tachibot-mcp 2.0.5 → 2.0.7

package/dist/src/tools/gemini-tools.js

@@ -5,23 +5,16 @@
  */
 import { z } from "zod";
 import { validateToolInput } from "../utils/input-validator.js";
+import { GEMINI_MODELS } from "../config/model-constants.js";
 // NOTE: dotenv is loaded in server.ts before any imports
 // No need to reload here - just read from process.env
 // Gemini API configuration
 const GEMINI_API_KEY = process.env.GOOGLE_API_KEY || process.env.GEMINI_API_KEY;
 const GEMINI_API_URL = "https://generativelanguage.googleapis.com/v1beta";
-// Available Gemini models (2025 - Latest)
-export var GeminiModel;
-(function (GeminiModel) {
-    // Primary models (Gemini 2.5 - preferred)
-    GeminiModel["FLASH"] = "gemini-2.5-flash";
-    GeminiModel["PRO"] = "gemini-2.5-pro";
-    GeminiModel["FLASH_LITE"] = "gemini-2.5-flash-lite";
-})(GeminiModel || (GeminiModel = {}));
 /**
  * Call Gemini API directly
  */
-export async function callGemini(prompt, model = GeminiModel.PRO, systemPrompt, temperature = 0.7, skipValidation = false) {
+export async function callGemini(prompt, model = GEMINI_MODELS.GEMINI_3_PRO, systemPrompt, temperature = 0.7, skipValidation = false) {
     if (!GEMINI_API_KEY) {
         return `[Gemini API key not configured. Add GOOGLE_API_KEY to .env file]`;
     }
@@ -144,10 +137,16 @@ export const geminiQueryTool = {
     description: "Query Gemini",
     parameters: z.object({
         prompt: z.string(),
-        model: z.enum(["pro", "flash"]).optional().default("pro")
+        model: z.enum(["gemini-3", "pro", "flash"]).optional().default("gemini-3")
     }),
     execute: async (args, { log }) => {
-        const model = args.model === "flash" ? GeminiModel.FLASH : GeminiModel.PRO;
+        let model = GEMINI_MODELS.GEMINI_3_PRO; // Default to Gemini 3
+        if (args.model === "flash") {
+            model = GEMINI_MODELS.FLASH;
+        }
+        else if (args.model === "pro") {
+            model = GEMINI_MODELS.PRO;
+        }
         return await callGemini(args.prompt, model);
     }
 };
@@ -174,7 +173,7 @@ IMPORTANT: Output a detailed written response with:
 4. Quick feasibility assessment
 
 Provide your complete analysis as visible text output.`;
-        const response = await callGemini(args.prompt, GeminiModel.PRO, systemPrompt, 0.9);
+        const response = await callGemini(args.prompt, GEMINI_MODELS.GEMINI_3_PRO, systemPrompt, 0.9);
         // If multiple rounds requested, we could iterate here
         // For now, return the single response
         return response;
@@ -208,7 +207,7 @@ Provide:
 2. ${args.focus === 'security' ? 'Security vulnerabilities' : 'Issues found'}
 3. Specific recommendations for improvement
 4. Code quality score (1-10) with justification`;
-        return await callGemini(`Analyze this code:\n\n\`\`\`${args.language || ''}\n${args.code}\n\`\`\``, GeminiModel.PRO, systemPrompt, 0.3);
+        return await callGemini(`Analyze this code:\n\n\`\`\`${args.language || ''}\n${args.code}\n\`\`\``, GEMINI_MODELS.GEMINI_3_PRO, systemPrompt, 0.3);
     }
 };
 /**
@@ -236,7 +235,7 @@ Format your response clearly with:
 ${args.type === 'sentiment' ? '- Overall sentiment\n- Confidence score\n- Emotional indicators' : ''}
 ${args.type === 'entities' ? '- People\n- Organizations\n- Locations\n- Other entities' : ''}
 ${args.type === 'key-points' ? '- Main arguments\n- Supporting points\n- Conclusions' : ''}`;
-        return await callGemini(`Analyze this text:\n\n${args.text}`, GeminiModel.PRO, systemPrompt, 0.3);
