vigthoria-cli 1.5.7 → 1.5.9

package/src/commands/chat.ts

@@ -142,7 +142,16 @@ export class ChatCommand {
     this.streamMode = options.stream !== false; // Default to true
     this.localMode = false; // Never use local mode - always use Vigthoria API
     
+    // CRITICAL: Agent mode REQUIRES stronger models to prevent hallucinations
+    // Use Vigthoria Cloud for complex agent tasks
     if (this.agentMode) {
+      // Upgrade model for agent mode if using a weak model
+      const weakModels = ['fast', 'mini', 'vigthoria-fast-1.7b'];
+      if (weakModels.includes(options.model)) {
+        this.logger.warn(`Agent mode works better with stronger models. Upgrading from '${options.model}' to 'code'`);
+        options.model = 'code';
+      }
+      
       this.tools = new AgenticTools(
         this.logger,
         projectPath,  // Use resolved path
@@ -196,18 +205,35 @@ export class ChatCommand {
   ): string {
     let prompt = `You are Vigthoria, the premier AI coding assistant of Vigthoria Technologies.
 
-IDENTITY:
-- Created by Vigthoria Technologies
+## IDENTITY & BRAND
+- Created by Vigthoria Technologies (NOT OpenAI, NOT Anthropic, NOT Microsoft)
 - Mascot: Viggy the Blue Queen
 - Philosophy: "Innovation through Intelligence"
-
-Project Context:
+- Always respond as "Vigthoria" - never claim to be another AI
+
+## VIGTHORIA ECOSYSTEM (Important - Know This!)
+Vigthoria is a comprehensive tech platform with these key services:
+- **Vigthoria Coder** (coder.vigthoria.io): AI-powered coding IDE and CLI (this tool!)
+- **Vigthoria Community** (community.vigthoria.io): Code sharing platform, NOT GitHub. When user mentions "repo", assume Vigthoria Community unless they say "GitHub"
+- **Vigthoria GoA** (agent.vigthoria.io): Graph of Agents - autonomous AI orchestration
+- **Vigthoria Meet** (meet.vigthoria.io): Video conferencing platform
+- **Vigthoria Pay**: Payment processing service
+- **Vigthoria B2C Hub**: Business-to-consumer services
+- **Model Router** (port 4009): Routes AI requests to optimal models
+
+## CRITICAL RULES - NEVER VIOLATE:
+1. **NEVER HALLUCINATE** - If you don't know something, say so. Don't make up file contents, directories, or project structures.
+2. **NEVER ASSUME FILE CONTENTS** - If asked about a file, you MUST use read_file tool first. Do NOT guess what's in a file.
+3. **NEVER CONFUSE PLATFORMS** - Vigthoria Community is NOT GitHub. Windows paths are NOT Unix paths.
+4. **ALWAYS USE TOOLS** - In agent mode, USE the tools to read files, not your imagination.
+
+## Project Context
 - Type: ${projectContext.type}
 - Root: ${options.project}
 - Key files: ${projectContext.files.slice(0, 10).join(', ')}
 ${projectContext.type === 'node' ? `- Dependencies: ${Object.keys(projectContext.dependencies).slice(0, 15).join(', ')}` : ''}
 
-CODE QUALITY STANDARDS (CRITICAL):
+## CODE QUALITY STANDARDS (CRITICAL):
 1. ALWAYS produce complete, production-ready code - no placeholders
 2. When creating UI/HTML/CSS:
    - Ensure proper color contrast (text readable against backgrounds)
@@ -219,12 +245,76 @@ CODE QUALITY STANDARDS (CRITICAL):
 6. Use semantic HTML, accessible patterns
 7. Show COMPLETE implementations
 
