}`);
@@ -263,14 +251,6 @@ export class ConfigurationManager {
         return result;
     }
 
-    /**
-     * Clear cached configuration (force reload on next access)
-     */
-    public clearCache(): void {
-        this.cachedConfig = null;
-        this.lastConfigUpdate = 0;
-    }
-
     // Private helper methods
 
     private async getAIEnabled(): Promise<boolean> {

apps/server/src/services/llm/pipeline/stages/model_selection_stage.ts

@@ -111,19 +111,13 @@ export class ModelSelectionStage extends BasePipelineStage<ModelSelectionInput,
             const { getValidModelConfig } = await import('../../config/configuration_helpers.js');
             const modelConfig = await getValidModelConfig(selectedProvider);
 
-            if (modelConfig) {
+            if (!modelConfig) {
-                // We have a valid configured model
+                throw new Error(`No default model configured for provider ${selectedProvider}. Please set a default model in your AI settings.`);
-                updatedOptions.model = modelConfig.model;
-            } else {
-                // No model configured, try to fetch and set a default from the service
-                const fetchedModel = await this.fetchAndSetDefaultModel(selectedProvider);
-                if (!fetchedModel) {
-                    throw new Error(`No default model configured for provider ${selectedProvider}. Please set a default model in your AI settings or ensure the provider service is available.`);
-                }
-                // Use the fetched model
-                updatedOptions.model = fetchedModel;
             }
 
+            // Use the configured model
+            updatedOptions.model = modelConfig.model;
+
             log.info(`Selected provider: ${selectedProvider}, model: ${updatedOptions.model}`);
 
             // Determine query complexity
@@ -183,20 +177,8 @@ export class ModelSelectionStage extends BasePipelineStage<ModelSelectionInput,
             return;
         }
 
-        // If no provider could be determined, try to use precedence
+        // Use the explicitly provided provider - no automatic fallbacks
         let selectedProvider = provider;
-        if (!selectedProvider) {
-            // List of providers in precedence order
-            const providerPrecedence = ['anthropic', 'openai', 'ollama'];
-
-            // Find the first available provider
-            for (const p of providerPrecedence) {
-                if (aiServiceManager.isProviderAvailable(p)) {
-                    selectedProvider = p as ServiceProviders;
-                    break;
-                }
-            }
-        }
 
         // Set the provider metadata in the options
         if (selectedProvider) {
@@ -218,47 +200,7 @@ export class ModelSelectionStage extends BasePipelineStage<ModelSelectionInput,
         }
     }
 
-    /**
-     * Determine model based on selected provider using the new configuration system
-     * This method is now simplified and delegates to the main model selection logic
-     */
-    private async determineDefaultModel(input: ModelSelectionInput): Promise<string> {
-        try {
-            // Use the same logic as the main process method
-            const { getValidModelConfig, getSelectedProvider } = await import('../../config/configuration_helpers.js');
-            const selectedProvider = await getSelectedProvider();
 
-            if (!selectedProvider) {
-                throw new Error('No AI provider is selected. Please select a provider in your AI settings.');
-            }
-
-            // Check if the provider is available through the service manager
-            if (!aiServiceManager.isProviderAvailable(selectedProvider)) {
-                throw new Error(`Selected provider ${selectedProvider} is not available`);
-            }
-
-            // Try to get a valid model config
-            const modelConfig = await getValidModelConfig(selectedProvider);
-            
-            if (!modelConfig) {
-                throw new Error(`No default model configured for provider ${selectedProvider}. Please configure a default model in your AI settings.`);
-            }
-
-            // Set provider metadata
-            if (!input.options.providerMetadata) {
-                input.options.providerMetadata = {
-                    provider: selectedProvider as 'openai' | 'anthropic' | 'ollama' | 'local',
-                    modelId: modelConfig.model
-                };
-            }
-
-            log.info(`Selected default model ${modelConfig.model} from provider ${selectedProvider}`);
-            return modelConfig.model;
-        } catch (error) {
-            log.error(`Error determining default model: ${error}`);
-            throw error; // Don't provide fallback defaults, let the error propagate
-        }
-    }
 
     /**
      * Get estimated context window for Ollama models
@@ -283,48 +225,5 @@ export class ModelSelectionStage extends BasePipelineStage<ModelSelectionInput,
         }
     }
 
-    /**
-     * Use AI service manager to get a configured model for the provider
-     * This eliminates duplication and uses the existing service layer
-     */
-    private async fetchAndSetDefaultModel(provider: ProviderType): Promise<string | null> {
-        try {
-            log.info(`Getting default model for provider ${provider} using AI service manager`);
-            
-            // Use the existing AI service manager instead of duplicating API calls
-            const service = await aiServiceManager.getInstance().getService(provider);
-            
-            if (!service || !service.isAvailable()) {
-                log.info(`Provider ${provider} service is not available`);
-                return null;
-            }
 