+        return await callGemini(`Analyze this text:\n\n${args.text}`, GEMINI_MODELS.GEMINI_3_PRO, systemPrompt, 0.3);
     }
 };
 /**
@@ -269,7 +268,7 @@ Focus on:
 - Main ideas and key arguments
 - Important facts and figures
 - Conclusions and implications`;
-        return await callGemini(`Summarize this content:\n\n${args.content}`, GeminiModel.PRO, systemPrompt, 0.3);
+        return await callGemini(`Summarize this content:\n\n${args.content}`, GEMINI_MODELS.GEMINI_3_PRO, systemPrompt, 0.3);
     }
 };
 /**
@@ -302,7 +301,7 @@ Make it specific and visually descriptive.`;
 ${args.style ? `Style: ${args.style}` : ''}
 ${args.mood ? `Mood: ${args.mood}` : ''}
 ${args.details ? `Additional details: ${args.details}` : ''}`;
-        return await callGemini(userPrompt, GeminiModel.PRO, systemPrompt, 0.7);
+        return await callGemini(userPrompt, GEMINI_MODELS.GEMINI_3_PRO, systemPrompt, 0.7);
     }
 };
 /**

package/dist/src/tools/grok-enhanced.js

@@ -6,16 +6,20 @@ import { z } from "zod";
 import { config } from "dotenv";
 import * as path from 'path';
 import { fileURLToPath } from 'url';
+import { getGrokApiKey, hasGrokApiKey } from "../utils/api-keys.js";
 const __filename = fileURLToPath(import.meta.url);
 const __dirname = path.dirname(__filename);
 config({ path: path.resolve(__dirname, '../../../.env') });
 // Grok API configuration
-const GROK_API_KEY = process.env.GROK_API_KEY || process.env.XAI_API_KEY;
+const GROK_API_KEY = getGrokApiKey();
 const GROK_API_URL = "https://api.x.ai/v1/chat/completions";
-// Grok models - Updated 2025-11-07 with new fast models
+// Grok models - Updated 2025-11-22 with correct API model names
 export var GrokModel;
 (function (GrokModel) {
-    // New fast models (2025) - PRIMARY USE
+    // Grok 4.1 models (Nov 2025) - LATEST & BEST (verified working)
+    GrokModel["GROK_4_1_FAST_REASONING"] = "grok-4-1-fast-reasoning";
+    GrokModel["GROK_4_1_FAST"] = "grok-4-1-fast-non-reasoning";
+    // Grok 4 fast models (2025) - Still good
     GrokModel["CODE_FAST"] = "grok-code-fast-1";
     GrokModel["GROK_4_FAST_REASONING"] = "grok-4-fast-reasoning";
     GrokModel["GROK_4_FAST"] = "grok-4-fast-non-reasoning";
@@ -35,7 +39,7 @@ export async function callGrokEnhanced(messages, options = {}) {
             content: `[Grok API key not configured. Add GROK_API_KEY or XAI_API_KEY to .env file]`
         };
     }
-    const { model = GrokModel.GROK_4_FAST_REASONING, // Changed: Use cheap fast reasoning by default
+    const { model = GrokModel.GROK_4_1_FAST_REASONING, // Updated 2025-11-22: Use latest Grok 4.1 by default
     temperature = 0.7, maxTokens = options.useHeavy ? 100000 : 4000, enableLiveSearch = false, searchSources = 100, // Default to 100 sources for cost control
     searchDomains = [], structuredOutput = false } = options;
     try {
@@ -140,9 +144,9 @@ Cite your sources when using web data.`
                 content: query
             }
         ];
-        log?.info(`Grok Scout: ${variant} research with ${enableLiveSearch ? 'live search' : 'knowledge base'} (using ${enableLiveSearch ? 'grok-4-fast-reasoning with search' : 'grok-4-fast-reasoning'})`);
+        log?.info(`Grok Scout: ${variant} research with ${enableLiveSearch ? 'live search' : 'knowledge base'} (using grok-4-1-fast-reasoning with enhanced reasoning)`);
         const result = await callGrokEnhanced(messages, {
-            model: GrokModel.GROK_4_FAST_REASONING, // Changed: Use fast reasoning (works with search)
+            model: GrokModel.GROK_4_1_FAST_REASONING, // Updated 2025-11-21: Use latest Grok 4.1
