polydev-ai 1.6.0 → 1.7.0

package/lib/cliManager.js

@@ -42,9 +42,6 @@ class CLIManager {
         id: 'claude_code',
         name: 'Claude Code',
         command: process.env.CLAUDE_CODE_PATH || 'claude',
-        // Model configuration: use env var or default to 'haiku' for speed
-        // Options: haiku (fastest), sonnet (balanced), opus (most capable)
-        model: process.env.POLYDEV_CLAUDE_MODEL || 'haiku',
         subcommands: {
           chat: [],
           version: ['--version'],
@@ -58,9 +55,6 @@ class CLIManager {
         id: 'codex_cli',
         name: 'Codex CLI',
         command: process.env.CODEX_CLI_PATH || 'codex',
-        // Model configuration: use env var or default to 'gpt-4o-mini' for speed
-        // Options: gpt-4o-mini (fastest), gpt-4o, o3-mini, o1, etc.
-        model: process.env.POLYDEV_CODEX_MODEL || 'gpt-4o-mini',
         subcommands: {
           chat: ['chat'],
           version: ['--version'],
@@ -78,8 +72,6 @@ class CLIManager {
         id: 'gemini_cli',
         name: 'Gemini CLI',
         command: process.env.GEMINI_CLI_PATH || 'gemini',
-        // Model configuration for Gemini
-        model: process.env.POLYDEV_GEMINI_MODEL || 'gemini-2.0-flash',
         subcommands: {
           chat: ['chat'],
           version: ['--version'],
@@ -386,7 +378,7 @@ This is a known issue with @google/gemini-cli@0.3.4 and older Node.js versions.`
     }
   }
 
-  async sendCliPrompt(providerId, prompt, mode = 'args', timeoutMs = null) {
+  async sendCliPrompt(providerId, prompt, mode = 'args', timeoutMs = null, model = null) {
     // Set provider-specific default timeouts (180s for all by default, complex prompts take time)
     if (timeoutMs === null) {
       timeoutMs = 180000; // 180 seconds default for all providers
@@ -432,6 +424,13 @@ This is a known issue with @google/gemini-cli@0.3.4 and older Node.js versions.`
         };
       }
 
+      // Log model being used
+      if (model) {
+        console.log(`[Polydev CLI] Using model for ${providerId}: ${model}`);
+      } else {
+        console.log(`[Polydev CLI] No model specified for ${providerId}, using CLI default`);
+      }
+
       const promptVariants = [
         provider.subcommands?.test_prompt ? [...provider.subcommands.test_prompt] : []
       ];
@@ -445,16 +444,15 @@ This is a known issue with @google/gemini-cli@0.3.4 and older Node.js versions.`
       if (providerId === 'codex_cli') {
         const execArgs = promptVariants.find(args => args.includes('exec')) || promptVariants[0];
         try {
-          // Pass model configuration to Codex
-          const content = await this.executeCodexExec(provider.command, execArgs, prompt, timeoutMs, provider.model);
+          const content = await this.executeCodexExec(provider.command, execArgs, prompt, timeoutMs, model);
           return {
             success: true,
             content,
             tokens_used: this.estimateTokens(prompt + content),
             latency_ms: Date.now() - startTime,
             provider: providerId,
-            model: provider.model,
             mode: 'args',
+            model_used: model || 'cli_default',
             timestamp: new Date()
           };
         } catch (error) {
@@ -463,7 +461,6 @@ This is a known issue with @google/gemini-cli@0.3.4 and older Node.js versions.`
             error: `CLI execution failed: ${error instanceof Error ? error.message : String(error)}`,
             latency_ms: Date.now() - startTime,
             provider: providerId,
-            model: provider.model,
             mode,
             timestamp: new Date()
           };
@@ -473,9 +470,25 @@ This is a known issue with @google/gemini-cli@0.3.4 and older Node.js versions.`
       let lastErrorMessage = null;
 
