tachibot-mcp 2.0.5 → 2.0.7

package/dist/src/workflows/tool-mapper.js

@@ -2,12 +2,13 @@
  * Tool Mapper - Maps workflow tool names to actual MCP tool implementations
  * Enables workflows to call real tools instead of returning placeholders
  */
-import { callGemini, GeminiModel } from "../tools/gemini-tools.js";
+import { callGemini } from "../tools/gemini-tools.js";
 import { getAllPerplexityTools } from "../tools/perplexity-tools.js";
 import { callOpenAI } from "../tools/openai-tools.js";
 import { callGrok, GrokModel } from "../tools/grok-tools.js";
-import { GPT51_MODELS, TOOL_DEFAULTS, } from "../config/model-constants.js";
+import { GPT51_MODELS, TOOL_DEFAULTS, GEMINI_MODELS, } from "../config/model-constants.js";
 import { validateToolInput } from "../utils/input-validator.js";
+import { hasGrokApiKey } from "../utils/api-keys.js";
 // Lazy load OpenRouter for Qwen models
 let callOpenRouter = null;
 let OpenRouterModel = null;
@@ -74,7 +75,8 @@ export async function executeWorkflowTool(toolName, input, options = {}) {
         if (typeof input === "string")
             return input;
         // Try all common parameter names (order matters!)
-        return input.requirements || // qwen_coder, task-specific
+        return input.thought || // think tool reflection
+            input.requirements || // qwen_coder, task-specific
             input.problem || // brainstorm, reasoning tools
             input.query || // search/ask tools
             input.topic || // research tools
@@ -144,7 +146,7 @@ export async function executeWorkflowTool(toolName, input, options = {}) {
             case "gemini_brainstorm":
             case "gemini_analyze_code":
             case "gemini_analyze_text":
-                actualModel = model === "flash" ? GeminiModel.FLASH : GeminiModel.PRO;
+                actualModel = model === "flash" ? GEMINI_MODELS.FLASH : GEMINI_MODELS.GEMINI_3_PRO;
                 return buildResult(await callGemini(prompt, actualModel, systemPrompt, temperature, options.skipValidation || false), actualModel);
             // ============ PERPLEXITY TOOLS ============
             case "perplexity_ask":
@@ -209,26 +211,51 @@ export async function executeWorkflowTool(toolName, input, options = {}) {
                 return buildResult(await callOpenAI(toMessages(prompt, systemPrompt), actualModel, temperature, maxTokens, "low", // reasoningEffort
                 false, // requireConfirmation
                 options.skipValidation || false), actualModel);
+            case "openai_reason":
+                // GPT-5 Pro with high reasoning effort for complex reasoning
+                actualModel = GPT51_MODELS.PRO;
+                return buildResult(await callOpenAI(toMessages(prompt, systemPrompt), actualModel, temperature, maxTokens, "high", // reasoningEffort
+                false, // requireConfirmation
+                options.skipValidation || false), actualModel);
+            case "openai_code_review":
+                // GPT-5.1 codex-mini for code review (medium reasoning)
+                actualModel = GPT51_MODELS.CODEX_MINI;
+                return buildResult(await callOpenAI(toMessages(prompt, systemPrompt || "You are an expert code reviewer. Provide thorough code review with specific, actionable feedback."), actualModel, 0.3, // Low temperature for consistent code review
+                maxTokens, "medium", // reasoningEffort
+                false, options.skipValidation || false), actualModel);
+            case "openai_explain":
+                // GPT-5.1 codex-mini for explanations (low reasoning)
+                actualModel = GPT51_MODELS.CODEX_MINI;
+                return buildResult(await callOpenAI(toMessages(prompt, systemPrompt || "You are an expert educator. Provide clear, engaging explanations."), actualModel, temperature, maxTokens, "low", // reasoningEffort
