tachibot-mcp 2.0.6 → 2.1.0

package/.env.example

@@ -22,8 +22,11 @@ TACHIBOT_PROFILE=balanced
 # Perplexity API (for web search, research, reasoning)
 PERPLEXITY_API_KEY=
 
-# Grok API (for code analysis, debugging, architecture)
-GROK_API_KEY=
+# Grok/xAI API (for code analysis, debugging, architecture)
+# Get from: https://console.x.ai/
+XAI_API_KEY=
+# Legacy alias (both work):
+# GROK_API_KEY=
 
 # OpenAI API (for GPT-5, analysis, comparison)
 OPENAI_API_KEY=
@@ -34,9 +37,16 @@ OPENAI_API_KEY=
 # Google Gemini API (for brainstorming, code/text analysis)
 GOOGLE_API_KEY=
 
-# OpenRouter API (for Qwen3 Coder and other models)
+# OpenRouter API (for Qwen3 Coder, Kimi, and gateway mode)
 OPENROUTER_API_KEY=
 
+# ===================================
+# OpenRouter Gateway Mode (Optional)
+# ===================================
+# When enabled, routes OpenAI/Gemini/Grok through OpenRouter
+# Benefits: Single API key, unified billing, automatic fallbacks
+# Note: Kimi/Qwen always use OpenRouter, Perplexity always uses direct API
+USE_OPENROUTER_GATEWAY=false
 
 # Anthropic API (optional - not currently used in Claude Code mode)
 # ANTHROPIC_API_KEY=

package/README.md

@@ -1,11 +1,46 @@
 # TachiBot MCP - Universal AI Orchestrator
 
-[![Version](https://img.shields.io/badge/version-2.0.0-blue.svg)](https://github.com/byPawel/tachibot-mcp)
+[![Version](https://img.shields.io/badge/version-2.1.0-blue.svg)](https://github.com/byPawel/tachibot-mcp)
 [![License](https://img.shields.io/badge/license-Apache%202.0-green.svg)](LICENSE)
 [![Node](https://img.shields.io/badge/node-%3E%3D18.0.0-brightgreen.svg)](https://nodejs.org)
 [![MCP](https://img.shields.io/badge/MCP-Compatible-purple.svg)](https://modelcontextprotocol.io)
 
-Multi-model AI orchestration platform with 31 tools (32 with competitive mode), advanced workflows, and intelligent prompt engineering. Works with Claude Code, Claude Desktop, Cursor, and any MCP-compatible client.
+Multi-model AI orchestration platform with 31 tools, advanced workflows, and intelligent prompt engineering. Works with Claude Code, Claude Desktop, Cursor, and any MCP-compatible client.
+
+---
+
+## 🔑 Bring Your Own Keys
+
+**Two ways to connect — pick what works for you:**
+
+| Mode | Keys Needed | Best For |
+|------|-------------|----------|
+| **Gateway Mode** | 2 keys (OpenRouter + Perplexity) | Simplicity, unified billing |
+| **Direct Mode** | 5+ keys (one per provider) | Maximum control, no middleman |
+
+### Gateway Mode (Recommended)
+Route GPT-5, Gemini, and Grok through OpenRouter — same key you need for Qwen/Kimi anyway.
+
+```bash
+OPENROUTER_API_KEY=sk-or-xxx     # GPT-5, Gemini, Grok, Qwen, Kimi
+PERPLEXITY_API_KEY=pplx-xxx      # Required separately (not on OpenRouter)
+USE_OPENROUTER_GATEWAY=true
+```
+
+### Direct Mode
+Use each provider's native API directly.
+
+```bash
+OPENAI_API_KEY=sk-xxx            # GPT-5
+GOOGLE_API_KEY=xxx               # Gemini
+GROK_API_KEY=xxx                 # Grok
+OPENROUTER_API_KEY=sk-or-xxx     # Qwen, Kimi (always via OpenRouter)
+PERPLEXITY_API_KEY=pplx-xxx      # Perplexity
+```
+
+Both modes work. Gateway mode just means fewer accounts to manage.
+
+Get keys: [OpenRouter](https://openrouter.ai) | [Perplexity](https://perplexity.ai)
 