       for (const promptArgs of promptVariants) {
-        // Add model parameter for Claude Code
-        const modelArgs = provider.model ? ['--model', provider.model] : [];
-        const args = Array.isArray(promptArgs) ? [...promptArgs, ...modelArgs, prompt] : [...modelArgs, prompt];
+        // Build args with model flag if specified
+        let args = Array.isArray(promptArgs) ? [...promptArgs] : [];
+        
+        // Add model flag based on CLI type
+        if (model) {
+          if (providerId === 'claude_code') {
+            // Claude Code uses --model flag
+            args = ['--model', model, ...args, prompt];
+          } else if (providerId === 'gemini_cli') {
+            // Gemini CLI uses -m flag
+            args = ['-m', model, ...args, prompt];
+          } else {
+            // Default: just append prompt
+            args = [...args, prompt];
+          }
+        } else {
+          args = [...args, prompt];
+        }
+
         try {
           const result = await this.executeCliCommand(
             provider.command,
@@ -493,8 +506,8 @@ This is a known issue with @google/gemini-cli@0.3.4 and older Node.js versions.`
               tokens_used: this.estimateTokens(prompt + content),
               latency_ms: Date.now() - startTime,
               provider: providerId,
-              model: provider.model,
               mode: 'args',
+              model_used: model || 'cli_default',
               timestamp: new Date()
             };
           }
@@ -502,6 +515,37 @@ This is a known issue with @google/gemini-cli@0.3.4 and older Node.js versions.`
           lastErrorMessage = result.error;
         } catch (error) {
           lastErrorMessage = error instanceof Error ? error.message : String(error);
+          
+          // If model was specified and command failed, retry without model (graceful fallback)
+          if (model && lastErrorMessage.includes('model')) {
+            console.log(`[Polydev CLI] Model ${model} may be invalid for ${providerId}, retrying without model flag`);
+            try {
+              const fallbackArgs = Array.isArray(promptArgs) ? [...promptArgs, prompt] : [prompt];
+              const fallbackResult = await this.executeCliCommand(
+                provider.command,
+                fallbackArgs,
+                'args',
+                timeoutMs,
+                undefined
+              );
+              
+              if (!fallbackResult.error) {
+                const content = this.cleanCliResponse(fallbackResult.stdout || '');
+                return {
+                  success: true,
+                  content,
+                  tokens_used: this.estimateTokens(prompt + content),
+                  latency_ms: Date.now() - startTime,
+                  provider: providerId,
+                  mode: 'args',
+                  model_used: 'cli_default_fallback',
+                  timestamp: new Date()
+                };
+              }
+            } catch (fallbackError) {
+              // Fallback also failed, continue with original error
+            }
+          }
         }
       }
 
@@ -623,7 +667,7 @@ This is a known issue with @google/gemini-cli@0.3.4 and older Node.js versions.`
     });
   }
 
-  async executeCodexExec(executable, commandArgs, prompt, timeoutMs, model) {
+  async executeCodexExec(executable, commandArgs, prompt, timeoutMs, model = null) {
     if (!executable) {
       throw new Error('Missing Codex executable');
     }
@@ -634,18 +678,25 @@ This is a known issue with @google/gemini-cli@0.3.4 and older Node.js versions.`
 
     const workingDir = process.cwd();
     
-    // Build args with model configuration if specified
-    const modelArgs = model ? ['-c', `model="${model}"`] : [];
-    const args = [
-      ...commandArgs,
-      ...modelArgs,
+    // Build args with optional model flag
+    // Codex CLI uses -m or --model flag
+    let args = [...commandArgs];
+    
+    // Add model flag if specified
+    if (model) {
+      args.push('-m', model);
+      console.log(`[CLI Debug] Codex using model: ${model}`);
+    }
+    
+    // Add standard flags and prompt
+    args.push(
       '--sandbox',
       'workspace-write',
       '--skip-git-repo-check',
       '--cd',
       workingDir,
       prompt
-    ];
+    );
 
     return new Promise((resolve, reject) => {
       const baseTmp = process.env.POLYDEV_CLI_TMPDIR || process.env.TMPDIR || os.tmpdir();

package/mcp/stdio-wrapper.js

@@ -116,6 +116,11 @@ class StdioMCPWrapper {
     
     // Smart refresh scheduler (will be started after initialization)
     this.refreshScheduler = null;
+    
+    // Cache for user model preferences (provider -> model)
+    this.userModelPreferences = null;
+    this.modelPreferencesCacheTime = null;
+    this.MODEL_PREFERENCES_CACHE_TTL = 5 * 60 * 1000; // 5 minutes cache
   }
 
   loadManifest() {
@@ -452,12 +457,26 @@ class StdioMCPWrapper {
       // Use reasonable timeout for CLI responses (180 seconds for complex prompts)
       const gracefulTimeout = Math.min(timeout_ms, 180000);
 