             enableLiveSearch,
             searchSources,
             searchDomains,
@@ -197,11 +201,11 @@ ${enableLiveSearch ? 'Use live search for current information when needed.' : ''
                 content: problem
             }
         ];
-        const modelName = useHeavy ? 'Grok-4-Heavy' : 'Grok-4-Fast-Reasoning';
-        const costInfo = useHeavy ? '$3/$15 (expensive!)' : '$0.20/$0.50 (cheap!)';
+        const modelName = useHeavy ? 'Grok-4-Heavy' : 'Grok-4.1';
+        const costInfo = useHeavy ? '$3/$15 (expensive!)' : '$0.20/$0.50 (latest!)';
         log?.info(`Using ${modelName} (${approach}) with ${enableLiveSearch ? 'live search' : 'knowledge base'} - Cost: ${costInfo}`);
         const result = await callGrokEnhanced(messages, {
-            model: useHeavy ? GrokModel.GROK_4_HEAVY : GrokModel.GROK_4_FAST_REASONING, // Changed: Use fast reasoning unless heavy requested
+            model: useHeavy ? GrokModel.GROK_4_HEAVY : GrokModel.GROK_4_1_FAST_REASONING, // Updated 2025-11-21: Use latest Grok 4.1
             useHeavy,
             enableLiveSearch,
             searchSources: 50,
@@ -253,14 +257,14 @@ export const grokFunctionTool = {
         ];
         // Make request with tools
         const requestBody = {
-            model: args.useHeavy ? GrokModel.GROK_4_HEAVY : GrokModel.GROK_4_FAST_REASONING, // Changed: Use fast reasoning unless heavy requested
+            model: args.useHeavy ? GrokModel.GROK_4_HEAVY : GrokModel.GROK_4_1_FAST, // Updated 2025-11-22: Use tool-calling optimized Grok 4.1 Fast Non-Reasoning
             messages,
             tools,
             tool_choice: "auto", // Let Grok decide when to call functions
             max_tokens: args.useHeavy ? 50000 : 5000,
             temperature: 0.3 // Lower for function calling
         };
-        log?.info(`Function calling with ${args.useHeavy ? 'Grok-4-Heavy ($3/$15)' : 'Grok-4-Fast-Reasoning ($0.20/$0.50)'}`);
+        log?.info(`Function calling with ${args.useHeavy ? 'Grok-4-Heavy ($3/$15)' : 'Grok-4.1-Fast ($0.20/$0.50, tool-calling optimized)'}`);
         try {
             const response = await fetch(GROK_API_URL, {
                 method: "POST",
@@ -318,13 +322,13 @@ Limit search to ${max_search_results} sources for cost control.`
                 content: `Search for: ${query}`
             }
         ];
-        log?.info(`Grok Search: ${max_search_results} sources, recency: ${recency} (using grok-4-fast-reasoning with search)`);
+        log?.info(`Grok Search: ${max_search_results} sources, recency: ${recency} (using grok-4-1-fast-reasoning with enhanced reasoning)`);
         // Extract domains from sources if specified
         const domains = sources
             ?.filter((s) => s.allowed_websites)
             ?.flatMap((s) => s.allowed_websites) || [];
         const result = await callGrokEnhanced(messages, {
-            model: GrokModel.GROK_4_FAST_REASONING, // Changed: Use fast reasoning with search
+            model: GrokModel.GROK_4_1_FAST_REASONING, // Updated 2025-11-21: Use latest Grok 4.1 with search
             enableLiveSearch: true,
             searchSources: max_search_results,
             searchDomains: domains,
@@ -340,7 +344,7 @@ Limit search to ${max_search_results} sources for cost control.`
  * Check if Grok is available
  */
 export function isGrokAvailable() {
-    return !!(GROK_API_KEY);
+    return hasGrokApiKey();
 }
 /**
  * Get Grok configuration status
@@ -348,10 +352,10 @@ export function isGrokAvailable() {
 export function getGrokStatus() {