+                false, options.skipValidation || false), actualModel);
             case "gpt5_analyze":
                 return buildResult(await callOpenAI(toMessages(prompt, systemPrompt), GPT51_MODELS.CODEX_MINI, 0.7, maxTokens), GPT51_MODELS.CODEX_MINI);
             case "openai_reason":
                 return buildResult(await callOpenAI(toMessages(prompt, systemPrompt), GPT51_MODELS.CODEX_MINI, temperature, maxTokens), GPT51_MODELS.CODEX_MINI);
             // ============ GPT-5 TOOLS ============
             case "gpt5":
+                // Map to flagship gpt-5.1
+                const gpt5Full = GPT51_MODELS.FULL; // gpt-5.1 flagship
+                return buildResult(await callOpenAI(toMessages(prompt, systemPrompt), gpt5Full, 0.7, maxTokens, "medium"), gpt5Full);
             case "gpt5_mini":
-            case "gpt5_nano":
-                // Map old names to new GPT-5.1 models
-                const gpt51Model = GPT51_MODELS.CODEX_MINI; // Always use cost-efficient codex-mini
-                return buildResult(await callOpenAI(toMessages(prompt, systemPrompt), gpt51Model, 0.7, maxTokens, "low"), gpt51Model);
+                // Map to gpt-5.1-codex-mini for code tasks (most workflows use for code)
+                const gpt5CodexMini = GPT51_MODELS.CODEX_MINI; // gpt-5.1-codex-mini
+                return buildResult(await callOpenAI(toMessages(prompt, systemPrompt), gpt5CodexMini, 0.7, maxTokens, "low"), gpt5CodexMini);
             // ============ GROK TOOLS ============
             case "grok":
             case "grok_reason":
-            case "grok_code":
-            case "grok_debug":
             case "grok_brainstorm":
-            case "grok_heavy": // Grok Heavy is just grok-4-0709 with more backend resources
             case "grok_search":
-                actualModel = GrokModel.GROK_4_FAST_REASONING; // Using fast reasoning (2M context, $0.20/$0.50)
+                // Use reasoning model for reasoning/creative tasks
+                actualModel = GrokModel.GROK_4_1_FAST_REASONING; // Latest 4.1 (2M context, $0.20/$0.50)
+                return buildResult(await callGrok(toMessages(prompt, systemPrompt), actualModel, temperature, maxTokens), actualModel);
+            case "grok_code":
+            case "grok_debug":
+                // Use non-reasoning model for code/debug (tool-calling optimized)
+                actualModel = GrokModel.GROK_4_1_FAST; // Latest 4.1 non-reasoning (2M context, $0.20/$0.50)
+                return buildResult(await callGrok(toMessages(prompt, systemPrompt), actualModel, temperature, maxTokens), actualModel);
+            case "grok_heavy": // Grok Heavy is grok-4-0709 with extended context
+                actualModel = GrokModel.GROK_4_HEAVY; // Expensive $3/$15
                 return buildResult(await callGrok(toMessages(prompt, systemPrompt), actualModel, temperature, maxTokens), actualModel);
             // ============ ADVANCED MODES ============
             case "verifier":
@@ -381,9 +408,9 @@ export function getAvailableTools() {
         tools.push("perplexity_ask", "perplexity_research", "perplexity_reason", "perplexity_code_search");
     }
     if (process.env.OPENAI_API_KEY) {
-        tools.push("openai_brainstorm", "gpt5_analyze", "openai_reason", "gpt5", "gpt5_mini", "gpt5_nano");
+        tools.push("openai_brainstorm", "openai_reason", "openai_code_review", "openai_explain", "gpt5_analyze", "openai_reason", "gpt5", "gpt5_mini");
     }
-    if (process.env.XAI_API_KEY) {
+    if (hasGrokApiKey()) {
         tools.push("grok", "grok_reason", "grok_code", "grok_debug", "grok_brainstorm", "grok_search");
     }
     // Add modes if available