 ---
 
@@ -29,6 +64,7 @@ Multi-model AI orchestration platform with 31 tools (32 with competitive mode),
 - **31 AI Tools:** Perplexity, Grok, GPT-5, Gemini, Qwen, Kimi (32 with competitive mode)
 - **Multi-Model Reasoning:** Challenger, Verifier, Scout modes
 - **Smart Routing:** Automatic model selection for optimal results
+- **OpenRouter Gateway:** Optional single API key for all providers (OpenAI, Gemini, Grok)
 
 ### 🔄 Advanced Workflows
 - **YAML-Based Workflows:** Define complex multi-step AI processes
@@ -38,10 +74,10 @@ Multi-model AI orchestration platform with 31 tools (32 with competitive mode),
 
 ### 🎯 Tool Profiles
 - **Minimal** (8 tools) - Budget-friendly, token-constrained
-- **Research Power** (15 tools) - Default, best balance
-- **Code Focus** (13 tools) - Software development
-- **Balanced** (17 tools) - General-purpose
-- **Full** (31 tools, 32 with competitive) - Maximum capability
+- **Research Power** (13 tools) - Default, best balance
+- **Code Focus** (15 tools) - Software development
+- **Balanced** (18 tools) - General-purpose
+- **Full** (31 tools, includes competitive mode) - Maximum capability
 
 ### 🔧 Developer Experience
 - **Claude Code Native** - First-class support
@@ -66,37 +102,41 @@ tachibot --version
 
 ### Setup
 
-1. **Add API Keys** (at least one):
-   ```bash
-   cp .env.example .env
-   # Edit .env and add your API keys
-   ```
-
-2. **Configure Profile** (optional):
-   ```bash
-   # Edit .env
-   TACHIBOT_PROFILE=research_power  # or minimal, code_focus, balanced, full
-   ```
-
-3. **Add to Claude Code/Desktop**:
-   ```json
-   {
-     "mcpServers": {
-       "tachibot": {
-         "command": "tachibot",
-         "env": {
-           "PERPLEXITY_API_KEY": "your-key",
-           "GROK_API_KEY": "your-key",
-           "OPENAI_API_KEY": "your-key",
-           "GEMINI_API_KEY": "your-key",
-           "OPENROUTER_API_KEY": "your-key"
-         }
-       }
-     }
-   }
-   ```
-
-4. **Restart Claude** and you're ready! 🎉
+**Option A: Single Key Mode (Recommended)**
+```json
+{
+  "mcpServers": {
+    "tachibot": {
+      "command": "tachibot",
+      "env": {
+        "OPENROUTER_API_KEY": "sk-or-xxx",
+        "PERPLEXITY_API_KEY": "pplx-xxx",
+        "USE_OPENROUTER_GATEWAY": "true"
+      }
+    }
+  }
+}
+```
+
+**Option B: Direct Provider Keys**
+```json
+{
+  "mcpServers": {
+    "tachibot": {
+      "command": "tachibot",
+      "env": {
+        "PERPLEXITY_API_KEY": "your-key",
+        "GROK_API_KEY": "your-key",
+        "OPENAI_API_KEY": "your-key",
+        "GOOGLE_API_KEY": "your-key",
+        "OPENROUTER_API_KEY": "your-key"
+      }
+    }
+  }
+}
+```
+
+**Restart Claude** and you're ready!
 
 See [Installation Guide](docs/INSTALLATION_BOTH.md) for detailed instructions.
 
@@ -106,9 +146,9 @@ See [Installation Guide](docs/INSTALLATION_BOTH.md) for detailed instructions.
 