-Guidelines:
+## Guidelines:
 - Provide working code first, explanations second
 - Be concise but thorough
 - Excellence is the standard - mediocrity is not acceptable
-
-Special Commands (user may use these):
+- If you need to see file contents, USE read_file FIRST before responding
+
+## VIGTHORIA CLI COMMANDS (Know these to help users):
+
+### Core Commands:
+- \`vigthoria chat\` (or \`vig c\`) - Start interactive chat (this session!)
+- \`vigthoria agent\` (or \`vig a\`) - Autonomous agent mode (can read/write files, run commands)
+- \`vigthoria edit <file>\` (or \`vig e\`) - Edit a file with AI assistance
+- \`vigthoria generate "desc"\` (or \`vig g\`) - Generate code from description
+- \`vigthoria explain <file>\` (or \`vig x\`) - Explain code in detail
+- \`vigthoria fix <file>\` (or \`vig f\`) - Fix bugs and issues
+- \`vigthoria review <file>\` (or \`vig r\`) - Code quality review
+
+### Repository Commands (Vigthoria Community - NOT GitHub):
+- \`vigthoria repo push\` - Push project to Vigthoria Community repo
+- \`vigthoria repo pull <name>\` - Pull project from Vigthoria repo
+- \`vigthoria repo list\` - List your projects in Vigthoria repo
+- \`vigthoria repo status\` - Show sync status
+- \`vigthoria repo share <name>\` - Generate shareable link
+- \`vigthoria repo clone <url>\` - Clone a public project
+- \`vigthoria repo delete <name>\` - Remove from repo
+
+### Hub/Marketplace Commands (Vigthoria API Modules):
+- \`vigthoria hub discover\` - Interactive module discovery
+- \`vigthoria hub list\` - List all API modules
+- \`vigthoria hub search <query>\` - Semantic search for modules
+- \`vigthoria hub activate <module>\` - Enable pay-as-you-go for module
+- \`vigthoria hub active\` - Show your active modules
+- \`vigthoria hub info <module>\` - Get module details
+
+### Deployment Commands (Vigthoria Hosting):
+- \`vigthoria deploy preview\` - Deploy to free preview URL
+- \`vigthoria deploy subdomain <name>\` - Deploy to yourname.vigthoria.io
+- \`vigthoria deploy custom <domain>\` - Deploy to custom domain
+- \`vigthoria deploy list\` - List your deployments
+- \`vigthoria deploy plans\` - Show hosting plans/pricing
+- \`vigthoria deploy status\` - Check deployment status
+- \`vigthoria deploy verify <domain>\` - Verify DNS config
+- \`vigthoria deploy remove <domain>\` - Remove deployment
+
+### Account & Config Commands:
+- \`vigthoria login\` - Authenticate with Vigthoria
+- \`vigthoria logout\` - Logout from account
+- \`vigthoria status\` - Show auth & subscription status
+- \`vigthoria config\` - Configure CLI settings
+- \`vigthoria update\` - Check for updates & upgrade CLI
+- \`vigthoria init\` - Initialize Vigthoria in project
+
+## NATURAL LANGUAGE UNDERSTANDING (Interpret User Intent):
+When users say these phrases, understand their intent:
+- "push it", "upload this", "save to repo" → Run \`vigthoria repo push\`
+- "pull my project", "download from repo" → Run \`vigthoria repo pull\`
+- "deploy this", "host it", "put it online" → Run \`vigthoria deploy\`
+- "show my repos", "list projects" → Run \`vigthoria repo list\`
+- "share this project" → Run \`vigthoria repo share\`
+- "check for updates", "upgrade cli" → Run \`vigthoria update\`
+- "what APIs are available", "show modules" → Run \`vigthoria hub list\`
+- "enable music api", "activate <module>" → Run \`vigthoria hub activate\`
+- "my deployments", "where is it hosted" → Run \`vigthoria deploy list\`
+- "make it live", "publish" → Run \`vigthoria deploy\`
+- "fix bugs", "debug this" → Run \`vigthoria fix <file>\`
+- "explain this code" → Run \`vigthoria explain <file>\`
+- "review my code" → Run \`vigthoria review <file>\`
+- "create a component", "generate code for" → Run \`vigthoria generate\`
+
+## Special Chat Commands (user may use these in chat):
 - /file <path> - Read and include a file in context
 - /edit <path> - Switch to file editing mode
 - /diff - Show pending changes
@@ -309,14 +399,44 @@ Special Commands (user may use these):
       input: process.stdin,
       output: process.stdout,
       prompt: chalk.cyan('you › '),
+      terminal: true,
     });
 
