feat(llm): redo chat storage, part 3

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

@@ -155,7 +155,7 @@ export class AIServiceManager implements IAIServiceManager {
             // Get precedence list from options
             let precedenceList: string[] = ['openai']; // Default to openai if not set
             const precedenceOption = await options.getOption('aiProviderPrecedence');
-            
+
             if (precedenceOption) {
                 try {
                     if (precedenceOption.startsWith('[') && precedenceOption.endsWith(']')) {
@@ -171,10 +171,10 @@ export class AIServiceManager implements IAIServiceManager {
                     log.error(`Error parsing precedence list: ${e}`);
                 }
             }
-            
+
             // Check for configuration issues with providers in the precedence list
             const configIssues: string[] = [];
-            
+
             // Check each provider in the precedence list for proper configuration
             for (const provider of precedenceList) {
                 if (provider === 'openai') {
@@ -198,20 +198,20 @@ export class AIServiceManager implements IAIServiceManager {
                 }
                 // Add checks for other providers as needed
             }
-            
+
             // Return warning message if there are configuration issues
             if (configIssues.length > 0) {
                 let message = 'There are issues with your AI provider configuration:';
-                
+
                 for (const issue of configIssues) {
                     message += `\n• ${issue}`;
                 }
-                
+
                 message += '\n\nPlease check your AI settings.';
-                
+
                 // Log warning to console
                 log.error('AI Provider Configuration Warning: ' + message);
-                
+
                 return message;
             }
 
@@ -279,9 +279,19 @@ export class AIServiceManager implements IAIServiceManager {
 
         // If a specific provider is requested and available, use it
         if (options.model && options.model.includes(':')) {
-            const [providerName, modelName] = options.model.split(':');
+            // Check if this is a provider prefix (e.g., "ollama:qwen3:30b")
+            // vs a model name with version (e.g., "qwen3:30b")
+            const parts = options.model.split(':');
+
+            // Only treat as provider:model if the first part is a known provider
+            const knownProviders = ['openai', 'anthropic', 'ollama', 'local'];
+            const potentialProvider = parts[0];
+
+            if (knownProviders.includes(potentialProvider) && availableProviders.includes(potentialProvider as ServiceProviders)) {
+                // This is a provider:model format
+                const providerName = potentialProvider;
+                const modelName = parts.slice(1).join(':'); // Rejoin the rest as model name
 
-            if (availableProviders.includes(providerName as ServiceProviders)) {
                 try {
                     const modifiedOptions = { ...options, model: modelName };
                     log.info(`[AIServiceManager] Using provider ${providerName} from model prefix with modifiedOptions.stream: ${modifiedOptions.stream}`);
@@ -291,6 +301,7 @@ export class AIServiceManager implements IAIServiceManager {
                     // If the specified provider fails, continue with the fallback providers
                 }
             }
+            // If not a provider prefix, treat the entire string as a model name and continue with normal provider selection
         }
 
         // Try each provider in order until one succeeds

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

@@ -20,44 +20,44 @@ export class MessagePreparationStage extends BasePipelineStage<MessagePreparatio
      */
     protected async process(input: MessagePreparationInput): Promise<{ messages: Message[] }> {
         const { messages, context, systemPrompt, options } = input;
-        
+
         // Determine provider from model string if available (format: "provider:model")
         let provider = 'default';
         if (options?.model && options.model.includes(':')) {
             const [providerName] = options.model.split(':');
             provider = providerName;
         }
-        
+
         // Check if tools are enabled
         const toolsEnabled = options?.enableTools === true;
-        
+
         log.info(`Preparing messages for provider: ${provider}, context: ${!!context}, system prompt: ${!!systemPrompt}, tools: ${toolsEnabled}`);
-        
+
         // Get appropriate formatter for this provider
         const formatter = MessageFormatterFactory.getFormatter(provider);
-        
+
         // Determine the system prompt to use
         let finalSystemPrompt = systemPrompt || SYSTEM_PROMPTS.DEFAULT_SYSTEM_PROMPT;
-        
+
         // If tools are enabled, enhance system prompt with tools guidance
         if (toolsEnabled) {
             const toolCount = toolRegistry.getAllTools().length;
             const toolsPrompt = `You have access to ${toolCount} tools to help you respond. When you need information that might be in the user's notes, use the search_notes tool to find relevant content or the read_note tool to read a specific note by ID. Use tools when specific information is required rather than making assumptions.`;
-            
+
             // Add tools guidance to system prompt
             finalSystemPrompt = finalSystemPrompt + '\n\n' + toolsPrompt;
             log.info(`Enhanced system prompt with tools guidance: ${toolCount} tools available`);
         }
-        
+
         // Format messages using provider-specific approach
         const formattedMessages = formatter.formatMessages(
             messages,
             finalSystemPrompt,
             context
         );
-        
+
         log.info(`Formatted ${messages.length} messages into ${formattedMessages.length} messages for provider: ${provider}`);
-        
+
         return { messages: formattedMessages };
     }
 }

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

@@ -234,7 +234,8 @@ export class ModelSelectionStage extends BasePipelineStage<ModelSelectionInput,
 
             // For backward compatibility, ensure model name is set without prefix
             if (options.model && options.model.includes(':')) {
-                options.model = modelName || options.model.split(':')[1];
+                const parsed = this.parseModelIdentifier(options.model);
+                options.model = modelName || parsed.model;
             }
 
             log.info(`Set provider metadata: provider=${selectedProvider}, model=${modelName}`);