     return {
         available: isGrokAvailable(),
-        model: GrokModel.GROK_4_FAST_REASONING,
+        model: GrokModel.GROK_4_1_FAST_REASONING,
         features: [
-            'Fast reasoning (grok-4-fast-reasoning: $0.20/$0.50, 2M→4M context)',
-            'Code specialist (grok-code-fast-1: 92 tok/sec)',
+            'Grok 4.1 Fast Reasoning (Nov 2025): Enhanced reasoning, creativity & emotional intelligence ($0.20/$0.50, 2M context)',
+            'Grok 4.1 Fast Non-Reasoning: Tool-calling optimized, agentic workflows ($0.20/$0.50, 2M context)',
             'Heavy mode available (grok-4-0709: $3/$15, use sparingly)',
             'Live web search with citations',
             'Function calling',

package/dist/src/tools/grok-tools.js

@@ -8,16 +8,20 @@ import * as path from 'path';
 import { fileURLToPath } from 'url';
 import { grokSearchTool } from './grok-enhanced.js';
 import { validateToolInput } from "../utils/input-validator.js";
+import { getGrokApiKey, hasGrokApiKey } from "../utils/api-keys.js";
 const __filename = fileURLToPath(import.meta.url);
 const __dirname = path.dirname(__filename);
 config({ path: path.resolve(__dirname, '../../../.env') });
 // Grok API configuration
-const GROK_API_KEY = process.env.GROK_API_KEY;
+const GROK_API_KEY = getGrokApiKey();
 const GROK_API_URL = "https://api.x.ai/v1/chat/completions";
-// Available Grok models - Updated 2025-11-07 with new fast models
+// Available Grok models - Updated 2025-11-22 with correct API model names
 export var GrokModel;
 (function (GrokModel) {
-    // New fast models (2025) - PRIMARY USE
+    // Grok 4.1 models (Nov 2025) - LATEST & BEST (verified working)
+    GrokModel["GROK_4_1_FAST_REASONING"] = "grok-4-1-fast-reasoning";
+    GrokModel["GROK_4_1_FAST"] = "grok-4-1-fast-non-reasoning";
+    // Grok 4 fast models (2025) - Still good
     GrokModel["CODE_FAST"] = "grok-code-fast-1";
     GrokModel["GROK_4_FAST_REASONING"] = "grok-4-fast-reasoning";
     GrokModel["GROK_4_FAST"] = "grok-4-fast-non-reasoning";
@@ -28,11 +32,11 @@ export var GrokModel;
 /**
  * Call Grok API
  */
-export async function callGrok(messages, model = GrokModel.GROK_4_FAST_REASONING, // Changed: Use cheap fast reasoning by default
+export async function callGrok(messages, model = GrokModel.GROK_4_1_FAST_REASONING, // Updated 2025-11-22: Use latest Grok 4.1 by default
 temperature = 0.7, maxTokens = 16384, // Increased default for comprehensive responses
 forceVisibleOutput = true) {
     if (!GROK_API_KEY) {
-        return `[Grok API key not configured. Add GROK_API_KEY to .env file]`;
+        return `[Grok API key not configured. Add XAI_API_KEY to .env file]`;
     }
     // Validate and sanitize message content
     const validatedMessages = messages.map((msg) => {
@@ -44,7 +48,9 @@ forceVisibleOutput = true) {
     });
     try {
         // For Grok 4 models, we need to handle reasoning tokens specially
-        const isGrok4 = model === GrokModel.GROK_4_FAST_REASONING ||
+        const isGrok4 = model === GrokModel.GROK_4_1_FAST_REASONING ||
+            model === GrokModel.GROK_4_1_FAST ||
+            model === GrokModel.GROK_4_FAST_REASONING ||
             model === GrokModel.GROK_4_FAST ||
             model === GrokModel.GROK_4_HEAVY;
         // Adjust prompt for Grok 4 to ensure visible output
@@ -113,7 +119,7 @@ export const grokReasonTool = {
         const messages = [
             {
                 role: "system",
-                content: `You are Grok, an expert at logical reasoning and problem-solving.