     let currentModel = options.model;
     let pendingChanges: { file: string; content: string } | null = null;
+    let isRunning = true;
+
+    // Handle unexpected close events
+    this.rl.on('close', () => {
+      if (isRunning) {
+        console.log(chalk.yellow('\n\n⚠ Session interrupted. Saving...'));
+        if (this.currentSession && this.messages.length > 1) {
+          this.sessionManager.save(this.currentSession);
+          console.log(chalk.gray(`Session saved: ${this.currentSession.id}`));
+        }
+        console.log(chalk.cyan('Run `vigthoria chat --resume` to continue.\n'));
+        isRunning = false;
+      }
+    });
+
+    // Handle SIGINT (Ctrl+C) gracefully
+    process.on('SIGINT', () => {
+      if (isRunning) {
+        isRunning = false;
+        console.log(chalk.yellow('\n\nExiting...'));
+        if (this.currentSession && this.messages.length > 1) {
+          this.sessionManager.save(this.currentSession);
+          console.log(chalk.gray(`Session saved: ${this.currentSession.id}`));
+        }
+        console.log(chalk.cyan('Goodbye! 👋\n'));
+        process.exit(0);
+      }
+    });
 
     this.rl.prompt();
 
     for await (const line of this.rl) {
+      if (!isRunning) break;
       const input = line.trim();
       
       if (!input) {
@@ -340,8 +460,24 @@ Special Commands (user may use these):
             
           case 'model':
             if (args[0]) {
-              currentModel = args[0];
-              this.logger.success(`Model switched to: ${currentModel}`);
+              const newModel = args[0];
+              currentModel = newModel;
+              
+              // Show branded model name
+              const modelInfo = this.config.getAvailableModels().find(m => m.id === newModel);
+              const isCloud = this.config.isCloudModel(newModel);
+              
+              if (isCloud) {
+                console.log();
+                console.log(chalk.magenta('  ☁️  Switched to: ') + chalk.magenta.bold(modelInfo?.name || 'Vigthoria Cloud'));
+                console.log(chalk.gray('      671B cloud model - ideal for complex tasks'));
+                console.log();
+              } else {
+                console.log();
+                console.log(chalk.green('  🏠 Switched to: ') + chalk.green.bold(modelInfo?.name || newModel));
+                console.log(chalk.gray('      Local model - fast, no API costs'));
+                console.log();
+              }
             } else {
               this.printModels();
             }
@@ -521,6 +657,9 @@ Special Commands (user may use these):
       }
 
       // Regular chat message
+      // Check if we should suggest Cloud upgrade for complex tasks
+      this.suggestCloudUpgrade(currentModel, input);
+      
       await this.chat(input, currentModel);
       
       // Save message to session
@@ -579,7 +718,22 @@ Special Commands (user may use these):
 