 ### Core Tools
 - 🔍 **Research:** `perplexity_ask`, `perplexity_research`, `grok_search`, `scout`
-- 🧠 **Reasoning:** `grok_reason`, `kimi_thinking`, `openai_brainstorm`, `focus`
-- 💡 **Analysis:** `gemini_brainstorm`, `gemini_analyze_text`, `qwen_coder`
-- ✅ **Validation:** `verifier`, `challenger`
+- 🧠 **Reasoning:** `grok_reason`, `openai_reason`, `perplexity_reason`, `kimi_thinking`, `focus`
+- 💡 **Analysis:** `gemini_brainstorm`, `gemini_analyze_text`, `openai_brainstorm`, `qwen_coder`
+- ✅ **Validation:** `verifier`*, `challenger`* (*multi-model workflows)
 - 💭 **Meta:** `think`, `nextThought`
 
 ### Advanced Modes
@@ -118,10 +158,14 @@ See [Installation Guide](docs/INSTALLATION_BOTH.md) for detailed instructions.
 - **Verifier** - Multi-model consensus verification
 
 ### Workflows
-- **Ultra Creative Brainstorm** - 15 steps, 10 techniques
-- **Iterative Problem Solver** - Research → Analyze → Solve
+- **Ultra Creative Brainstorm** - 15 steps, 10 prompt techniques
 - **Code Architecture Review** - Systematic code analysis
+- **Code Review** - PR and code review workflow
+- **UX Research Flow** - User experience research
 - **Accessibility Audit** - WCAG compliance checking
+- **Creative Brainstorm** - Quick ideation workflow
+- **PingPong** - Multi-model debate/collaboration
+- **System Workflows** - `verifier`, `challenger`, `scout`
 - **Custom Workflows** - Build your own in YAML
 
 ---
@@ -183,7 +227,7 @@ We welcome contributions! See [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines.
 
 ## 📄 License
 
-Apache License 2.0 - see [LICENSE](LICENSE) for details.
+AGPL-3.0 - see [LICENSE](LICENSE) for details.
 
 ---
 
@@ -196,6 +240,6 @@ Apache License 2.0 - see [LICENSE](LICENSE) for details.
 
 ---
 
-**Made with ❤️ by the TachiBot Team**
+**Made with ❤️ by [@byPawel](https://github.com/byPawel)**
 