+                content: `You are Grok 4.1, an expert at logical reasoning and problem-solving with enhanced emotional intelligence.
 ${approachPrompts[approach]}.
 ${context ? `Context: ${context}` : ''}`
             },
@@ -122,8 +128,8 @@ ${context ? `Context: ${context}` : ''}`
                 content: problem
             }
         ];
-        // Use GROK_4_FAST_REASONING by default (3x cheaper!), GROK_4_HEAVY only if explicitly requested
-        const model = useHeavy ? GrokModel.GROK_4_HEAVY : GrokModel.GROK_4_FAST_REASONING;
+        // Use GROK_4_1_FAST_REASONING by default (latest with enhanced reasoning!), GROK_4_HEAVY only if explicitly requested
+        const model = useHeavy ? GrokModel.GROK_4_HEAVY : GrokModel.GROK_4_1_FAST_REASONING;
         const maxTokens = useHeavy ? 100000 : 16384; // 100k for heavy, 16k for normal reasoning
         log?.info(`Using Grok model: ${model} for deep reasoning (max tokens: ${maxTokens}, cost: ${useHeavy ? 'expensive $3/$15' : 'cheap $0.20/$0.50'})`);
         return await callGrok(messages, model, 0.7, maxTokens, true);
@@ -154,7 +160,7 @@ export const grokCodeTool = {
         const messages = [
             {
                 role: "system",
-                content: `You are Grok Code Fast, an expert programmer and code analyst.
+                content: `You are Grok 4.1 Fast, an expert programmer and code analyst with advanced tool-calling capabilities.
 Task: ${taskPrompts[task]}
 ${language ? `Language: ${language}` : ''}
 ${requirements ? `Requirements: ${requirements}` : ''}`
@@ -164,8 +170,8 @@ ${requirements ? `Requirements: ${requirements}` : ''}`
                 content: `Code:\n\`\`\`${language || ''}\n${code}\n\`\`\``
             }
         ];
-        log?.info(`Using Grok Code Fast (92 tok/sec coding specialist)`);
-        return await callGrok(messages, GrokModel.CODE_FAST, 0.2, 4000, true);
+        log?.info(`Using Grok 4.1 Fast Non-Reasoning (2M context, tool-calling optimized, $0.20/$0.50)`);
+        return await callGrok(messages, GrokModel.GROK_4_1_FAST, 0.2, 4000, true);
     }
 };
 /**
@@ -208,8 +214,8 @@ Analyze the issue systematically:
                 content: prompt
             }
         ];
-        log?.info(`Using Grok Code Fast for debugging (specialized code model)`);
-        return await callGrok(messages, GrokModel.CODE_FAST, 0.3, 3000, true);
+        log?.info(`Using Grok 4.1 Fast Non-Reasoning for debugging (tool-calling optimized, $0.20/$0.50)`);
+        return await callGrok(messages, GrokModel.GROK_4_1_FAST, 0.3, 3000, true);
     }
 };
 /**
@@ -240,8 +246,8 @@ ${constraints ? `Constraints: ${constraints}` : ''}`
                 content: requirements
             }
         ];
-        log?.info(`Using Grok 4 Fast Reasoning for architecture (cheap reasoning model)`);
-        return await callGrok(messages, GrokModel.GROK_4_FAST_REASONING, 0.6, 4000, true);
+        log?.info(`Using Grok 4.1 Fast Reasoning for architecture (latest model, $0.20/$0.50)`);
+        return await callGrok(messages, GrokModel.GROK_4_1_FAST_REASONING, 0.6, 4000, true);
     }
 };
 /**
@@ -271,9 +277,9 @@ ${constraints ? `Constraints: ${constraints}` : 'No constraints - think freely!'