     } catch (error) {
       spinner.stop();
-      this.logger.error('Failed to get response:', (error as Error).message);
+      const errMsg = (error as Error).message || 'Unknown error';
+      
+      // Check for specific error types
+      if (errMsg.includes('ECONNREFUSED') || errMsg.includes('ENOTFOUND')) {
+        this.logger.error('Connection failed: Unable to reach AI service');
+        console.log(chalk.gray('  Check your internet connection or try again later.'));
+      } else if (errMsg.includes('timeout') || errMsg.includes('ETIMEDOUT')) {
+        this.logger.error('Request timed out: AI service took too long to respond');
+        console.log(chalk.gray('  Try a shorter query or check service status.'));
+      } else if (errMsg.includes('401') || errMsg.includes('Unauthorized')) {
+        this.logger.error('Authentication failed: Your session may have expired');
+        console.log(chalk.gray('  Run `vigthoria login` to re-authenticate.'));
+      } else {
+        this.logger.error('Failed to get response:', errMsg);
+      }
+      
       // Remove failed user message
       this.messages.pop();
     }
@@ -782,6 +936,18 @@ Special Commands (user may use these):
     console.log(chalk.cyan('/exit') + '         - Exit Vigthoria (auto-saves)');
     console.log(chalk.cyan('/quit') + '         - Same as /exit');
     console.log();
+    console.log(chalk.yellow('── CLI Commands (run in terminal) ──'));
+    console.log(chalk.gray('  vigthoria repo push     - Push to Vigthoria Community'));
+    console.log(chalk.gray('  vigthoria repo pull     - Pull from your repos'));
+    console.log(chalk.gray('  vigthoria deploy        - Deploy & host project'));
+    console.log(chalk.gray('  vigthoria hub           - Browse API modules'));
+    console.log(chalk.gray('  vigthoria update        - Update CLI to latest'));
+    console.log();
+    console.log(chalk.yellow('💡 Natural Language Tips:'));
+    console.log(chalk.gray('  Say "push it" → I\'ll help with vigthoria repo push'));
+    console.log(chalk.gray('  Say "deploy this" → I\'ll guide you through hosting'));
+    console.log(chalk.gray('  Say "what APIs are available" → I\'ll show hub modules'));
+    console.log();
     
     if (this.agentMode) {
       console.log(chalk.yellow('Agent Mode Tools:'));
@@ -876,10 +1042,60 @@ Special Commands (user may use these):
     const models = this.config.getAvailableModels();
     
     console.log();
-    this.logger.section('Available Models');
-    models.forEach(m => {
-      console.log(chalk.cyan(m.id.padEnd(20)) + chalk.gray(m.description));
+    this.logger.section('═══ VIGTHORIA MODELS ═══');
+    console.log();
+    
+    // Group by tier
+    const localModels = models.filter(m => m.tier === 'local');
+    const cloudModels = models.filter(m => m.tier === 'cloud');
+    
+    console.log(chalk.green.bold('  🏠 VIGTHORIA LOCAL (Self-hosted, fast, no API cost)'));
+    console.log(chalk.gray('  ─────────────────────────────────────────────────'));
+    localModels.forEach(m => {
+      const isDefault = m.id === 'code';
+      const marker = isDefault ? chalk.yellow(' ★ DEFAULT') : '';
+      console.log(chalk.cyan('    ' + m.id.padEnd(15)) + chalk.white(m.name.padEnd(25)) + chalk.gray(m.description) + marker);
     });
+    
+    if (cloudModels.length > 0) {
+      console.log();
+      console.log(chalk.magenta.bold('  ☁️  VIGTHORIA CLOUD (Premium, for complex tasks)'));
+      console.log(chalk.gray('  ─────────────────────────────────────────────────'));
+      cloudModels.forEach(m => {
+        console.log(chalk.cyan('    ' + m.id.padEnd(15)) + chalk.white(m.name.padEnd(25)) + chalk.gray(m.description));
+      });
+      console.log();
+      console.log(chalk.yellow('  💡 Tip: Use /model cloud for complex multi-file tasks'));
+    } else {
+      console.log();
+      console.log(chalk.yellow('  💡 Upgrade to Pro for Vigthoria Cloud (671B models)'));
+    }
     console.log();
   }
+
+  // Suggest Cloud upgrade for complex tasks
+  private suggestCloudUpgrade(currentModel: string, prompt: string): boolean {
+    // Don't suggest if already on cloud
+    if (this.config.isCloudModel(currentModel)) {
+      return false;
+    }
+    
+    // Check if task seems complex
+    if (this.config.isComplexTask(prompt)) {
+      console.log();
+      console.log(chalk.yellow('╔══════════════════════════════════════════════════════════╗'));
+      console.log(chalk.yellow('║') + chalk.white.bold('  💡 This looks like a complex task!                      ') + chalk.yellow('║'));
+      console.log(chalk.yellow('║') + chalk.gray('  Current: ') + chalk.cyan(currentModel.padEnd(43)) + chalk.yellow('║'));
+      console.log(chalk.yellow('║') + chalk.gray('  Suggested: ') + chalk.magenta('Vigthoria Cloud (671B)'.padEnd(41)) + chalk.yellow('║'));
+      console.log(chalk.yellow('║                                                          ║'));
+      console.log(chalk.yellow('║') + chalk.white('  Type ') + chalk.cyan('/model cloud') + chalk.white(' for better results on:           ') + chalk.yellow('║'));
+      console.log(chalk.yellow('║') + chalk.gray('    • Multi-file refactoring                              ') + chalk.yellow('║'));
+      console.log(chalk.yellow('║') + chalk.gray('    • Architecture decisions                              ') + chalk.yellow('║'));
+      console.log(chalk.yellow('║') + chalk.gray('    • Complex feature implementation                      ') + chalk.yellow('║'));
+      console.log(chalk.yellow('╚══════════════════════════════════════════════════════════╝'));
+      console.log();
+      return true;
+    }
+    return false;
+  }
 }