 *Transform your AI workflow with intelligent multi-model orchestration.*

package/dist/src/config/model-constants.js

@@ -1,30 +1,37 @@
 /**
  * Centralized Model Names and Constants
+ * Named by PROVIDER (not model version) for consistency and future-proofing
  * Use these constants instead of hardcoded strings in workflows and tools
  */
-// OpenAI GPT-5.1 Models (November 2025)
-export const GPT51_MODELS = {
-    FULL: "gpt-5.1", // Full reasoning model ($1.25/$10 per 1M tokens)
-    CODEX_MINI: "gpt-5.1-codex-mini", // Coding optimized, cost-efficient ($0.25/$2 per 1M tokens) - DEFAULT
-    CODEX: "gpt-5.1-codex", // Advanced coding ($1.25/$10 per 1M tokens)
+// =============================================================================
+// OPENAI MODELS (provider-based naming)
+// =============================================================================
+// NOTE: Codex models use /v1/responses endpoint, non-codex use /v1/chat/completions
+export const OPENAI_MODELS = {
+    // General purpose (use /v1/chat/completions)
+    FULL: "gpt-5.1", // Flagship: reasoning/fallback ($10/$30, 2M context)
+    PRO: "gpt-5-pro", // Premium: complex orchestration ($20/$60, 4M context)
+    // Code specialized (use /v1/responses endpoint!)
+    CODEX_MINI: "gpt-5.1-codex-mini", // Workhorse: 70-80% of code tasks ($2/$6, 256K)
+    CODEX: "gpt-5.1-codex", // Power: complex code tasks ($15/$45, 1M context)
+    CODEX_MAX: "gpt-5.1-codex-max", // Frontier: deep analysis & multi-file refactoring
 };
-// GPT-5.1 Reasoning Effort Levels
-export const GPT51_REASONING = {
+// OpenAI Reasoning Effort Levels (for models that support it)
+export const OPENAI_REASONING = {
     NONE: "none", // No extra reasoning (fastest, cheapest)
     LOW: "low", // Light reasoning
     MEDIUM: "medium", // Balanced reasoning (default)
     HIGH: "high", // Maximum reasoning (slowest, most thorough)
 };
-// OpenAI GPT-4 Models (Legacy - mapped to GPT-5.1)
-export const GPT4_MODELS = {
-    O_MINI: "gpt-5.1-codex-mini", // Cost-efficient
-    O: "gpt-5.1", // Current best
-    _1_MINI: "gpt-4.1-mini", // Best value with 1M context
-};
-// Google Gemini Models (2025)
+// =============================================================================
+// GEMINI MODELS (Google)
+// =============================================================================
 export const GEMINI_MODELS = {
-    FLASH: "gemini-2.5-flash", // Latest fast model
-    PRO: "gemini-2.5-pro", // Most advanced reasoning
+    // Gemini 3 (November 2025 - Latest)
+    GEMINI_3_PRO: "gemini-3-pro-preview", // Latest: structured outputs & multimodal, 1M context
+    // Gemini 2.5 (Previous generation - still available)
+    FLASH: "gemini-2.5-flash", // Fast model
+    PRO: "gemini-2.5-pro", // Advanced reasoning
     FLASH_LITE: "gemini-2.5-flash-lite", // Cost-effective
 };
 // Perplexity Models
@@ -32,12 +39,12 @@ export const PERPLEXITY_MODELS = {
     SONAR_PRO: "sonar-pro", // Main search model
     SONAR_REASONING: "sonar-reasoning-pro", // Reasoning model
 };
-// Grok Models (xAI) - Updated 2025-11-21 with Grok 4.1
+// Grok Models (xAI) - Updated 2025-11-22 with correct API model names
 export const GROK_MODELS = {
     // Grok 4.1 models (Nov 2025) - LATEST & BEST