+      // Fetch user's model preferences (cached, non-blocking on failure)
+      let modelPreferences = {};
+      try {
+        modelPreferences = await this.fetchUserModelPreferences();
+      } catch (prefError) {
+        console.error('[Stdio Wrapper] Model preferences fetch failed (will use CLI defaults):', prefError.message);
+      }
+
       let localResults = [];
 
       if (provider_id) {
         // Specific provider requested - use only that one
         console.error(`[Stdio Wrapper] Using specific provider: ${provider_id}`);
-        const result = await this.cliManager.sendCliPrompt(provider_id, prompt, mode, gracefulTimeout);
+        const model = modelPreferences[provider_id] || null;
+        if (model) {
+          console.error(`[Stdio Wrapper] Using user's preferred model for ${provider_id}: ${model}`);
+        } else {
+          console.error(`[Stdio Wrapper] No model preference for ${provider_id}, using CLI default`);
+        }
+        const result = await this.cliManager.sendCliPrompt(provider_id, prompt, mode, gracefulTimeout, model);
         localResults = [{ provider_id, ...result }];
       } else {
         // No specific provider - use ALL available local CLIs
@@ -473,7 +492,11 @@ class StdioMCPWrapper {
           // Run all CLI prompts concurrently
           const cliPromises = availableProviders.map(async (providerId) => {
             try {
-              const result = await this.cliManager.sendCliPrompt(providerId, prompt, mode, gracefulTimeout);
+              const model = modelPreferences[providerId] || null;
+              if (model) {
+                console.error(`[Stdio Wrapper] Using user's preferred model for ${providerId}: ${model}`);
+              }
+              const result = await this.cliManager.sendCliPrompt(providerId, prompt, mode, gracefulTimeout, model);
               return { provider_id: providerId, ...result };
             } catch (error) {
               console.error(`[Stdio Wrapper] CLI ${providerId} failed:`, error.message);
@@ -1020,6 +1043,87 @@ class StdioMCPWrapper {
     }
   }
 
+  /**
+   * Fetch user's model preferences from API keys
+   * Returns a map of CLI provider -> default_model
+   */
+  async fetchUserModelPreferences() {
+    // Check cache first
+    if (this.userModelPreferences && this.modelPreferencesCacheTime) {
+      const cacheAge = Date.now() - this.modelPreferencesCacheTime;
+      if (cacheAge < this.MODEL_PREFERENCES_CACHE_TTL) {
+        console.error('[Stdio Wrapper] Using cached model preferences');
+        return this.userModelPreferences;
+      }
+    }
+
+    console.error('[Stdio Wrapper] Fetching user model preferences from API...');
+    
+    try {
+      // Call the dedicated model-preferences endpoint
+      const response = await fetch('https://www.polydev.ai/api/model-preferences', {
+        method: 'GET',
+        headers: {
+          'Authorization': `Bearer ${this.userToken}`,
+          'User-Agent': 'polydev-stdio-wrapper/1.0.0'
+        }
+      });
+
+      if (!response.ok) {
+        console.error('[Stdio Wrapper] Failed to fetch model preferences:', response.status);
+        return this.userModelPreferences || {};
+      }
+
+      const result = await response.json();
+      
+      if (result.success && result.modelPreferences) {
+        // Cache the preferences
+        this.userModelPreferences = result.modelPreferences;
+        this.modelPreferencesCacheTime = Date.now();
+        
+        console.error('[Stdio Wrapper] Model preferences loaded:', JSON.stringify(result.modelPreferences));
+        return result.modelPreferences;
+      } else {
+        console.error('[Stdio Wrapper] No model preferences in response');
+        return this.userModelPreferences || {};
+      }
+
+    } catch (error) {
+      console.error('[Stdio Wrapper] Error fetching model preferences:', error.message);
+      return this.userModelPreferences || {};
+    }
+  }
+
+  /**
+   * Map provider name to CLI provider ID
+   */
+  mapProviderToCli(provider) {
+    const providerLower = (provider || '').toLowerCase().trim();
+    
+    // Map provider names to CLI tool IDs
+    const providerMap = {
+      'anthropic': 'claude_code',
+      'anthropic-ai': 'claude_code',
+      'claude': 'claude_code',
+      'openai': 'codex_cli',
+      'open-ai': 'codex_cli',
+      'gpt': 'codex_cli',
+      'google': 'gemini_cli',
+      'google-ai': 'gemini_cli',
+      'gemini': 'gemini_cli'
+    };
+    
+    return providerMap[providerLower] || null;
+  }
+
+  /**
+   * Get model for a specific CLI provider
+   */
+  async getModelForProvider(providerId) {
+    const preferences = await this.fetchUserModelPreferences();
+    return preferences[providerId] || null;
+  }
+
   async start() {
     console.log('Starting Polydev Stdio MCP Wrapper...');
     

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "polydev-ai",
-  "version": "1.6.0",
+  "version": "1.7.0",
   "description": "Agentic workflow assistant with CLI integration - get diverse perspectives from multiple LLMs when stuck or need enhanced reasoning",
   "keywords": [
     "mcp",