tachibot-mcp 2.23.1 → 2.24.0

package/CHANGELOG.md

@@ -5,6 +5,19 @@ All notable changes to TachiBot MCP will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [2.23.3] - 2026-06-17
+
+### Changed
+- **GLM-5.1 → GLM-5.2** (`z-ai/glm-5.2`, released Jun 13 on OpenRouter, verified via grok_search + web): Zhipu's open-weights flagship for long-horizon tasks — usable 1M-token context (up from agentic-tier 5.1), two thinking-effort levels (High/Max). SWE-Bench Pro 62.1 (up from 58.4), Terminal-Bench 2.1 81.0 (up from 62.0); beats GPT-5.5 on several long-horizon coding benchmarks at ~1/6 the cost. List price $1.40/$4.40 per M (262K max output). Powers `glm_reason` and the `glm` juror. GLM-5.1 kept as quota fallback via `MODEL_FALLBACKS` (GLM-5.2 → GLM-5.1 → GLM-5); the `glm` substring already routes 5.2 to the 600s reasoning timeout.
+
+## [2.23.2] - 2026-06-15
+
+### Changed
+- **Kimi K2.6 → K2.7-Code** (`moonshotai/kimi-k2.7-code`, released Jun 12 on OpenRouter, verified via grok_search): coding-specialized variant built on K2.6 — +21.8% on Kimi Code Bench v2 with lower token use, 262K context, native multimodal, always-thinking mode. List price $0.75/$3.50 per M. Now powers all Kimi tools — `kimi_thinking`, `kimi_code`, `kimi_decompose`, `kimi_long_context` (`openrouter-tools.ts`) and the `kimi` juror (`jury-tool.ts`). K2.6 kept as quota fallback via `MODEL_FALLBACKS` (K2.7-Code → K2.6 → K2-Thinking); the `kimi` substring already routes K2.7-Code to the 600s reasoning timeout (regression test added).
+
+### Notes
+- No general-purpose Kimi K2.7 exists yet — K2.7-Code is the only post-K2.6 release. The coding-tuned model now also serves the reasoning/long-context tools and the kimi juror.
+
 ## [2.23.1] - 2026-06-11
 
 ### Changed

package/README.md

@@ -4,18 +4,18 @@
 
 ### Multi-Model AI Orchestration Platform
 
-[![Version](https://img.shields.io/badge/version-2.15.2-blue.svg)](https://www.npmjs.com/package/tachibot-mcp)
-[![Tools](https://img.shields.io/badge/tools-57_active-brightgreen.svg)](#-tool-ecosystem-57-tools)
+[![Version](https://img.shields.io/badge/version-2.24.0-blue.svg)](https://www.npmjs.com/package/tachibot-mcp)
+[![Tools](https://img.shields.io/badge/tools-61_active-brightgreen.svg)](#-tool-ecosystem-61-tools)
 [![License](https://img.shields.io/badge/license-AGPL--3.0-green.svg)](LICENSE)
-[![Node](https://img.shields.io/badge/node-%3E%3D18.0.0-brightgreen.svg)](https://nodejs.org)
+[![Node](https://img.shields.io/badge/node-%3E%3D22.0.0-brightgreen.svg)](https://nodejs.org)
 [![MCP](https://img.shields.io/badge/MCP-Compatible-purple.svg)](https://modelcontextprotocol.io)
 
-**57 AI tools. 12 providers. One protocol.**
+**61 AI tools. 12 providers. One protocol.**
 
-Orchestrate Perplexity, Grok, GPT-5.5, Gemini, Qwen, Kimi K2.6, and MiniMax M3
+Orchestrate Perplexity, Grok, GPT-5.5, Gemini, Qwen, Kimi K2.7-Code, and MiniMax M3
 from Claude Code, Claude Desktop, Cursor, or any MCP client.
 