                 content: `Brainstorm creative solutions for: ${topic}`
             }
         ];
-        // Use GROK_4_FAST for creative brainstorming (cheap, fast), GROK_4_HEAVY only if explicitly requested
-        const model = forceHeavy ? GrokModel.GROK_4_HEAVY : GrokModel.GROK_4_FAST;
-        log?.info(`Brainstorming with Grok model: ${model} (Heavy: ${forceHeavy}, cost: ${forceHeavy ? 'expensive $3/$15' : 'cheap $0.20/$0.50'})`);
+        // Use GROK_4_1_FAST_REASONING for creative brainstorming (needs reasoning for creativity), GROK_4_HEAVY only if explicitly requested
+        const model = forceHeavy ? GrokModel.GROK_4_HEAVY : GrokModel.GROK_4_1_FAST_REASONING;
+        log?.info(`Brainstorming with Grok model: ${model} (Heavy: ${forceHeavy}, cost: ${forceHeavy ? 'expensive $3/$15' : 'cheap $0.20/$0.50 - latest 4.1'})`);
         return await callGrok(messages, model, 0.95, 4000); // High temperature for creativity
     }
 };
@@ -281,7 +287,7 @@ ${constraints ? `Constraints: ${constraints}` : 'No constraints - think freely!'
  * Check if Grok is available
  */
 export function isGrokAvailable() {
-    return !!GROK_API_KEY;
+    return hasGrokApiKey();
 }
 export function getAllGrokTools() {
     if (!isGrokAvailable()) {

package/dist/src/tools/openai-tools.js

@@ -70,12 +70,14 @@ const ResponsesAPISchema = z.object({
         total_tokens: z.number().optional()
     }).optional()
 });
-// Available OpenAI models (GPT-5.1 family)
+// Available OpenAI GPT-5 models (optimized for Claude Code)
 export var OpenAI51Model;
 (function (OpenAI51Model) {
     OpenAI51Model["FULL"] = "gpt-5.1";
+    OpenAI51Model["PRO"] = "gpt-5-pro";
     OpenAI51Model["CODEX_MINI"] = "gpt-5.1-codex-mini";
     OpenAI51Model["CODEX"] = "gpt-5.1-codex";
+    OpenAI51Model["CODEX_MAX"] = "gpt-5.1-codex-max";
 })(OpenAI51Model || (OpenAI51Model = {}));
 /**
  * Call OpenAI API with model fallback support
@@ -111,34 +113,33 @@ reasoningEffort = "low", requireConfirmation = false, skipValidation = false) {
     for (const currentModel of modelsToTry) {
         console.error(`🔍 TRACE: Trying model: ${currentModel}`);
         try {
-            // GPT-5.1 models use /v1/responses, others use /v1/chat/completions
-            const isGPT51 = currentModel.startsWith('gpt-5.1');
-            const endpoint = isGPT51 ? OPENAI_RESPONSES_URL : OPENAI_CHAT_URL;
+            // Codex models use /v1/responses, non-codex use /v1/chat/completions
+            const isCodex = currentModel.includes('codex');
+            const endpoint = isCodex ? OPENAI_RESPONSES_URL : OPENAI_CHAT_URL;
             let requestBody;
-            // GPT-5.1 uses Responses API format, others use Chat Completions format
-            if (isGPT51) {
-                // Responses API format - NO temperature, use reasoning.effort instead
+            if (isCodex) {
+                // Responses API format for codex models
                 requestBody = {
                     model: currentModel,
                     input: validatedMessages,
                     max_output_tokens: maxTokens,
                     stream: false,
                     reasoning: {
-                        effort: reasoningEffort // "none", "low", "medium", "high"
+                        effort: reasoningEffort
                     }
                 };
             }
             else {
-                // Chat Completions format
