⚡ OpenRouter Copilot Model Manager

Browse, filter, and add OpenRouter models to VS Code Copilot Chat with a rich visual interface.

✨ Features

🔍 Browse 400+ Models — Search and filter OpenRouter's entire model catalog
👁️ Capability Filtering — Filter by Vision, Tool Calling (Function Call), Free models
💰 Pricing Display — See input/output costs per million tokens
📊 Context & Output Info — View context window and max output token limits
🚀 Native Copilot Integration — Models register directly as a VS Code LanguageModelChatProvider — no external config files needed
🔧 Full Agentic Mode — Tool calling, terminal access, file operations, and web search through Copilot Chat
💭 Thinking/Reasoning Display — See model reasoning (Claude, DeepSeek-R1, Qwen3, etc.)
🎚️ Thinking Effort Control — Pick the reasoning depth per request straight from the Copilot model-picker's Thinking Effort submenu, or set a per-model default in the browser panel
🔄 Auto Retry with Backoff — Graceful handling of rate limits (429) and server errors
⏱️ Configurable Timeout — Prevent hanging requests with customizable timeouts
📈 Usage Stats — Token usage displayed in the status bar after each request
🔑 Secure API Key Storage — Keys stored in VS Code's SecretStorage
⚙️ Rich Settings — Temperature, max tokens, timeout, retries, custom API endpoint, and more

🚀 Quick Start

Install the extension (.vsix file)
Open Command Palette → OpenRouter: Set API Key
Open Command Palette → OpenRouter: Browse Models
Browse, filter, and add models
Click "Apply to Copilot"
Your models now appear in Copilot Chat model picker! 🎉

📦 Install from VSIX

code --install-extension openrouter-copilot-model-manager-1.0.4.vsix

Or: VS Code → Extensions → ... menu → Install from VSIX...

🔧 Commands

Command	Description
`OpenRouter: Browse Models`	Open the Model Browser panel
`OpenRouter: Set API Key`	Set/update your OpenRouter API key
`OpenRouter: Sync Models from API`	Manually fetch latest models

⚙️ Settings

All settings are under openrouterModelManager.* in VS Code Settings:

Setting	Default	Description
`cache.ttlMinutes`	`60`	Cache TTL in minutes before auto-refresh
`requestTimeoutSeconds`	`60`	Request timeout in seconds
`maxRetries`	`3`	Max retry attempts for failed requests
`defaultTemperature`	`null`	Default temperature (0.0–2.0), empty = model default
`defaultMaxTokens`	`null`	Default max output tokens, empty = model default
`apiEndpoint`	`https://openrouter.ai/api/v1`	Custom API endpoint URL
`enablePromptCaching`	`true`	OpenRouter automatic prompt caching (cuts agent-mode input costs)
`sanitizeBase64Content`	`true`	Strip long base64 blobs from prompts (avoids guardrail 403 blocks, saves tokens)
`defaultReasoningEffort`	`null`	Default thinking effort for reasoning models (`none`/`minimal`/`low`/`medium`/`high`/`xhigh`/`max`); empty = each model's catalog default
`logLevel`	`info`	Log level: debug, info, warn, error

🧠 Supported Capabilities

Capability	Status
Chat (text-to-text)	✅
Tool Calling (Function Call)	✅
Vision / Image Input	✅
Thinking/Reasoning	✅
Thinking Effort picker	✅
System Messages	✅
Streaming	✅
Retry with Backoff	✅
Usage Statistics	✅

🙏 Credits

The Copilot model-picker Thinking Effort integration was adapted from @Irvingouj's fork (c2dc1de), with the model-id heuristic removed and without enabling proposed APIs, so this build stays installable from the Marketplace on stable VS Code.

📄 License

MIT

OpenRouter Copilot Model Manager

Hüsamettin Ulutaş