-    _4_1: "grok-4.1", // Latest: 2M context, $0.20/$0.50, enhanced reasoning & creativity
-    _4_1_FAST: "grok-4.1-fast", // Tool-calling optimized: 2M context, $0.20/$0.50, agentic workflows
-    // Previous fast models (2025) - Still good
+    _4_1_FAST_REASONING: "grok-4-1-fast-reasoning", // Latest: 2M context, $0.20/$0.50, enhanced reasoning
+    _4_1_FAST_NON_REASONING: "grok-4-1-fast-non-reasoning", // Tool-calling optimized: 2M context, $0.20/$0.50
+    // Grok 4 fast models (2025) - Still good
     CODE_FAST: "grok-code-fast-1", // Coding specialist: 256K→2M, $0.20/$1.50, 92 tok/sec
     _4_FAST_REASONING: "grok-4-fast-reasoning", // Cheap reasoning: 2M→4M, $0.20/$0.50
     _4_FAST: "grok-4-fast-non-reasoning", // Fast general: 2M→4M, $0.20/$0.50
@@ -45,18 +52,43 @@ export const GROK_MODELS = {
     _4_HEAVY: "grok-4-0709", // Multi-agent: 256K→2M, $3/$15 (expensive!)
     _3: "grok-3", // Legacy with search: 256K→2M
 };
-// Kimi Models (Moonshot AI via OpenRouter) - Added 2025-11-07
+// Kimi Models (Moonshot AI via OpenRouter)
 export const KIMI_MODELS = {
     K2_THINKING: "moonshotai/kimi-k2-thinking", // 1T MoE, 32B active - Leading open-source agentic reasoning (256k context)
 };
+// Qwen Models (Alibaba via OpenRouter)
+export const QWEN_MODELS = {
+    CODER_PLUS: "qwen/qwen3-coder-plus", // Code specialist (32K context)
+    CODER: "qwen/qwen3-coder", // Standard coder
+    QWQ_32B: "qwen/qwq-32b", // Deep reasoning
+};
+// =============================================================================
+// OPENROUTER MODELS (Unified - all models accessible via OpenRouter)
+// =============================================================================
+export const OPENROUTER_MODELS = {
+    // Qwen models
+    ...QWEN_MODELS,
+    // Kimi models
+    ...KIMI_MODELS,
+};
+// =============================================================================
+// PROVIDERS - All provider constants in one place
+// =============================================================================
+export const PROVIDERS = {
+    openai: OPENAI_MODELS,
+    google: GEMINI_MODELS,
+    xai: GROK_MODELS,
+    perplexity: PERPLEXITY_MODELS,
+    openrouter: OPENROUTER_MODELS,
+};
 // All models combined for validation
 export const ALL_MODELS = {
-    ...GPT51_MODELS,
-    ...GPT4_MODELS,
+    ...OPENAI_MODELS,
     ...GEMINI_MODELS,
     ...PERPLEXITY_MODELS,
     ...GROK_MODELS,
     ...KIMI_MODELS,
+    ...QWEN_MODELS,
 };
 // Common workflow settings
 export const DEFAULT_WORKFLOW_SETTINGS = {
@@ -65,154 +97,152 @@ export const DEFAULT_WORKFLOW_SETTINGS = {
     retries: 3,
     timeout: 30000, // 30 seconds
 };
-// Tool-specific defaults for ALL tools
+// ============================================================================
+// CURRENT_MODELS - SINGLE BUMP POINT FOR MODEL VERSIONS
+// ============================================================================
+// When new models release, update ONLY this section!
+// All tools automatically use the new models.
+// ============================================================================
+export const CURRENT_MODELS = {
+    openai: {
+        reason: OPENAI_MODELS.PRO, // Deep reasoning
+        brainstorm: OPENAI_MODELS.FULL, // Creative ideation
+        code: OPENAI_MODELS.CODEX_MINI, // Code tasks (cheap & fast)
