polydev-ai 1.8.59 → 1.8.62

package/README.md

@@ -1,219 +1,180 @@
 # Polydev AI
 
-**Advanced Model Context Protocol Platform with Multi-LLM Integrations**
+**Multi-model AI perspectives for your coding agents.**
 
-<!-- Last updated: 2024-12-12 -->
+Get insights from GPT 5.2, Claude Opus 4.5, Gemini 3, and Grok 4.1 — all through one MCP server.
 
-[polydev.ai](https://polydev.ai) | Live Platform
+[![npm version](https://img.shields.io/npm/v/polydev-ai.svg)](https://www.npmjs.com/package/polydev-ai)
+[![SWE-bench Verified](https://img.shields.io/badge/SWE--bench-74.6%25-brightgreen)](https://polydev.ai/articles/swe-bench-paper)
+[![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](https://opensource.org/licenses/MIT)
 
 ---
 
-## Overview
+## Why Polydev?
 
-Polydev AI is an advanced Model Context Protocol (MCP) platform providing comprehensive multi-LLM integrations, subscription-based CLI access, OAuth bridges, and advanced tooling for AI development.
+**Stop copy-pasting between ChatGPT, Claude, and Gemini.** Get all their perspectives in your IDE with one request.
 
-## Features
+| Metric | Result |
+|--------|--------|
+| **SWE-bench Verified** | 74.6% Resolve@2 |
+| **Cost vs Claude Opus** | 62% lower |
+| **Response time** | 10-40 seconds |
 
-### 🤖 Comprehensive LLM Integration
+> *"Different models have different blind spots. Combining their perspectives eliminates yours."*
 
-- **API-Based Providers**: Direct integration with 8+ providers (Anthropic, OpenAI, Google, etc.)
-- **Subscription-Based CLI Access**: Use your existing ChatGPT Plus, Claude Pro, GitHub Copilot subscriptions
-- **Unified Interface**: Single API for all providers with consistent streaming responses
-- **Auto-Detection**: Automatic CLI tool discovery and path configuration
-
-### 🔧 CLI Provider Support
-
-| Provider | Integration | Authentication |
-|----------|-------------|----------------|
-| **Codex CLI** | Access GPT-5 with high reasoning | ChatGPT Plus subscription |
-| **Claude Code CLI** | Claude via Anthropic | Anthropic Pro subscription |
-| **Gemini CLI** | Google Cloud | Google Cloud authentication |
-| **GitHub Copilot** | VS Code Language Model API | GitHub Copilot subscription |
-
-### 🛠 Advanced Tooling
-
-- **Model Context Protocol (MCP)**: Hosted MCP server with OAuth authentication
-- **Multi-Authentication**: Both OAuth and API token support for maximum flexibility
-- **Process Execution**: Cross-platform CLI management with timeout handling
-- **Path Auto-Discovery**: Smart detection of CLI installations across Windows, macOS, Linux
-- **Real-time Status**: Live CLI availability and authentication checking
-
-### 🔒 Security & Authentication
-
-- **Encrypted Storage**: Browser-based API key encryption using SubtleCrypto API
-- **OAuth Bridges**: Secure authentication flows
-- **Subscription Auth**: No API costs - use existing subscriptions
-- **Local Storage**: Keys never leave your device
+---
 
-### 📊 Monitoring & Analytics
+## Quick Start
 
-- **PostHog Integration**: Advanced user analytics and feature tracking
-- **BetterStack Monitoring**: System health and performance monitoring
-- **Upstash Redis**: High-performance caching layer
-- **Supabase Auth**: Robust authentication system
+### 1. Get your free API token
 
-## Tech Stack
+**[polydev.ai/dashboard/mcp-tokens](https://polydev.ai/dashboard/mcp-tokens)**
 
-| Layer | Technology |
-|-------|------------|
-| **Frontend** | Next.js 15 (App Router), React 18, TypeScript, Tailwind CSS, shadcn/ui |
-| **Backend** | Supabase (PostgreSQL + Auth), Upstash Redis |
-| **AI Integration** | Custom TypeScript handlers for 8+ LLM providers |
-| **CLI Integration** | Cross-platform process execution utilities |
-| **Streaming** | Server-Sent Events for real-time responses |
-| **Monitoring** | PostHog Analytics, BetterStack |
+| Tier | Messages/Month | Price |
+|------|----------------|-------|
+| **Free** | 1,000 | $0 |
+| **Pro** | 10,000 | $19/mo |
 