-[Get Started](#-quick-start) · [View Tools](#-tool-ecosystem-57-tools) · [Documentation](https://tachibot.com/docs)
+[Get Started](#-quick-start) · [View Tools](#-tool-ecosystem-61-tools) · [Documentation](https://tachibot.com/docs)
 
 <br>
 
@@ -70,7 +70,7 @@ Techniques are embedded directly in tool system prompts for automatic applicatio
 
 ## Skills (Claude Code)
 
-TachiBot ships with 12 slash commands for Claude Code. These orchestrate the tools into powerful workflows:
+TachiBot ships with 14 slash commands for Claude Code. These orchestrate the tools into powerful workflows:
 
 | Skill | What it does | Example |
 |-------|-------------|---------|
@@ -85,6 +85,8 @@ TachiBot ships with 12 slash commands for Claude Code. These orchestrate the too
 | `/lens` | Long-context analysis over Kimi's 256K window | `/lens find inconsistencies in this spec` |
 | `/reflect` | Grounded reflexion loop — critique vs external evidence | `/reflect harden this auth middleware` |
 | `/tot` | Tree-of-Thought: branch → jury-prune → synthesize | `/tot design a rate limiter` |
+| `/review` | Multi-model diff review — panel + Gemini judge verdict | `/review` (or paste a diff) |
+| `/redteam` | Adversarial plan red-team — pre-mortem, risks, plan edits | `/redteam <paste plan>` |
 | `/tachi` | Help - see available skills, tools, key status | `/tachi` |
 
 Skills automatically adapt to your configured API keys. Even with just 1-2 providers, all skills work.
@@ -96,7 +98,7 @@ Skills automatically adapt to your configured API keys. Even with just 1-2 provi
 ## Key Features
 
 ### Multi-Model Intelligence
-- **57 AI Tools** across 12 providers &mdash; Perplexity, Grok, GPT-5, Gemini, Qwen, Kimi, MiniMax, DeepSeek, GLM (Zhipu), StepFun, ERNIE (Baidu), plus free local models (Ollama / LM Studio / llama.cpp / vLLM)
+- **61 AI Tools** across 12 providers &mdash; Perplexity, Grok, GPT-5, Gemini, Qwen, Kimi, MiniMax, DeepSeek, GLM (Zhipu), StepFun, ERNIE (Baidu), plus free local models (Ollama / LM Studio / llama.cpp / vLLM)
 - **Gemini 3.5 Flash** (`gemini-3.5-flash`, GA May 19 2026) &mdash; Flash/search tier; reasoning default stays `gemini-3.1-pro-preview`
 - **Multi-Model Council** &mdash; planner_maker synthesizes plans from 5+ models into bite-sized TDD steps
 - **Smart Routing** &mdash; Automatic model selection for optimal results
@@ -111,12 +113,12 @@ Skills automatically adapt to your configured API keys. Even with just 1-2 provi
 ### Tool Profiles
 | Profile | Tools | Best For |
 |---------|-------|----------|
-| **Minimal** | 12 | Quick tasks, low token budget |
+| **Minimal** | 13 | Quick tasks, low token budget |
 | **Research Power** | 35 | Deep investigation, multi-source |
-| **Code Focus** | 34 | Software development, SWE tasks |
-| **Balanced** | 45 | General-purpose, mixed workflows |
-| **Heavy Coding** (default) | 50 | Max code tools + agentic workflows |
-| **Full** | 57 | Everything enabled |
+| **Code Focus** | 39 | Software development, SWE tasks |
+| **Balanced** | 50 | General-purpose, mixed workflows |
+| **Heavy Coding** | 54 | Max code tools + agentic workflows |
+| **Full** (default) | 61 | Everything enabled |
 