+        explain: OPENAI_MODELS.CODEX_MINI, // Explanations
+    },
+    grok: {
+        reason: GROK_MODELS._4_1_FAST_REASONING,
+        code: GROK_MODELS._4_1_FAST_NON_REASONING,
+        debug: GROK_MODELS._4_1_FAST_NON_REASONING,
+        brainstorm: GROK_MODELS._4_1_FAST_REASONING,
+        search: GROK_MODELS._4_1_FAST_REASONING,
+        architect: GROK_MODELS._4_1_FAST_REASONING,
+    },
+    gemini: {
+        default: GEMINI_MODELS.GEMINI_3_PRO,
+    },
+    perplexity: {
+        search: PERPLEXITY_MODELS.SONAR_PRO,
+        reason: PERPLEXITY_MODELS.SONAR_REASONING,
+    },
+    openrouter: {
+        kimi: KIMI_MODELS.K2_THINKING,
+        qwen: QWEN_MODELS.CODER_PLUS,
+    }
+};
+// Tool-specific defaults - References CURRENT_MODELS for easy bumping
 export const TOOL_DEFAULTS = {
-    // OpenAI GPT-5.1 tools
-    openai_gpt5_reason: {
-        model: GPT51_MODELS.FULL,
-        reasoning_effort: GPT51_REASONING.HIGH,
+    // OpenAI tools
+    openai_reason: {
+        model: CURRENT_MODELS.openai.reason,
+        reasoning_effort: OPENAI_REASONING.HIGH,
         maxTokens: 4000,
         temperature: 0.7,
     },
     openai_brainstorm: {
-        model: GPT51_MODELS.CODEX_MINI,
-        reasoning_effort: GPT51_REASONING.MEDIUM,
+        model: CURRENT_MODELS.openai.brainstorm,
+        reasoning_effort: OPENAI_REASONING.MEDIUM,
         maxTokens: 2000,
         temperature: 0.9,
     },
-    openai_compare: {
-        model: GPT51_MODELS.CODEX_MINI,
-        reasoning_effort: GPT51_REASONING.LOW,
-        maxTokens: 2000,
-        temperature: 0.7,
-    },
     openai_code_review: {
-        model: GPT51_MODELS.CODEX_MINI,
-        reasoning_effort: GPT51_REASONING.MEDIUM,
+        model: CURRENT_MODELS.openai.code,
+        reasoning_effort: OPENAI_REASONING.MEDIUM,
         maxTokens: 2000,
         temperature: 0.3,
     },
     openai_explain: {
-        model: GPT51_MODELS.CODEX_MINI,
-        reasoning_effort: GPT51_REASONING.LOW,
+        model: CURRENT_MODELS.openai.explain,
+        reasoning_effort: OPENAI_REASONING.LOW,
         maxTokens: 1500,
         temperature: 0.7,
     },
     // Gemini tools
-    gemini_query: {
-        model: GEMINI_MODELS.PRO,
-        maxTokens: 2048,
-        temperature: 0.7,
-    },
     gemini_brainstorm: {
-        model: GEMINI_MODELS.PRO,
+        model: CURRENT_MODELS.gemini.default,
         maxTokens: 2048,
         temperature: 0.9,
     },
     gemini_analyze_code: {
-        model: GEMINI_MODELS.PRO,
+        model: CURRENT_MODELS.gemini.default,
         maxTokens: 2048,
         temperature: 0.3,
     },
     gemini_analyze_text: {
-        model: GEMINI_MODELS.PRO,
+        model: CURRENT_MODELS.gemini.default,
         maxTokens: 2048,
         temperature: 0.5,
     },
     // Perplexity tools
     perplexity_ask: {
-        model: PERPLEXITY_MODELS.SONAR_PRO,
+        model: CURRENT_MODELS.perplexity.search,
         maxTokens: 2000,
         temperature: 0.7,
     },
     perplexity_reason: {
-        model: PERPLEXITY_MODELS.SONAR_REASONING,
+        model: CURRENT_MODELS.perplexity.reason,
         maxTokens: 4000,
         temperature: 0.7,
     },
     perplexity_research: {
-        model: PERPLEXITY_MODELS.SONAR_PRO,