package/docs/API_KEYS.md

@@ -48,7 +48,7 @@ TachiBot MCP works with multiple AI providers to offer diverse capabilities. You
 |----------|--------------|----------------|
 | **Perplexity** | Research, web search | `perplexity_ask`, `perplexity_research`, `perplexity_reason`, `scout` (default) |
 | **Grok/xAI** | Live search, reasoning | `grok_search`, `grok_reason`, `grok_code`, `grok_debug`, `grok_architect`, `grok_brainstorm`, `scout` (with grok) |
-| **OpenAI** | GPT-5 models | `openai_brainstorm`, `openai_compare`, `focus` (some modes), `verifier`, `challenger` |
+| **OpenAI** | GPT-5 models | `openai_brainstorm`, `openai_reason`, `openai_code_review`, `openai_explain`, `focus` (some modes), `verifier`, `challenger` |
 | **Google** | Gemini models | `gemini_brainstorm`, `gemini_analyze_code`, `gemini_analyze_text`, `verifier`, `scout` |
 | **OpenRouter** | Qwen models | `qwen_coder`, `qwen_competitive` |
 
@@ -116,15 +116,19 @@ Grok (by xAI) provides live web search, reasoning, and code analysis.
 
 #### Models Available
 
-- **grok-4** - Latest reasoning model
-- **grok-4-0709** - Specific version
-- **grok-4-heavy** - Extended context (256k tokens)
+- **grok-4-1-fast-reasoning** - Latest (Nov 2025): Enhanced reasoning, creativity & emotional intelligence (2M context)
+- **grok-4-1-fast-non-reasoning** - Tool-calling optimized: Fast inference, agentic workflows (2M context)
+- **grok-4-fast-reasoning** - Previous reasoning model
+- **grok-4-0709** - Heavy model (expensive, use sparingly)
+- **grok-code-fast-1** - Coding specialist
 
 #### Pricing
 
 | Model | Input | Output | Notes |
 |-------|-------|--------|-------|
-| Grok-4 | $5.00 / 1M tokens | $15.00 / 1M tokens | Standard |
+| Grok-4.1 | $0.20 / 1M tokens | $0.50 / 1M tokens | Latest & best! |
+| Grok-4.1-fast | $0.20 / 1M tokens | $0.50 / 1M tokens | Tool-calling optimized |
+| Grok-4 | $5.00 / 1M tokens | $15.00 / 1M tokens | Previous version |
 | Grok-4-heavy | $10.00 / 1M tokens | $30.00 / 1M tokens | 256k context |
 | **Live Search** | **$5 / 1k sources** | - | Extra cost per search! |
 
@@ -203,7 +207,7 @@ OpenAI provides GPT-5 models for brainstorming, comparison, and reasoning.
 
 - Single `openai_brainstorm` (gpt-5-mini): ~$0.01 - $0.03
 - Single `openai_brainstorm` (gpt-5): ~$0.15 - $0.40
-- Single `openai_compare`: ~$0.02 - $0.05
+- Single `openai_code_review`: ~$0.02 - $0.05
 
 **Tip:** Use `model: "gpt-5-mini"` by default, only use `gpt-5` for complex tasks.
 

package/docs/TOOLS_REFERENCE.md

@@ -22,9 +22,8 @@
   - [grok_architect](#grok_architect)
   - [grok_brainstorm](#grok_brainstorm)
 - [OpenAI Suite](#openai-suite)
-  - [openai_gpt5_reason](#openai_gpt5_reason)
+  - [openai_reason](#openai_reason)
   - [openai_brainstorm](#openai_brainstorm)
-  - [openai_compare](#openai_compare)
   - [openai_code_review](#openai_code_review)
   - [openai_explain](#openai_explain)
 - [Gemini Suite](#gemini-suite)
@@ -514,7 +513,7 @@ perplexity_reason({
 
 ### grok_search
 
-Cost-optimized web search using Grok-4's live search with advanced filtering.
+Cost-optimized web search using Grok-4.1's live search with advanced filtering and enhanced reasoning.
 