-## Supported LLM Providers
+### 2. Install
 
-| Provider | Models | Context Window | Features |
-|----------|--------|----------------|----------|
-| **Anthropic** | Claude 3.5 Sonnet, Haiku, Opus | 200K tokens | Best for reasoning and code |
-| **OpenAI** | GPT-4o, GPT-4 Turbo, GPT-3.5 | 128K tokens | Versatile, widely adopted |
-| **Google Gemini** | Gemini 1.5 Pro, Flash | 1M+ tokens | Large context window |
-| **OpenRouter** | 100+ models | Varies | Access to multiple providers |
-| **Groq** | Open-source models | Varies | Ultra-fast inference |
-| **Perplexity** | Search-optimized models | Varies | AI search and reasoning |
-| **DeepSeek** | Reasoning models | Varies | Advanced reasoning capabilities |
-| **Mistral AI** | European AI models | Varies | Strong performance, EU-based |
+```bash
+npx polydev-ai@latest
+```
 
-## MCP Tools Available
+---
 
-- **Research**: Exa (web search), DeepWiki, Context7
-- **Storage**: Supabase, Upstash Redis, Memory (knowledge graph)
-- **Development**: GitHub, Git, Filesystem
-- **Infrastructure**: Vercel, Stripe
-- **AI**: Polydev (multi-model consultation)
-- **Communication**: Resend (email)
+## Setup
 
-## Architecture
+### Claude Code
 
-```
-┌─────────────────┐    ┌──────────────────┐    ┌─────────────────┐
-│   Frontend UI   │────│  Process Utils   │────│   CLI Tools     │
-│   (React/TS)    │    │  (Node.js)       │    │   (External)    │
-└─────────────────┘    └──────────────────┘    └─────────────────┘
-         │                        │                        │
-         │                        │                        │
-         ▼                        ▼                        ▼
-┌─────────────────┐    ┌──────────────────┐    ┌─────────────────┐
-│ LLM Service     │    │ CLI Handlers     │    │ Subscriptions   │
-│ (Unified API)   │    │ (Per Provider)   │    │ (ChatGPT+, etc) │
-└─────────────────┘    └──────────────────┘    └─────────────────┘
-         │                        │
-         │                        │
-         ▼                        ▼
-┌─────────────────┐    ┌──────────────────┐
-│   Supabase      │    │   MCP Platform   │
-│   (Auth + DB)   │    │   (16+ Tools)    │
-└─────────────────┘    └──────────────────┘
+```bash
+claude mcp add polydev -- npx -y polydev-ai@latest
 ```
 
-## Quick Start
+Then set your token:
+```bash
+export POLYDEV_USER_TOKEN="pd_your_token_here"
+```
 
-### Prerequisites
+Or add to `~/.claude.json`:
+
+```json
+{
+  "mcpServers": {
+    "polydev": {
+      "command": "npx",
+      "args": ["-y", "polydev-ai@latest"],
+      "env": {
+        "POLYDEV_USER_TOKEN": "pd_your_token_here"
+      }
+    }
+  }
+}
+```
 
-- Node.js 18+
-- npm or yarn package manager
-- (Optional) CLI tools for subscription-based access:
-  - Codex CLI for ChatGPT Plus integration
-  - Claude Code CLI for Anthropic Pro integration
-  - Gemini CLI for Google Cloud integration
-  - VS Code with GitHub Copilot for Copilot integration
+### Cursor / Windsurf / Cline
+
+Add to your MCP config:
+
+```json
+{
+  "mcpServers": {
+    "polydev": {
+      "command": "npx",
+      "args": ["-y", "polydev-ai@latest"],
+      "env": {
+        "POLYDEV_USER_TOKEN": "pd_your_token_here"
+      }
+    }
+  }
+}
+```
 
-### Installation
+### OpenAI Codex CLI
 
-```bash
-# Clone the repository
-git clone https://github.com/backspacevenkat/polydev-ai.git
-cd polydev-ai
+Add to `~/.codex/config.toml`:
 
-# Install dependencies
-npm install
+```toml
+[mcp_servers.polydev]
+command = "npx"
+args = ["-y", "polydev-ai@latest"]
 
-# Configure environment variables
-cp .env.example .env.local
+[mcp_servers.polydev.env]
+POLYDEV_USER_TOKEN = "pd_your_token_here"
 
-# Start development server
-npm run dev
+[mcp_servers.polydev.timeouts]
+tool_timeout = 180
+session_timeout = 600
 ```
 