 ### Developer Experience
 - **Claude Code** &mdash; First-class support
@@ -134,7 +136,15 @@ Skills automatically adapt to your configured API keys. Even with just 1-2 provi
 npm install -g tachibot-mcp
 ```
 
-### Setup
+### Claude Code (one-liner)
+
+```bash
+claude mcp add tachibot -- npx -y -p tachibot-mcp tachibot
+```
+
+Then verify with `/mcp`. Add API keys with `--env`, e.g. `--env OPENROUTER_API_KEY=sk-or-xxx --env PERPLEXITY_API_KEY=pplx-xxx`.
+
+### Setup (Claude Desktop)
 
 **Gateway Mode (Recommended)** &mdash; 2 keys, all providers:
 
@@ -178,22 +188,22 @@ See [Installation Guide](docs/INSTALLATION_BOTH.md) for detailed instructions.
 
 ---
 
-## Tool Ecosystem (57 Tools)
+## Tool Ecosystem (61 Tools)
 
-### Research & Search (6)
-`perplexity_ask` &#183; `perplexity_research` &#183; `perplexity_reason` &#183; `grok_search` &#183; `openai_search` &#183; `gemini_search`
+### Research & Search (5)
+`perplexity_ask` &#183; `perplexity_reason` &#183; `grok_search` &#183; `openai_search` &#183; `gemini_search`
 
 ### Reasoning & Planning (13)
 `grok_reason` &#183; `openai_reason` &#183; `qwen_reason` &#183; `qwq_reason` &#183; `kimi_thinking` &#183; `kimi_decompose` &#183; `deepseek_reason` &#183; `glm_reason` &#183; `stepfun_reason` &#183; `ernie_reason` &#183; `planner_maker` &#183; `planner_runner` &#183; `list_plans`
 
-### Code Intelligence (9)
-`kimi_code` &#183; `grok_code` &#183; `grok_debug` &#183; `qwen_coder` &#183; `qwen_algo` &#183; `qwen_competitive` &#183; `deepseek_algo` &#183; `minimax_code` &#183; `minimax_agent`
+### Code Intelligence (10)
+`kimi_code` &#183; `grok_code` &#183; `grok_debug` &#183; `qwen_coder` &#183; `qwen_algo` &#183; `qwen_competitive` &#183; `deepseek_algo` &#183; `minimax_code` &#183; `minimax_agent` &#183; `testgen`
 
-### Analysis & Judgment (11)
-`gemini_analyze_text` &#183; `gemini_analyze_code` &#183; `gemini_judge` &#183; `jury` &#183; `gemini_brainstorm` &#183; `openai_brainstorm` &#183; `openai_code_review` &#183; `openai_explain` &#183; `grok_brainstorm` &#183; `grok_architect` &#183; `kimi_long_context`
+### Analysis & Judgment (14)
+`gemini_analyze_text` &#183; `gemini_analyze_code` &#183; `gemini_judge` &#183; `jury` &#183; `diff_review` &#183; `plan_critique` &#183; `gemini_brainstorm` &#183; `openai_brainstorm` &#183; `openai_code_review` &#183; `openai_explain` &#183; `grok_brainstorm` &#183; `grok_architect` &#183; `security_review` &#183; `kimi_long_context`
 
-### Meta & Orchestration (5)
-`think` &#183; `nextThought` &#183; `focus` &#183; `tachi` &#183; `usage_stats`
+### Meta & Orchestration (6)
+`think` &#183; `nextThought` &#183; `focus` &#183; `tachi` &#183; `doctor` &#183; `usage_stats`
 
 ### Workflows (9)
 `workflow` &#183; `workflow_start` &#183; `continue_workflow` &#183; `list_workflows` &#183; `create_workflow` &#183; `visualize_workflow` &#183; `workflow_status` &#183; `validate_workflow` &#183; `validate_workflow_file`

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

@@ -87,12 +87,14 @@ export const GROK_MODELS = {
     _3: "grok-3", // Legacy with search: 256K→2M
 };
 // Kimi Models (Moonshot AI via OpenRouter)
-// K2.6 released Apr 20, 2026 - 1T MoE, leads SWE-bench Pro for long-horizon coding
-// K2.5 released Jan 27, 2026 - kept as fallback
+// K2.7-Code released Jun 12, 2026 - coding-specialized (built on K2.6), +21.8% on Kimi Code Bench v2, lower token use, 262K ctx, multimodal, always-thinking
+// K2.6 released Apr 20, 2026 - 1T MoE, leads SWE-bench Pro for long-horizon coding (fallback)
+// K2.5 released Jan 27, 2026 - kept as legacy
 export const KIMI_MODELS = {
     K2_THINKING: "moonshotai/kimi-k2-thinking", // 1T MoE, 32B active - agentic reasoning (256k context)
-    K2_6: "moonshotai/kimi-k2.6", // CURRENT: 1T MoE, SWE-Pro leader, ~$0.74/$4.65
-    K2_5: "moonshotai/kimi-k2.5", // Previous: multimodal + agent swarm (fallback)
+    K2_7_CODE: "moonshotai/kimi-k2.7-code", // CURRENT (Jun 12, 2026): coding-specialized, +21.8% Code Bench v2, 262K ctx, $0.75/$3.50
+    K2_6: "moonshotai/kimi-k2.6", // Previous (Apr 2026): 1T MoE, SWE-Pro leader (fallback)