-            // Check if the service has a method to get available models
-            if (typeof (service as any).getAvailableModels === 'function') {
-                try {
-                    const models = await (service as any).getAvailableModels();
-                    if (models && models.length > 0) {
-                        // Use the first available model - no hardcoded preferences
-                        const selectedModel = models[0];
-                        
-                        // Import server-side options to update the default model
-                        const optionService = (await import('../../../options.js')).default;
-                        const optionKey = `${provider}DefaultModel` as const;
-                        
-                        await optionService.setOption(optionKey, selectedModel);
-                        log.info(`Set default ${provider} model to: ${selectedModel}`);
-                        return selectedModel;
-                    }
-                } catch (modelError) {
-                    log.error(`Error fetching models from ${provider} service: ${modelError}`);
-                }
-            }
-            
-            log.info(`Provider ${provider} does not support dynamic model fetching`);
-            return null;
-        } catch (error) {
-            log.error(`Error getting default model for provider ${provider}: ${error}`);
-            return null;
-        }
-    }
 }

apps/server/src/services/llm/providers/providers.ts

@@ -26,7 +26,11 @@ export function getOpenAIOptions(
         }
 
         const baseUrl = options.getOption('openaiBaseUrl') || PROVIDER_CONSTANTS.OPENAI.BASE_URL;
-        const modelName = opts.model || options.getOption('openaiDefaultModel') || PROVIDER_CONSTANTS.OPENAI.DEFAULT_MODEL;
+        const modelName = opts.model || options.getOption('openaiDefaultModel');
+
+        if (!modelName) {
+            throw new Error('No OpenAI model configured. Please set a default model in your AI settings.');
+        }
 
         // Create provider metadata
         const providerMetadata: ModelMetadata = {
@@ -87,7 +91,11 @@ export function getAnthropicOptions(
         }
 
         const baseUrl = options.getOption('anthropicBaseUrl') || PROVIDER_CONSTANTS.ANTHROPIC.BASE_URL;
-        const modelName = opts.model || options.getOption('anthropicDefaultModel') || PROVIDER_CONSTANTS.ANTHROPIC.DEFAULT_MODEL;
+        const modelName = opts.model || options.getOption('anthropicDefaultModel');
+
+        if (!modelName) {
+            throw new Error('No Anthropic model configured. Please set a default model in your AI settings.');
+        }
 
         // Create provider metadata
         const providerMetadata: ModelMetadata = {
@@ -150,8 +158,12 @@ export async function getOllamaOptions(
             throw new Error('Ollama API URL is not configured');
         }
 
-        // Get the model name - no prefix handling needed now
+        // Get the model name - no defaults, must be configured by user
-        let modelName = opts.model || options.getOption('ollamaDefaultModel') || 'llama3';
+        let modelName = opts.model || options.getOption('ollamaDefaultModel');
+
+        if (!modelName) {
+            throw new Error('No Ollama model configured. Please set a default model in your AI settings.');
+        }
 
         // Create provider metadata
         const providerMetadata: ModelMetadata = {
@@ -249,4 +261,4 @@ async function getOllamaModelContextWindow(modelName: string): Promise<number> {
         log.info(`Error getting context window for model ${modelName}: ${error}`);
         return MODEL_CAPABILITIES['default'].contextWindowTokens; // Default fallback
     }
-}
+}

apps/server/src/services/options.ts

@@ -82,6 +82,26 @@ function setOption<T extends OptionNames>(name: T, value: string | OptionDefinit
     } else {
         createOption(name, value, false);
     }
+
+    // Clear current AI provider when AI-related options change
+    const aiOptions = [
+        'aiSelectedProvider', 'openaiApiKey', 'openaiBaseUrl', 'openaiDefaultModel',
+        'anthropicApiKey', 'anthropicBaseUrl', 'anthropicDefaultModel',
+        'ollamaBaseUrl', 'ollamaDefaultModel'
+    ];
+
+    if (aiOptions.includes(name)) {
+        // Import dynamically to avoid circular dependencies
+        setImmediate(async () => {
+            try {
+                const aiServiceManager = (await import('./llm/ai_service_manager.js')).default;
+                aiServiceManager.getInstance().clearCurrentProvider();
+                console.log(`Cleared AI provider after ${name} option changed`);
+            } catch (error) {
+                console.log(`Could not clear AI provider: ${error}`);
+            }
+        });
+    }
 }
 
 /**