-Open the application at http://localhost:3000
-
-### Quick Configuration
-
-1. **API Key Setup**: Go to Settings → API Keys tab to configure traditional API access
-2. **CLI Setup**: Go to Settings → CLI Subscriptions tab to set up subscription-based access
-3. **Provider Selection**: Choose your preferred LLM provider from the dropdown
-4. **Test Integration**: Use the chat interface to test your configuration
+---
 
-## Environment Variables
+## Usage
 
-```env
-# Supabase
-NEXT_PUBLIC_SUPABASE_URL=your_supabase_url
-NEXT_PUBLIC_SUPABASE_ANON_KEY=your_supabase_anon_key
-SUPABASE_SERVICE_ROLE_KEY=your_service_role_key
+Once connected, your agent can call:
 
-# PostHog Analytics
-NEXT_PUBLIC_POSTHOG_KEY=your_posthog_key
-NEXT_PUBLIC_POSTHOG_HOST=https://us.i.posthog.com
+```typescript
+{
+  "tool": "get_perspectives",
+  "arguments": {
+    "prompt": "How should I refactor this authentication flow?",
+    "user_token": "pd_your_token_here"
+  }
+}
+```
 
-# Upstash Redis
-UPSTASH_REDIS_REST_URL=your_upstash_redis_url
-UPSTASH_REDIS_REST_TOKEN=your_upstash_redis_token
+Or just mention "polydev" or "perspectives" in your prompt:
 
-# BetterStack Logging
-BETTERSTACK_LOGS_TOKEN=your_betterstack_token
 ```
+"Use polydev to debug this infinite loop"
+"Get perspectives on: Should I use Redis or PostgreSQL for caching?"
+```
+
+Returns structured perspectives from multiple models with reasoning and recommendations.
 
-## CLI Provider Setup
+---
 
-### Codex CLI (ChatGPT Plus Integration)
+## How It Works
 
-```bash
-# Install and authenticate
-codex auth
-codex --version
+```
+Your Agent → Polydev → [GPT 5.2, Claude Opus 4.5, Gemini 3, Grok 4.1] → Synthesized Answer
 ```
 
-### Claude Code CLI (Anthropic Pro Integration)
+When your AI agent gets stuck, Polydev consults multiple frontier models simultaneously and returns their perspectives. One API call, four expert opinions.
 
-```bash
-# Install and authenticate
-claude login
-claude --version
-```
+---
 
-### Gemini CLI (Google Cloud Integration)
+## Research
 
-```bash
-# Install Google Cloud SDK and authenticate
-gcloud auth login
-gcloud auth application-default login
-```
+Our approach achieves **74.6% on SWE-bench Verified** (Resolve@2), matching Claude Opus at 62% lower cost.
 
-### GitHub Copilot Integration
+| Approach | Resolution Rate | Cost/Instance |
+|----------|-----------------|---------------|
+| Claude Haiku (baseline) | 64.6% | $0.18 |
+| + Polydev consultation | 66.6% | $0.24 |
+| **Resolve@2 (best of both)** | **74.6%** | $0.37 |
+| Claude Opus (reference) | 74.4% | $0.97 |
 
-1. Install VS Code with GitHub Copilot extension
-2. Sign in with your GitHub account that has Copilot access
-3. The application will detect VS Code and Copilot availability automatically
+**[Read the full paper →](https://polydev.ai/articles/swe-bench-paper)**
 
-## Development Status
+---
 
-**Current Status**: Active Development
+## Links
 
-The platform is fully functional for:
-- Multi-LLM chat interface with streaming
-- API key management with client-side encryption
-- CLI subscription integration
-- MCP server with 16+ tools
-- Real-time streaming responses
+- **Website:** [polydev.ai](https://polydev.ai)
+- **Dashboard:** [polydev.ai/dashboard](https://polydev.ai/dashboard)
+- **npm:** [npmjs.com/package/polydev-ai](https://www.npmjs.com/package/polydev-ai)
+- **Research:** [SWE-bench Paper](https://polydev.ai/articles/swe-bench-paper)
+
+---
 
 ## License
 
-MIT
+MIT License - see [LICENSE](LICENSE) for details.
 