package/src/index.ts

@@ -157,11 +157,12 @@ async function main() {
     });
 
   // Agent command - Agentic mode (Vigthoria Autonomous)
+  // Uses Vigthoria v3 Code 30B or Vigthoria Cloud for complex tasks
   program
     .command('agent')
     .alias('a')
     .description('Start agentic mode - AI can read/write files, run commands')
-    .option('-m, --model <model>', 'Select AI model', 'code')
+    .option('-m, --model <model>', 'Select AI model (code, cloud, ultra)', 'code')
     .option('-p, --project <path>', 'Set project context path', process.cwd())
     .option('--auto-approve', 'Auto-approve all actions (dangerous!)', false)
     .action(async (options) => {

package/src/utils/api.ts

@@ -73,7 +73,7 @@ export class APIClient {
       httpsAgent,
       headers: {
         'Content-Type': 'application/json',
-        'User-Agent': `Vigthoria-CLI/${process.env.npm_package_version || '1.4.5'}`,
+        'User-Agent': `Vigthoria-CLI/${process.env.npm_package_version || '1.5.7'}`,
       },
     });
 
@@ -84,7 +84,7 @@ export class APIClient {
       httpsAgent,
       headers: {
         'Content-Type': 'application/json',
-        'User-Agent': `Vigthoria-CLI/${process.env.npm_package_version || '1.4.5'}`,
+        'User-Agent': `Vigthoria-CLI/${process.env.npm_package_version || '1.5.7'}`,
       },
     });
 
@@ -231,18 +231,21 @@ export class APIClient {
   }
 
   /**
-   * Chat API - Supports multiple inference strategies
+   * Chat API - SIMPLIFIED Architecture (NO Ollama!)
    * 
-   * For authenticated users: Uses Vigthoria Cloud API
-   * For local users: Uses local Ollama or Model Router
+   * 1. Vigthoria Cloud API (coder.vigthoria.io) - Primary for authenticated users
+   * 2. Vigthoria Inference Service (port 8010) - Native 30B model
+   * 3. Model Router (port 4009) - Routes to OpenRouter (Kimi K2.5, DeepSeek)
+   * 
+   * NO OLLAMA FALLBACK - That causes hallucinations with small models!
    */
   async chat(messages: ChatMessage[], model: string, useLocal: boolean = false): Promise<ChatResponse> {
     const resolvedModel = this.resolveModelId(model);
     