+        model: CURRENT_MODELS.perplexity.search,
         maxTokens: 3000,
         temperature: 0.7,
     },
-    // Grok tools - UPDATED 2025-11-21 with Grok 4.1
-    grok: {
-        model: GROK_MODELS._4_1, // Latest: Enhanced reasoning & creativity
-        maxTokens: 4000,
-        temperature: 0.7,
-    },
+    // Grok tools
     grok_reason: {
-        model: GROK_MODELS._4_1, // Latest: Enhanced reasoning with lower hallucination
+        model: CURRENT_MODELS.grok.reason,
         maxTokens: 8000,
         temperature: 0.7,
     },
     grok_code: {
-        model: GROK_MODELS._4_1_FAST, // Tool-calling optimized, better than code-fast-1
+        model: CURRENT_MODELS.grok.code,
         maxTokens: 4000,
         temperature: 0.3,
     },
     grok_search: {
-        model: GROK_MODELS._4_1, // Latest with enhanced reasoning
+        model: CURRENT_MODELS.grok.search,
         maxTokens: 3000,
         temperature: 0.7,
     },
     grok_brainstorm: {
-        model: GROK_MODELS._4_1, // Latest: Enhanced creativity & emotional intelligence
+        model: CURRENT_MODELS.grok.brainstorm,
         maxTokens: 4000,
         temperature: 0.9,
     },
     grok_architect: {
-        model: GROK_MODELS._4_1, // Latest: Enhanced architecture reasoning
+        model: CURRENT_MODELS.grok.architect,
         maxTokens: 4000,
         temperature: 0.6,
     },
     grok_debug: {
-        model: GROK_MODELS._4_1_FAST, // Tool-calling optimized for debugging
+        model: CURRENT_MODELS.grok.debug,
         maxTokens: 3000,
         temperature: 0.3,
     },
-    // Qwen tools (via OpenRouter)
+    // OpenRouter tools
     qwen_coder: {
         maxTokens: 4000,
         temperature: 0.5,
     },
-    // Kimi tools (via OpenRouter)
     kimi_thinking: {
-        model: KIMI_MODELS.K2_THINKING,
-        maxTokens: 16000, // Large for detailed reasoning chains
-        temperature: 0.7, // Higher for creative reasoning
+        model: CURRENT_MODELS.openrouter.kimi,
+        maxTokens: 16000,
+        temperature: 0.7,
     },
-    // Meta tools (think, focus, code_reviewer, etc.)
+    // Meta tools
     think: {
-        model: GPT51_MODELS.FULL,
-        reasoning_effort: GPT51_REASONING.HIGH,
+        model: CURRENT_MODELS.openai.reason,
+        reasoning_effort: OPENAI_REASONING.HIGH,
         maxTokens: 500,
         temperature: 0.7,
     },
     focus: {
-        model: GPT51_MODELS.CODEX_MINI,
-        reasoning_effort: GPT51_REASONING.LOW,
+        model: CURRENT_MODELS.openai.code,
+        reasoning_effort: OPENAI_REASONING.LOW,
         maxTokens: 2000,
         temperature: 0.8,
     },
-    code_reviewer: {
-        model: GPT51_MODELS.CODEX_MINI,
-        reasoning_effort: GPT51_REASONING.MEDIUM,
-        maxTokens: 2000,
-        temperature: 0.5,
-    },
-    test_architect: {
-        model: GPT51_MODELS.CODEX_MINI,
-        reasoning_effort: GPT51_REASONING.MEDIUM,
-        maxTokens: 2000,
-        temperature: 0.6,
-    },
-    documentation_writer: {
-        model: GPT51_MODELS.CODEX_MINI,
-        reasoning_effort: GPT51_REASONING.LOW,
-        maxTokens: 2000,
-        temperature: 0.7,
-    },
 };
 // Default tool to use in workflows if not specified
 export const DEFAULT_WORKFLOW_TOOL = "openai_brainstorm";