-## Links
+---
 
-- **Website**: [polydev.ai](https://polydev.ai)
-- **Repository**: [github.com/backspacevenkat/polydev-ai](https://github.com/backspacevenkat/polydev-ai)
+<p align="center">
+  <b>Built by <a href="https://polydev.ai">Polydev AI</a></b><br>
+  <i>Multi-model consultation for better code</i>
+</p>

package/lib/cliManager.js

@@ -578,16 +578,39 @@ This is a known issue with @google/gemini-cli@0.3.4 and older Node.js versions.`
         // Build args with model flag if specified
         let args = Array.isArray(promptArgs) ? [...promptArgs] : [];
         
+        // Normalize model names to CLI-compatible formats
+        let cliModel = model;
+        if (model && providerId === 'claude_code') {
+          // Map common model names to Claude CLI aliases/full names
+          const claudeModelMap = {
+            'claude-opus-4-5': 'opus',
+            'claude-opus-4.5': 'opus',
+            'claude-4.5-opus': 'opus',
+            'claude-opus-4-5-20250514': 'opus',
+            'claude-sonnet-4-5': 'sonnet',
+            'claude-sonnet-4.5': 'sonnet', 
+            'claude-4.5-sonnet': 'sonnet',
+            'claude-sonnet-4-5-20250514': 'sonnet',
+            'claude-3-5-sonnet': 'sonnet',
+            'claude-3-5-haiku': 'haiku',
+            'claude-haiku-3-5': 'haiku',
+          };
+          cliModel = claudeModelMap[model.toLowerCase()] || model;
+          if (cliModel !== model) {
+            console.log(`[Polydev CLI] Mapped model '${model}' to Claude CLI alias '${cliModel}'`);
+          }
+        }
+        
         // Add model flag based on CLI type
-        if (model) {
+        if (cliModel) {
           if (providerId === 'claude_code') {
             // Claude Code uses --model flag
-            args = ['--model', model, ...args, prompt];
+            args = ['--model', cliModel, ...args, prompt];
           } else if (providerId === 'gemini_cli') {
             // Gemini CLI: -m for model, -p for prompt (headless mode)
             // Add prompt prefix to prevent tool planning in non-interactive mode
             const geminiPrompt = `Answer directly without using any tools, file operations, or searches. Do not say "I will search" or "I will look up". Provide your analysis immediately.\n\n${prompt}`;
-            args = ['-m', model, '-p', geminiPrompt];
+            args = ['-m', cliModel, '-p', geminiPrompt];
           } else {
             // Default: just append prompt
             args = [...args, prompt];

package/mcp/manifest.json

@@ -1,7 +1,7 @@
 {
   "name": "polydev-perspectives",
   "version": "1.3.0",
-  "description": "Agentic workflow assistant - get diverse perspectives from multiple LLMs when stuck or need enhanced reasoning",
+  "description": "Multi-model AI perspectives - query GPT 5.2, Claude Opus 4.5, Gemini 3, and Grok 4.1 simultaneously. Get diverse perspectives when stuck or need enhanced reasoning. Achieved 74.6% on SWE-bench Verified.",
   "author": "Polydev AI",
   "license": "MIT",
   "main": "server.js",

package/mcp/stdio-wrapper.js

@@ -1706,4 +1706,33 @@ if (require.main === module) {
   });
 }
 
+/**
+ * Smithery sandbox server factory
+ * Creates a mock server instance for Smithery's capability scanning
+ * This allows Smithery to discover tools/resources without real credentials
+ */
+function createSandboxServer() {
+  // Return a minimal server that exposes our tool definitions for scanning
+  // No real API calls will be made - this is just for capability discovery
+  const fs = require('fs');
+  const path = require('path');
+  
+  const manifestPath = path.join(__dirname, 'manifest.json');
+  const manifest = JSON.parse(fs.readFileSync(manifestPath, 'utf8'));
+  
+  return {
+    serverInfo: {
+      name: manifest.name,
+      version: manifest.version
+    },
+    capabilities: { tools: {} },
+    tools: manifest.tools.map(tool => ({
+      name: tool.name,
+      description: tool.description,
+      inputSchema: tool.inputSchema
+    }))
+  };
+}
+
 module.exports = StdioMCPWrapper;
+module.exports.createSandboxServer = createSandboxServer;

package/package.json

@@ -1,8 +1,11 @@
 {
   "name": "polydev-ai",
-  "version": "1.8.59",
+  "version": "1.8.62",
+  "engines": {
+    "node": ">=20.x <=22.x"
+  },
   "mcpName": "io.github.backspacevenkat/perspectives",
-  "description": "Agentic workflow assistant with CLI integration - get diverse perspectives from multiple LLMs when stuck or need enhanced reasoning",
+  "description": "Multi-model AI perspectives for coding agents - query GPT 5.2, Claude Opus 4.5, Gemini 3, and Grok 4.1 simultaneously through one MCP server",
   "keywords": [
     "mcp",
     "model-context-protocol",