 #### Schema
 
@@ -640,7 +639,7 @@ grok_search({
 
 ### grok_reason
 
-Deep logical reasoning with Grok-4 using first principles.
+Deep logical reasoning with Grok-4.1 using first principles and enhanced emotional intelligence.
 
 #### Schema
 
@@ -686,7 +685,7 @@ grok_reason({
 
 ### grok_code
 
-Code analysis and optimization with Grok-4.
+Code analysis and optimization with Grok-4.1 Fast (tool-calling optimized).
 
 #### Schema
 
@@ -738,7 +737,7 @@ grok_code({
 
 ### grok_debug
 
-Deep debugging assistance with Grok-4.
+Deep debugging assistance with Grok-4.1 Fast.
 
 #### Schema
 
@@ -772,7 +771,7 @@ grok_debug({
 
 ### grok_architect
 
-System architecture and design with Grok-4.
+System architecture and design with Grok-4.1.
 
 #### Schema
 
@@ -800,7 +799,7 @@ grok_architect({
 
 ### grok_brainstorm
 
-Creative brainstorming using Grok-4-heavy.
+Creative brainstorming using Grok-4.1 with enhanced creativity and emotional intelligence.
 
 #### Schema
 
@@ -928,41 +927,6 @@ openai_brainstorm({
 
 ---
 
-### openai_compare
-
-Multi-option consensus analysis with GPT-5.
-
-#### Schema
-
-```typescript
-{
-  topic: string;                    // REQUIRED
-  options: string[];                // REQUIRED - Options to compare
-  criteria?: string[];              // Evaluation criteria
-  includeRecommendation?: boolean;  // Default: true
-}
-```
-
-#### Example Calls
-
-**Compare frameworks:**
-```typescript
-openai_compare({
-  topic: "JavaScript framework selection",
-  options: ["React", "Vue", "Svelte", "Angular"],
-  criteria: [
-    "Learning curve",
-    "Performance",
-    "Community support",
-    "Ecosystem maturity",
-    "Job market demand"
-  ],
-  includeRecommendation: true
-})
-```
-
----
-
 ## Gemini Suite
 
 ### gemini_brainstorm

package/package.json

@@ -2,7 +2,7 @@
   "name": "tachibot-mcp",
   "mcpName": "io.github.byPawel/tachibot-mcp",
   "displayName": "TachiBot MCP - Universal AI Orchestrator",
-  "version": "2.0.5",
+  "version": "2.0.7",
   "type": "module",
   "main": "dist/src/server.js",
   "bin": {

package/profiles/balanced.json

@@ -13,8 +13,7 @@
     "grok_architect": false,
     "grok_brainstorm": false,
     "grok_search": true,
-    "openai_gpt5_reason": true,
-    "openai_compare": false,
+    "openai_reason": true,
     "openai_brainstorm": true,
     "openai_code_review": false,
     "openai_explain": false,

package/profiles/code_focus.json

@@ -13,8 +13,7 @@
     "grok_architect": false,
     "grok_brainstorm": false,
     "grok_search": false,
-    "openai_gpt5_reason": false,
-    "openai_compare": false,
+    "openai_reason": false,
     "openai_brainstorm": false,
     "openai_code_review": true,
     "openai_explain": false,

package/profiles/debug_intensive.json

@@ -22,7 +22,6 @@
     "grok_architect": false,
     "grok_brainstorm": false,
     "grok_search": false,
-    "openai_compare": false,
     "openai_brainstorm": false,
     "gemini_brainstorm": false,
     "gemini_analyze_text": false,