-    // If authenticated and not forcing local, use Vigthoria Cloud API
-    if (!useLocal && this.config.isAuthenticated()) {
+    // STRATEGY 1: Vigthoria Cloud API (authenticated users)
+    if (this.config.isAuthenticated()) {
       try {
-        this.logger.debug(`Attempting Vigthoria Cloud API call to /api/ai/chat with model: ${resolvedModel}`);
+        this.logger.debug(`Vigthoria Cloud API: ${resolvedModel}`);
         const response = await this.client.post('/api/ai/chat', {
           messages,
           model: resolvedModel,
@@ -258,118 +261,62 @@ export class APIClient {
             usage: response.data.usage,
           };
         }
-        this.logger.debug(`Cloud API returned success=false: ${JSON.stringify(response.data)}`);
+        this.logger.debug(`Cloud API returned success=false`);
       } catch (error: any) {
         const errMsg = error.response?.data?.error || error.message || 'Unknown error';
-        const status = error.response?.status || 'no status';
-        this.logger.debug(`Vigthoria Cloud API failed (${status}): ${errMsg}`);
+        this.logger.debug(`Vigthoria Cloud API failed: ${errMsg}`);
+        // Continue to fallback strategies
       }
-    } else {
-      this.logger.debug(`Skipping cloud API - useLocal: ${useLocal}, isAuthenticated: ${this.config.isAuthenticated()}`);
     }
 
-    // Strategy 1: Try Vigthoria Inference Server directly (NATIVE models on port 8010)
+    // STRATEGY 2: Vigthoria Inference Service (port 8010) - Native 30B model
     try {
+      this.logger.debug(`Trying Vigthoria Inference Service (8010): ${resolvedModel}`);
       const response = await axios.post('http://localhost:8010/v1/chat/completions', {
         model: resolvedModel,
         messages,
         max_tokens: this.config.get('preferences').maxTokens,
         temperature: 0.7,
         stream: false,
-      }, { timeout: 120000 });
+      }, { timeout: 180000 }); // 3 min timeout for 30B model
 
       if (response.data.choices && response.data.choices.length > 0) {
         return {
-          id: response.data.id || `vigthoria-native-${Date.now()}`,
+          id: response.data.id || `vigthoria-inference-${Date.now()}`,
           message: response.data.choices[0].message?.content || response.data.choices[0].text,
           model: response.data.model || model,
           usage: response.data.usage,
         };
       }
     } catch (error) {
-      this.logger.debug('Vigthoria Inference Server (8010) failed, trying Model Router...');
+      this.logger.debug('Vigthoria Inference Service (8010) unavailable');
     }
 
-    // Strategy 2: Try local Model Router's Vigthoria chat endpoint
+    // STRATEGY 3: Model Router (port 4009) - Routes to OpenRouter (Kimi/DeepSeek)
     try {
-      const response = await axios.post('http://localhost:4009/api/vigthoria/chat', {
-        messages,
+      this.logger.debug(`Trying Model Router (4009): ${resolvedModel}`);
+      const response = await axios.post('http://localhost:4009/v1/chat/completions', {
         model: resolvedModel,
-        maxTokens: this.config.get('preferences').maxTokens,
+        messages,
+        max_tokens: this.config.get('preferences').maxTokens,
         temperature: 0.7,
         stream: false,
-      }, { timeout: 120000 });
+      }, { timeout: 180000 });
 
-      if (response.data.success !== false) {
+      if (response.data.choices && response.data.choices.length > 0) {
         return {
           id: response.data.id || `router-${Date.now()}`,
-          message: response.data.response || response.data.message || response.data.content,
+          message: response.data.choices[0].message?.content || response.data.choices[0].text,
           model: response.data.model || model,
           usage: response.data.usage,
         };
       }
     } catch (error) {
-      this.logger.debug('Model Router (4009) failed, trying Ollama directly...');
+      this.logger.debug('Model Router (4009) unavailable');
     }
 