+                // Chat Completions format for non-codex GPT-5 models (gpt-5.1, gpt-5-pro)
                 requestBody = {
                     model: currentModel,
                     messages: validatedMessages,
                     temperature,
-                    max_tokens: maxTokens,
+                    max_completion_tokens: maxTokens, // GPT-5 requires max_completion_tokens (not max_tokens)
                     stream: false
                 };
             }
-            console.error(`🔍 TRACE: Using ${isGPT51 ? '/v1/responses' : '/v1/chat/completions'} endpoint`);
+            console.error(`🔍 TRACE: Using ${isCodex ? '/v1/responses' : '/v1/chat/completions'} endpoint`);
             const response = await fetch(endpoint, {
                 method: "POST",
                 headers: {
@@ -159,16 +160,15 @@ reasoningEffort = "low", requireConfirmation = false, skipValidation = false) {
                 throw new Error(lastError);
             }
             const rawData = await response.json();
-            // Parse based on API type - they have DIFFERENT response formats!
+            // Parse based on API type
             let rawContent;
-            if (isGPT51) {
-                // Validate and parse Responses API format
+            if (isCodex) {
+                // Responses API format
                 const parseResult = ResponsesAPISchema.safeParse(rawData);
                 if (parseResult.success) {
                     const data = parseResult.data;
                     const messageOutput = data.output.find(item => item.type === 'message');
                     rawContent = messageOutput?.content?.[0]?.text;
-                    // Capture reasoning info
                     if (data.reasoning) {
                         console.error(`🔍 TRACE: Reasoning effort: ${data.reasoning.effort}`);
                     }
@@ -178,7 +178,7 @@ reasoningEffort = "low", requireConfirmation = false, skipValidation = false) {
                 }
             }
             else {
-                // Validate and parse Chat Completions API format
+                // Chat Completions format
                 const parseResult = ChatCompletionResponseSchema.safeParse(rawData);
                 if (parseResult.success) {
                     const chatData = parseResult.data;
@@ -225,17 +225,16 @@ reasoningEffort = "low", skipValidation = false) {
         return { ...msg, content: validation.sanitized };
     });
     try {
-        // GPT-5.1 models use /v1/responses, others use /v1/chat/completions
-        const isGPT51 = model.startsWith('gpt-5.1');
-        const endpoint = isGPT51 ? OPENAI_RESPONSES_URL : OPENAI_CHAT_URL;
+        // Codex models use /v1/responses, non-codex use /v1/chat/completions
+        const isCodex = model.includes('codex');
+        const endpoint = isCodex ? OPENAI_RESPONSES_URL : OPENAI_CHAT_URL;
         let requestBody;
-        // GPT-5.1 uses Responses API format, others use Chat Completions format
-        if (isGPT51) {
-            // Responses API format - NO temperature, use reasoning.effort instead
+        if (isCodex) {
+            // Responses API format for codex models
             requestBody = {
                 model: model,
                 input: validatedMessages,
-                max_output_tokens: maxTokens,
+                max_output_tokens: maxTokens, // NOT max_completion_tokens or max_tokens!
