polydev-ai 1.8.61 → 1.8.63

package/README.md

@@ -2,30 +2,74 @@
 
 **Multi-model AI perspectives for your coding agents.**
 
-Get insights from GPT-5.2, Claude Opus 4.5, Gemini 3 Pro, and Grok 4.1 — all through one MCP server.
+Get insights from GPT 5.2, Claude Opus 4.5, Gemini 3, and Grok 4.1 — all through one MCP server.
+
+[![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)
+
+---
+
+## Why Polydev?
+
+**Stop copy-pasting between ChatGPT, Claude, and Gemini.** Get all their perspectives in your IDE with one request.
+
+| Metric | Result |
+|--------|--------|
+| **SWE-bench Verified** | 74.6% Resolve@2 |
+| **Cost vs Claude Opus** | 62% lower |
+| **Response time** | 10-40 seconds |
+
+> *"Different models have different blind spots. Combining their perspectives eliminates yours."*
 
 ---
 
 ## Quick Start
 
-```bash
-npx polydev-ai@latest
-```
+### 1. Get your free API token
 
-## What It Does
+**[polydev.ai/dashboard/mcp-tokens](https://polydev.ai/dashboard/mcp-tokens)**
 
-When your AI agent gets stuck, Polydev consults multiple frontier models simultaneously and returns their perspectives. One API call, four expert opinions.
+| Tier | Messages/Month | Price |
+|------|----------------|-------|
+| **Free** | 1,000 | $0 |
+| **Pro** | 10,000 | $19/mo |
 
+### 2. Install
+
+```bash
+npx polydev-ai@latest
 ```
-Your Agent → Polydev → [GPT-5.2, Claude, Gemini, Grok] → Synthesized Answer
-```
+
+---
 
 ## Setup
 
 ### Claude Code
 
 ```bash
-claude mcp add polydev-ai -- npx polydev-ai@latest
+claude mcp add polydev -- npx -y polydev-ai@latest
+```
+
+Then set your token:
+```bash
+export POLYDEV_USER_TOKEN="pd_your_token_here"
+```
+
+Or add to `~/.claude.json`:
+
+```json
+{
+  "mcpServers": {
+    "polydev": {
+      "command": "npx",
+      "args": ["-y", "polydev-ai@latest"],
+      "env": {
+        "POLYDEV_USER_TOKEN": "pd_your_token_here"
+      }
+    }
+  }
+}
 ```
 
 ### Cursor / Windsurf / Cline
@@ -35,28 +79,102 @@ Add to your MCP config:
 ```json
 {
   "mcpServers": {
-    "polydev-ai": {
+    "polydev": {
       "command": "npx",
-      "args": ["polydev-ai@latest"]
+      "args": ["-y", "polydev-ai@latest"],
+      "env": {
+        "POLYDEV_USER_TOKEN": "pd_your_token_here"
+      }
     }
   }
 }
 ```
 
+### OpenAI Codex CLI
+
+Add to `~/.codex/config.toml`:
+
+```toml
+[mcp_servers.polydev]
+command = "npx"
+args = ["-y", "polydev-ai@latest"]
+
+[mcp_servers.polydev.env]
+POLYDEV_USER_TOKEN = "pd_your_token_here"
+
+[mcp_servers.polydev.timeouts]
+tool_timeout = 180
+session_timeout = 600
+```
+
+---
+
 ## Usage
 
 Once connected, your agent can call:
 
+```typescript
+{
+  "tool": "get_perspectives",
+  "arguments": {
+    "prompt": "How should I refactor this authentication flow?",
+    "user_token": "pd_your_token_here"
+  }
+}
+```
+
+Or just mention "polydev" or "perspectives" in your prompt:
+
 ```
-polydev.getPerspectives("How should I refactor this authentication flow?")
+"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.
 
-## Why Multi-Model?
+---
+
+## How It Works
+
+```
+Your Agent → Polydev → [GPT 5.2, Claude Opus 4.5, Gemini 3, Grok 4.1] → Synthesized Answer
+```
+
+When your AI agent gets stuck, Polydev consults multiple frontier models simultaneously and returns their perspectives. One API call, four expert opinions.
+
+---
+
+## Research
+
+Our approach achieves **74.6% on SWE-bench Verified** (Resolve@2), matching Claude Opus at 62% lower cost.
+
+| 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 |
+
+**[Read the full paper →](https://polydev.ai/articles/swe-bench-paper)**
+
+---
+
+## Links
+
+- **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
 
-Different models have different blind spots. Our [research](https://polydev.ai/articles/swe-bench-paper) shows that consulting multiple models improves success rates by 10%+ on coding benchmarks — at lower cost than using a single frontier model.
+MIT License - see [LICENSE](LICENSE) for details.
 
 ---
 
-**[Get Started](https://polydev.ai)** · **[Docs](https://polydev.ai/docs/mcp-integration)** · **[Research](https://polydev.ai/articles/swe-bench-paper)**
+<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

@@ -1,11 +1,11 @@
 {
   "name": "polydev-ai",
-  "version": "1.8.61",
+  "version": "1.8.63",
   "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",