package/dist/src/config/model-defaults.js

@@ -5,53 +5,67 @@
  * with smart defaults that balance cost and quality.
  *
  * Environment variables allow users to override defaults via Claude Desktop config.
+ *
+ * DRY: Imports model names from model-constants.ts - update there to bump versions.
+ */
+import { GEMINI_MODELS, OPENAI_MODELS, GROK_MODELS, PERPLEXITY_MODELS, KIMI_MODELS, QWEN_MODELS } from './model-constants.js';
+/**
+ * Default model selections by provider
+ * Strategy: Always use latest models (quality over cost)
+ * Update model-constants.ts to bump versions
  */
+const MODELS = {
+    // Google Gemini
+    GEMINI: GEMINI_MODELS.GEMINI_3_PRO, // gemini-3-pro-preview
+    // OpenAI
+    OPENAI: OPENAI_MODELS.CODEX_MINI, // gpt-5.1-codex-mini (default)
+    OPENAI_REASON: OPENAI_MODELS.FULL, // gpt-5.1 (deep reasoning)
+    // xAI Grok
+    GROK: GROK_MODELS._4_1_FAST_REASONING, // grok-4-1-fast-reasoning
+    // Perplexity
+    PERPLEXITY: PERPLEXITY_MODELS.SONAR_PRO, // sonar-pro
+    PERPLEXITY_REASON: PERPLEXITY_MODELS.SONAR_REASONING, // sonar-reasoning-pro
+    // OpenRouter
+    QWEN: QWEN_MODELS.CODER_PLUS, // qwen/qwen3-coder-plus
+    KIMI: KIMI_MODELS.K2_THINKING, // moonshotai/kimi-k2-thinking
+};
 /**
  * Get Scout model configuration
- *
- * Defaults:
- * - quick_scout: Flash + gpt-5.1-codex-mini (speed + cost efficient)
- * - research_scout: Pro + gpt-5.1-codex-mini (quality + cost balance)
+ * All variants use Gemini 3 Pro (latest & best)
  */
 export function getScoutModels() {
     const quick = process.env.SCOUT_QUICK_MODELS?.split(',').map(m => m.trim()) ||
-        ['qwen/qwen3-coder-plus', 'gemini-2.5-flash', 'gpt-5.1-codex-mini'];
+        [MODELS.QWEN, MODELS.GEMINI, MODELS.OPENAI];
     const research = process.env.SCOUT_RESEARCH_MODELS?.split(',').map(m => m.trim()) ||
-        ['qwen/qwen3-coder-plus', 'gemini-2.5-pro', 'gpt-5.1-codex-mini'];
+        [MODELS.QWEN, MODELS.GEMINI, MODELS.OPENAI];
     return { quick, research };
 }
 /**
  * Get Challenger model configuration
- *
- * Defaults: Pro + gpt-5.1-codex-mini (quality for critical analysis, cost efficient)
+ * Uses Gemini 3 Pro for critical analysis
  */
 export function getChallengerModels() {
     return process.env.CHALLENGER_MODELS?.split(',').map(m => m.trim()) ||
-        ['qwen/qwen3-coder-plus', 'gemini-2.5-pro', 'gpt-5.1-codex-mini'];
+        [MODELS.QWEN, MODELS.GEMINI, MODELS.OPENAI];
 }
 /**
  * Get Verifier model configuration
- *
- * Defaults:
- * - quick_verify: Flash + gpt-5.1-codex-mini (fast checks, cost efficient)
- * - standard modes: Pro + gpt-5.1-codex-mini (quality + cost balance)
- * - deep_verify: Pro + gpt-5.1 (maximum quality for critical verification)
+ * All variants use Gemini 3 Pro; deep uses gpt-5.1 for max reasoning
  */
 export function getVerifierModels() {
     const quick = process.env.VERIFIER_QUICK_MODELS?.split(',').map(m => m.trim()) ||
-        ['qwen/qwen3-coder-plus', 'gemini-2.5-flash', 'gpt-5.1-codex-mini'];
+        [MODELS.QWEN, MODELS.GEMINI, MODELS.OPENAI];
     const deep = process.env.VERIFIER_DEEP_MODELS?.split(',').map(m => m.trim()) ||
-        ['qwen/qwen3-coder-plus', 'gemini-2.5-pro', 'gpt-5.1'];
+        [MODELS.QWEN, MODELS.GEMINI, MODELS.OPENAI_REASON];
     const standard = process.env.VERIFIER_STANDARD_MODELS?.split(',').map(m => m.trim()) ||
-        ['qwen/qwen3-coder-plus', 'gemini-2.5-pro', 'gpt-5.1-codex-mini'];
+        [MODELS.QWEN, MODELS.GEMINI, MODELS.OPENAI];
     return { quick, deep, standard };
 }
 /**
- * Get default fallback models (used when variant has no specific config)
- *
- * Default: Pro + gpt-5.1-codex-mini (balanced quality and cost)
+ * Get default fallback models
+ * Uses Gemini 3 Pro + codex-mini
  */
 export function getDefaultModels() {
     return process.env.DEFAULT_MODELS?.split(',').map(m => m.trim()) ||
-        ['qwen/qwen3-coder-plus', 'gemini-2.5-pro', 'gpt-5.1-codex-mini'];
+        [MODELS.QWEN, MODELS.GEMINI, MODELS.OPENAI];
 }