+    K2_5: "moonshotai/kimi-k2.5", // Legacy: multimodal + agent swarm
 };
 // MiniMax Models (MiniMax via OpenRouter)
 // M3 released May 31, 2026 - 1M ctx, MSA sparse attention (~1/20 compute at 1M), native multimodal
@@ -186,7 +188,7 @@ export const CURRENT_MODELS = {
         reason: PERPLEXITY_MODELS.SONAR_REASONING, // sonar-reasoning-pro $2/$8 per M
     },
     openrouter: {
-        kimi: KIMI_MODELS.K2_6, // K2.6 (Apr 2026): 1T MoE, SWE-bench Pro leader
+        kimi: KIMI_MODELS.K2_7_CODE, // K2.7-Code (Jun 2026): coding-specialized, built on K2.6
         qwen: QWEN_MODELS.CODER_NEXT, // Qwen3-Coder-Next: 80B/3B MoE, 262K ctx, SWE >70% (no 3.6-coder yet)
         qwen_reason: QWEN_MODELS.MAX_THINKING, // 235B MoE thinking mode (HMMT 98%) — still best for reasoning
         minimax: MINIMAX_MODELS.M3, // M3: 1M ctx, MSA sparse attention, agentic/coding
@@ -304,7 +306,7 @@ export const TOOL_DEFAULTS = {
         temperature: 0.3, // Lower for precise reasoning
     },
     kimi_thinking: {
-        model: KIMI_MODELS.K2_6, // K2.6 (Apr 2026): 1T MoE, SWE-bench Pro leader
+        model: KIMI_MODELS.K2_7_CODE, // K2.7-Code (Jun 2026): coding-specialized, built on K2.6
         maxTokens: 16000,
         temperature: 0.7,
     },
@@ -370,6 +372,7 @@ export const MODEL_DISPLAY_NAMES = {
     "sonar-reasoning-pro": "perplexity-reason",
     // Kimi (Moonshot)
     "moonshotai/kimi-k2-thinking": "kimi-k2",
+    "moonshotai/kimi-k2.7-code": "kimi-k2.7-code",
     "moonshotai/kimi-k2.6": "kimi-k2.6",
     "moonshotai/kimi-k2.5": "kimi-k2.5",
     "moonshotai/kimi-k2.5-thinking": "kimi-k2.5",
@@ -424,6 +427,7 @@ export const MODEL_PRICING = {
     "sonar-reasoning-pro": 0.006, // avoid - expensive reasoning tokens
     // OpenRouter models - Kimi
     "moonshotai/kimi-k2-thinking": 0.002,
+    "moonshotai/kimi-k2.7-code": 0.002125, // ($0.75 + $3.50) / 2 / 1000 (Jun 12, 2026)
     "moonshotai/kimi-k2.6": 0.0027, // ($0.74 + $4.65) / 2 / 1000 (Apr 2026)
     "moonshotai/kimi-k2.5": 0.003,
     "moonshotai/kimi-k2.5-thinking": 0.003,

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

@@ -28,7 +28,7 @@ const MODELS = {
     // OpenRouter
     QWEN: QWEN_MODELS.CODER_NEXT, // qwen/qwen3-coder-next (80B/3B MoE, 262K ctx)
     QWEN_REASON: QWEN_MODELS.MAX_THINKING, // qwen/qwen3-235b-a22b-thinking (235B MoE)
-    KIMI: KIMI_MODELS.K2_6, // moonshotai/kimi-k2.6 (Apr 2026 - SWE-bench Pro leader)
+    KIMI: KIMI_MODELS.K2_7_CODE, // moonshotai/kimi-k2.7-code (Jun 2026 - coding-specialized, built on K2.6)
     MINIMAX: MINIMAX_MODELS.M3, // minimax/minimax-m3 (May 2026 - 1M ctx, MSA sparse attention)
 };
 /**

package/dist/src/config/timeout-config.js

@@ -36,8 +36,8 @@ export function getTimeoutConfig() {
  * Thinking/reasoning/swarm models get extended timeout (600s default).
  * Standard models get 180s default.
  *
- * NOTE: All Kimi K2 variants (incl. 'moonshotai/kimi-k2.6') run the Agent Swarm
- * and need the extended timeout — but their model IDs don't contain 'thinking',
+ * NOTE: All Kimi K2 variants (incl. 'moonshotai/kimi-k2.7-code', which is always in
+ * thinking mode) need the extended timeout — but their model IDs don't contain 'thinking',
  * so we match 'kimi' explicitly. Same applies to DeepSeek V4 and Zhipu GLM, whose
  * reasoning passes are slow but whose IDs lack the 'thinking'/'reasoning' marker.
  * This keeps primary and fallback consistent and lets these tools inherit 600s
@@ -77,14 +77,19 @@ export function formatTimeout(ms) {
     return `${minutes}m ${remainingSeconds}s`;
 }
 /**
- * Create timeout promise for use with Promise.race()
+ * Create timeout promise for use with Promise.race(), plus a cancel handle.
+ * The race's loser leaves its timer scheduled unless the caller cancels it —
+ * without that, a fast-resolving `promise` still leaves this setTimeout alive
+ * for the full `ms`, holding the event loop open.
  */
 export function createTimeoutPromise(ms, message) {
-    return new Promise((_, reject) => {
-        setTimeout(() => {
+    let timer;