-    // Strategy 3: Try Ollama directly (for local development/testing)
-    try {
-      const ollamaModel = this.resolveToOllamaModel(model);
-      const prompt = this.formatMessagesForOllama(messages);
-      
-      const response = await axios.post('http://localhost:11434/api/generate', {
-        model: ollamaModel,
-        prompt,
-        stream: false,
-      }, { timeout: 120000 });
-
-      return {
-        id: `ollama-${Date.now()}`,
-        message: response.data.response,
-        model: ollamaModel,
-        usage: {
-          prompt_tokens: response.data.prompt_eval_count || 0,
-          completion_tokens: response.data.eval_count || 0,
-          total_tokens: (response.data.prompt_eval_count || 0) + (response.data.eval_count || 0),
-        },
-      };
-    } catch (error) {
-      this.logger.debug('Ollama failed...');
-    }
-
-    throw new Error('AI service unavailable. Please check your authentication with `vigthoria login` or try again later.');
-  }
-
-  // Map CLI model names to Ollama model names (for local/offline fallback)
-  // Vigthoria_v3_Code_30B runs on qwen3-coder base via Vigthoria Cloud
-  // Local users with Ollama can use the base model for offline work
-  private resolveToOllamaModel(model: string): string {
-    const ollamaMap: Record<string, string> = {
-      'fast': 'qwen3:0.6b',
-      'mini': 'smollm2:135m',
-      'code': 'qwen3-coder:latest',           // Vigthoria_v3_Code_30B (cloud) / qwen3-coder (local fallback)
-      'balanced': 'phi3:mini',
-      'creative': 'gemma3:latest',
-      'vigthoria-fast-1.7b': 'qwen3:0.6b',
-      'vigthoria-mini-0.6b': 'smollm2:135m',
-      'vigthoria-v3-code-30b': 'qwen3-coder:latest',  // Vigthoria_v3_Code_30B
-      'vigthoria-v2-code-8b': 'qwen3-coder:latest',   // Legacy v2
-      'vigthoria-balanced-4b': 'phi3:mini',
-      'vigthoria-creative-9b-v4': 'gemma3:latest',
-    };
-    return ollamaMap[model] || model;
-  }
-
-  // Format messages for Ollama's generate API
-  private formatMessagesForOllama(messages: ChatMessage[]): string {
-    return messages
-      .filter(m => m.role !== 'system' || messages.indexOf(m) === 0)
-      .map(m => {
-        if (m.role === 'system') return `System: ${m.content}\n\n`;
-        if (m.role === 'user') return `User: ${m.content}\n`;
-        return `Assistant: ${m.content}\n`;
-      })
-      .join('') + 'Assistant:';
+    // NO OLLAMA FALLBACK! That causes hallucinations!
+    throw new Error('AI service unavailable. Please ensure Vigthoria services are running or login with `vigthoria login`.');
   }
 
   // Streaming chat
@@ -543,108 +490,51 @@ export class APIClient {
     return response.data;
   }
 
-  // Model resolution - maps short names to actual Vigthoria model IDs
+  // Model resolution - maps Vigthoria model names to internal IDs
+  // INTERNAL USE ONLY - users see only Vigthoria branding
   private resolveModelId(shortName: string): string {
     const modelMap: Record<string, string> = {
       // ═══════════════════════════════════════════════════════════════
-      // Vigthoria Native Models - Core Suite
+      // VIGTHORIA LOCAL - Self-hosted models
       // ═══════════════════════════════════════════════════════════════
       'fast': 'vigthoria-fast-1.7b',
-      'mini': 'vigthoria-mini-0.6b',
+      'mini': 'vigthoria-fast-1.7b',
       'balanced': 'vigthoria-balanced-4b',
-      
-      // Code Models - Current & Future Versions
-      'code': 'vigthoria-v2-code-8b',
-      'code-v2': 'vigthoria-v2-code-8b',
-      'code-v2-8b': 'vigthoria-v2-code-8b',
-      'code-v3': 'vigthoria-v3-code-8b',      // Future
-      'code-v3-8b': 'vigthoria-v3-code-8b',   // Future
-      'code-v3-32b': 'vigthoria-v3-code-32b', // Future large model
-      'code-v4': 'vigthoria-v4-code-8b',      // Future
-      'code-32b': 'vigthoria-v3-code-32b',    // Future large model alias
-      
-      // Creative Models
       'creative': 'vigthoria-creative-9b-v4',
-      'creative-v3': 'vigthoria-creative-9b-v3',
-      'creative-v4': 'vigthoria-creative-9b-v4',
-      
-      // Music Model
-      'music': 'vigthoria-music-master-4b',
-      'music-master': 'vigthoria-music-master-4b',
       