                 stream: false,
                 reasoning: {
                     effort: reasoningEffort // "none", "low", "medium", "high"
@@ -243,17 +242,17 @@ reasoningEffort = "low", skipValidation = false) {
             };
         }
         else {
-            // Chat Completions format
+            // Chat Completions format for non-codex GPT-5 models (gpt-5.1, gpt-5-pro)
             requestBody = {
                 model: model,
                 messages: validatedMessages,
                 temperature,
-                max_tokens: maxTokens,
+                max_completion_tokens: maxTokens, // GPT-5 requires max_completion_tokens (not max_tokens)
                 stream: false
             };
         }
-        console.error(`🔍 TRACE: Using ${isGPT51 ? '/v1/responses' : '/v1/chat/completions'} endpoint`);
-        console.error(`🔍 TRACE: Model params: max_tokens=${maxTokens}, temperature=${temperature}${isGPT51 ? `, reasoning_effort=${reasoningEffort}` : ''}`);
+        console.error(`🔍 TRACE: Using ${isCodex ? '/v1/responses' : '/v1/chat/completions'} endpoint`);
+        console.error(`🔍 TRACE: Model params: ${isCodex ? `max_output_tokens=${maxTokens}, reasoning_effort=${reasoningEffort}` : `max_completion_tokens=${maxTokens}, temperature=${temperature}`}`);
         const response = await fetch(endpoint, {
             method: "POST",
             headers: {
@@ -270,7 +269,7 @@ reasoningEffort = "low", skipValidation = false) {
         const rawData = await response.json();
         // Parse based on API type - they have DIFFERENT response formats!
         let rawContent;
-        if (isGPT51) {
+        if (isCodex) {
             // Validate and parse Responses API format
             const parseResult = ResponsesAPISchema.safeParse(rawData);
             if (parseResult.success) {
@@ -378,7 +377,7 @@ export const gpt5MiniReasonTool = {
     }
 };
 export const openaiGpt5ReasonTool = {
-    name: "openai_gpt5_reason",
+    name: "openai_reason",
     description: "Mathematical reasoning using GPT-5.1 with high reasoning effort",
     parameters: z.object({
         query: z.string(),
@@ -409,37 +408,6 @@ ${args.context ? `Context: ${args.context}` : ''}`
         return await callOpenAI(messages, OpenAI51Model.FULL, 0.7, 4000, "high");
     }
 };
-/**
- * OpenAI Compare Tool
- * Multi-option comparison and consensus building using GPT-5.1-codex-mini
- */
-export const openaiCompareTool = {
-    name: "openai_compare",
-    description: "Multi-model consensus",
-    parameters: z.object({
-        topic: z.string(),
-        options: z.array(z.string()),
-        criteria: z.string().optional(),
-        includeRecommendation: z.boolean().optional().default(true)
-    }),
-    execute: async (args, { log }) => {
-        const optionsList = args.options.map((opt, i) => `${i + 1}. ${opt}`).join('\n');
-        const messages = [
-            {
-                role: "system",
-                content: `You are an expert at comparative analysis and decision-making.
-Compare the given options systematically.
-${args.criteria ? `Criteria: ${args.criteria}` : 'Consider: pros, cons, trade-offs, and suitability'}
-${args.includeRecommendation ? 'Provide a clear recommendation with justification.' : ''}`
-            },
-            {
-                role: "user",
-                content: `Topic: ${args.topic}\n\nOptions:\n${optionsList}`
-            }
-        ];
-        return await callOpenAI(messages, OpenAI51Model.CODEX_MINI, 0.7, 3000, "low");
-    }
-};
 /**
  * OpenAI Brainstorm Tool
  * Creative ideation and brainstorming
@@ -582,7 +550,6 @@ export function getAllOpenAITools() {
     }
     return [
         openaiGpt5ReasonTool, // GPT-5.1 reasoning (high effort)
-        openaiCompareTool, // GPT-5.1-codex-mini comparison (low effort)
         openAIBrainstormTool, // GPT-5.1-codex-mini brainstorming (medium effort)
         openaiCodeReviewTool, // GPT-5.1-codex-mini code review (medium effort)
         openaiExplainTool // GPT-5.1-codex-mini explanations (low effort)