+    const promise = new Promise((_, reject) => {
+        timer = setTimeout(() => {
             reject(new Error(message || `Operation timed out after ${formatTimeout(ms)}`));
         }, ms);
     });
+    return { promise, cancel: () => clearTimeout(timer) };
 }
 /**
  * Wrap promise with timeout
@@ -93,10 +98,13 @@ export async function withTimeout(promise, timeoutMs, toolName) {
     const timeoutMessage = toolName
         ? `${toolName} operation timed out after ${formatTimeout(timeoutMs)}`
         : `Operation timed out after ${formatTimeout(timeoutMs)}`;
-    return Promise.race([
-        promise,
-        createTimeoutPromise(timeoutMs, timeoutMessage)
-    ]);
+    const raceTimeout = createTimeoutPromise(timeoutMs, timeoutMessage);
+    try {
+        return await Promise.race([promise, raceTimeout.promise]);
+    }
+    finally {
+        raceTimeout.cancel();
+    }
 }
 /**
  * Check if we should show progress (operation exceeds threshold)
@@ -127,7 +135,9 @@ export const SMART_TIMEOUT_DEFAULTS = {
         },
         openai: {
             base: 60000, // 60 seconds - GPT-5.4 reasoning needs more time
-            max: 180000 // 3 minutes - high/xhigh reasoning effort
+            max: 600000 // 10 minutes - GPT-5.5 high/xhigh effort can exceed the
+            // OpenAI SDK's 900s default; 180s was cutting off real
+            // reasoning runs (verified 2026-07 deep-research)
         },
         anthropic: {
             base: 20000, // 20 seconds

package/dist/src/memory/index.js

@@ -9,7 +9,7 @@ export * from './memory-manager.js';
 // Provider exports
 export * from './providers/local-provider.js';
 export * from './providers/mem0-provider.js';
-// export * from './providers/dokoro-provider.js';
+export * from './providers/dokoro-provider.js';
 // export * from './providers/hybrid-provider.js';
 // Main API
 export { getMemoryManager, resetMemoryManager, HierarchicalMemoryManager } from './memory-manager.js';

package/dist/src/memory/memory-config.js

@@ -109,9 +109,8 @@ export function validateMemoryConfig(config) {
     if (config.provider === 'mem0' && !config.mem0?.apiKey && !process.env.MEM0_API_KEY) {
         errors.push('Mem0 provider requires API key (MEM0_API_KEY or config.mem0.apiKey)');
     }
-    if (config.provider === 'dokoro' && !config.dokoro?.connectionString && !process.env.DOKORO_CONNECTION) {
-        errors.push('Dokoro provider requires connection string');
-    }
+    // Dokoro is file-backed: connectionString is optional (falls back to
+    // DOKORO_PATH env or {cwd}/dokoro), so no validation needed here.
     if (config.provider === 'local' && !config.local?.path) {
         errors.push('Local provider requires storage path');
     }

package/dist/src/memory/memory-manager.js

@@ -6,7 +6,7 @@ import { mergeMemoryConfig, validateMemoryConfig } from './memory-config.js';
 import { LocalProvider } from './providers/local-provider.js';
 import { Mem0Provider } from './providers/mem0-provider.js';
 import { randomBytes } from 'crypto';
-// import { DokoroProvider } from './providers/dokoro-provider.js';
+import { DokoroProvider } from './providers/dokoro-provider.js';
 // import { HybridProvider } from './providers/hybrid-provider.js';
 /**
  * Main memory manager that coordinates all providers
@@ -245,9 +245,15 @@ export class HierarchicalMemoryManager {
                     return null;
                 }
             case 'dokoro':
-                // TODO: Implement Dokoro provider
-                console.warn('Dokoro provider not yet implemented');
-                return null;
+                try {
+                    const provider = new DokoroProvider(this.config.dokoro || {});
+                    await provider.initialize();
+                    return provider;
+                }
+                catch (error) {
+                    console.error('Failed to create Dokoro provider:', error);
+                    return null;
+                }
             case 'local':
                 return await this.createLocalProvider();
             case 'hybrid':