package/profiles/full.json

@@ -1,5 +1,5 @@
 {
-  "description": "All tools enabled for maximum capability (~Xk tokens, 32 tools)",
+  "description": "All tools enabled for maximum capability (~Xk tokens, 31 tools)",
   "tools": {
     "think": true,
     "focus": true,
@@ -13,8 +13,7 @@
     "grok_architect": true,
     "grok_brainstorm": true,
     "grok_search": true,
-    "openai_gpt5_reason": true,
-    "openai_compare": true,
+    "openai_reason": true,
     "openai_brainstorm": true,
     "openai_code_review": true,
     "openai_explain": true,

package/profiles/minimal.json

@@ -13,8 +13,7 @@
     "grok_architect": false,
     "grok_brainstorm": false,
     "grok_search": false,
-    "openai_gpt5_reason": false,
-    "openai_compare": false,
+    "openai_reason": false,
     "openai_brainstorm": false,
     "openai_code_review": false,
     "openai_explain": false,

package/profiles/research_power.json

@@ -13,8 +13,7 @@
     "grok_architect": false,
     "grok_brainstorm": false,
     "grok_search": true,
-    "openai_gpt5_reason": false,
-    "openai_compare": false,
+    "openai_reason": false,
     "openai_brainstorm": true,
     "openai_code_review": false,
     "openai_explain": false,

package/profiles/workflow_builder.json

@@ -13,8 +13,7 @@
     "grok_architect": false,
     "grok_brainstorm": false,
     "grok_search": false,
-    "openai_gpt5_reason": false,
-    "openai_compare": false,
+    "openai_reason": false,
     "openai_brainstorm": false,
     "openai_code_review": false,
     "openai_explain": false,

package/smithery.yaml

@@ -1,7 +1,7 @@
 # Smithery configuration file: https://smithery.ai/docs/config#smitheryyaml
 name: "tachibot-mcp"
-description: "Multi-model AI orchestration platform with 26 tools across Perplexity, Grok, OpenAI GPT-5, Gemini, and Qwen. Features YAML workflows, 5 token-optimized profiles (4k-19k), smart routing, and cost controls. Build complex AI pipelines with variable passing and parallel execution."
-version: "2.0.0"
+description: "Multi-model AI orchestration platform with 31 tools across Perplexity, Grok, OpenAI GPT-5, Gemini, and Qwen. Features YAML workflows, 5 token-optimized profiles (4k-20k), smart routing, and cost controls. Build complex AI pipelines with variable passing and parallel execution."
+version: "2.0.6"
 
 startCommand:
   type: stdio

package/tools.config.json

@@ -25,7 +25,7 @@
       "grok_brainstorm",
       "grok_search"
     ],
-    "openai": ["openai_compare", "openai_brainstorm"],
+    "openai": ["openai_brainstorm", "openai_reason", "openai_code_review", "openai_explain"],
     "gemini": [
       "gemini_brainstorm",
       "gemini_analyze_code",
@@ -38,7 +38,12 @@
       "workflow",
       "list_workflows",
       "create_workflow",
-      "visualize_workflow"
+      "visualize_workflow",
+      "workflow_start",
+      "continue_workflow",
+      "workflow_status",
+      "validate_workflow",
+      "validate_workflow_file"
     ],
     "collaborative": ["pingpong", "qwen_competitive"]
   },
@@ -59,8 +64,10 @@
       "grok_architect": true,
       "grok_brainstorm": true,
       "grok_search": true,
-      "openai_compare": true,
       "openai_brainstorm": true,
+      "openai_reason": true,
+      "openai_code_review": true,
+      "openai_explain": true,
       "gemini_brainstorm": true,
       "gemini_analyze_code": true,
       "gemini_analyze_text": true,
@@ -74,6 +81,11 @@
       "list_workflows": true,
       "create_workflow": true,
       "visualize_workflow": true,
+      "workflow_start": true,
+      "continue_workflow": true,
+      "workflow_status": true,
+      "validate_workflow": true,
+      "validate_workflow_file": true,
       "pingpong": true,
       "qwen_competitive": false
     }

package/workflows/code-architecture-review.yaml

@@ -176,10 +176,12 @@ steps:
   # ═══════════════════════════════════════════════════════════════════════════
 