-      // Aliases
-      'pro': 'vigthoria-balanced-4b',
-      'ultra': 'vigthoria-creative-9b-v4',
-      
-      // Legacy Vigthoria models
-      'master': 'vigthoria_master:latest',
-      'c1': 'vigthoria_c1_m:latest',
-      'm1': 'vigthoria_m1_m:latest',
+      // Code Models - 30B is the default powerhouse
+      'code': 'qwen3-coder:latest',           // Internal: qwen3-coder 30B
+      'code-30b': 'qwen3-coder:latest',
+      'code-8b': 'vigthoria-v2-code-8b',
+      'pro': 'qwen3-coder:latest',
       
       // ═══════════════════════════════════════════════════════════════
-      // External Models (via Ollama fallback)
+      // VIGTHORIA CLOUD - Premium cloud models (internal routing)
       // ═══════════════════════════════════════════════════════════════
-      'qwen-coder': 'qwen2.5-coder:7b',
-      'qwen-coder-7b': 'qwen2.5-coder:7b',
-      'qwen-coder-32b': 'qwen2.5-coder:32b',
-      'deepseek': 'deepseek-coder-v2:latest',
-      'deepseek-r1': 'deepseek-r1:8b',
-      'deepseek-r1-32b': 'deepseek-r1:32b',
-      'llama3': 'llama3:8b-instruct-q4_0',
-      'mixtral': 'mixtral:8x7b-instruct-v0.1-q4_K_M',
+      'cloud': 'deepseek-v3.1:671b-cloud',
+      'cloud-reason': 'moonshotai/kimi-k2.5',
+      'agent': 'deepseek-v3.1:671b-cloud',
+      'ultra': 'deepseek-v3.1:671b-cloud',
     };
 
-    // If already a full model ID (contains 'vigthoria' or known prefix), return as-is
-    if (shortName.includes('vigthoria') || shortName.includes(':') || shortName.includes('-')) {
-      // Check if it's a known short name first
+    // If already a full model ID, return as-is
+    if (shortName.includes('vigthoria') || shortName.includes('/') || shortName.includes(':')) {
       if (modelMap[shortName]) {
         return modelMap[shortName];
       }
-      // Otherwise assume it's a direct model ID
       return shortName;
     }
 
-    return modelMap[shortName] || shortName;
+    return modelMap[shortName] || 'qwen3-coder:latest';  // Default to 30B
   }
 
-  // Health check with short timeout and multiple fallbacks
+  // Health check
   async healthCheck(): Promise<boolean> {
-    const apiUrl = this.config.get('apiUrl') || 'https://coder.vigthoria.io';
-    
     try {
-      // Try the main API health endpoint first with shorter timeout
-      const response = await this.client.get('/api/health', {
-        timeout: 10000,  // 10 second timeout for health check
-      });
+      const response = await this.client.get('/api/health', { timeout: 10000 });
       return response.data?.status === 'ok' || response.data?.healthy === true;
-    } catch (error) {
-      // Fallback 1: try root health endpoint with axios
-      try {
-        const fallback = await this.client.get('/health', {
-          timeout: 5000,
-        });
-        return fallback.data?.status === 'ok' || fallback.data?.healthy === true || fallback.status === 200;
-      } catch {
-        // Fallback 2: try native fetch (Node 18+ built-in, works better on Windows)
-        try {
-          const controller = new AbortController();
-          const timeoutId = setTimeout(() => controller.abort(), 5000);
-          
-          const fetchResponse = await fetch(`${apiUrl}/health`, {
-            method: 'GET',
-            signal: controller.signal,
-          });
-          clearTimeout(timeoutId);
-          
-          if (fetchResponse.ok) {
-            const data = await fetchResponse.json() as { status?: string; healthy?: boolean };
-            return data?.status === 'ok' || data?.healthy === true;
-          }
-          return fetchResponse.ok;
-        } catch {
-          return false;
-        }
-      }
+    } catch {
+      return false;
     }
   }
 }