   - name: consensus
-    tool: openai_compare
+    tool: openai_brainstorm
     input:
-      topic: "Final architecture recommendations for: ${query}"
-      options:
+      problem: |
+        Synthesize final architecture recommendations for: ${query}
+
+        Combine these expert analyses into actionable recommendations:
         - "Grok's SOLID analysis: ${grok-solid-analysis.output}"
         - "Gemini's pattern analysis: ${gemini-pattern-analysis.output}"
         - "Qwen's CQRS evaluation: ${qwen-cqrs-evaluation.output}"

package/workflows/creative-brainstorm-yaml.yaml

@@ -15,7 +15,7 @@ variables:
 steps:
   # Step 1: Claude Thinking - Problem Framing
   - name: claude-thinking
-    tool: openai_gpt5_reason  # Using GPT-5 Mini for structured thinking
+    tool: openai_reason  # Using GPT-5 Mini for structured thinking
     input:
       query: |
         Analyze and structure the brainstorming request: ${query}

package/workflows/pingpong.yaml

@@ -121,10 +121,12 @@ steps:
   # ═══════════════════════════════════════════════════════════════════════════
 
   - name: consensus
-    tool: openai_compare
+    tool: openai_brainstorm
     input:
-      topic: "Final consensus on: ${query}"
-      options:
+      problem: |
+        Synthesize final consensus on: ${query}
+
+        Analyze these perspectives and provide a unified recommendation:
         - "Grok's analysis: ${challenge-grok.output}"
         - "Gemini's synthesis: ${challenge-gemini.output}"
         - "Qwen's technical review: ${challenge-qwen.output}"

package/workflows/system/README.md

@@ -25,7 +25,7 @@ workflow verifier --query "Python is faster than JavaScript for all use cases"
 ```
 
 **Models Used:**
-1. GPT-5 Mini (openai_compare)
+1. GPT-5 Mini (openai_brainstorm)
 2. Gemini 2.5 (gemini_analyze_text)
 3. Grok 4 (grok_code)
 4. Qwen Coder (qwen_coder)

package/workflows/system/verifier.yaml

@@ -10,12 +10,15 @@ steps:
   # Each model gets 10k tokens, saved to disk
 
   - name: verify_gpt5_mini
-    tool: openai_compare
+    tool: openai_brainstorm
     input:
-      topic: "Verify the following claim or statement"
-      options:
-        - "${query}"
-      includeRecommendation: false
+      problem: |
+        Verify the following claim or statement:
+        ${query}
+
+        Provide verification analysis including accuracy assessment, evidence, and confidence level.
+      style: "systematic"
+      quantity: 3
     saveToFile: true
     maxTokens: 10000
     output:

package/workflows/ultra-creative-brainstorm.yaml

@@ -8,10 +8,10 @@ version: "3.0"
 settings:
   optimization:
     enabled: true
-    smartRouting: true
+    smartRouting: false  # Disabled - was selecting invalid gpt-5-nano model
     compressPrompts: false
   autoSynthesis:
-    enabled: true
+    enabled: false  # Disabled for testing
     tokenThreshold: 25000
     checkpointInterval: 12000
     synthesisTool: 'gemini_analyze_text'
@@ -195,7 +195,7 @@ steps:
     input:
       query: "${query} - explore creative applications across domains"
       mode: "code-brainstorm"
-      models: ["gpt-5.1", "gemini-2.5", "grok-4-fast-reasoning"]
+      models: ["gpt-5.1", "gemini-2.5-flash", "grok-4-1-fast-reasoning"]
       rounds: 3
       context: "Research: ${research_findings}\nIdeas: ${innovative_solutions}\nPatterns: ${patterns}"
     saveToFile: true