GPTSL for Copilot Chat

VS Code extension that exposes GPTSL AI Gateway models in the VS Code Copilot Chat model picker and shows API key usage in the status bar.

Features

Copilot Chat Model Provider

This extension registers GPTSL as a VS Code Language Model Chat provider. After the API key is configured, GPTSL models can appear in Copilot Chat contexts that support third-party language model providers.

Users must configure their own global gateway base URL. Built-in default models are provided without any embedded gateway URL, and each model can define token limits, vision support, API mode, and temperature. The provider supports both OpenAI Responses-style gateway requests and Anthropic messages-style gateway requests.

Usage Status Bar

The right-side VS Code status bar shows GPTSL usage for the configured API key.

Status bar behavior:

• Shows usage as a percentage by default.
• Can show current spend amount instead.
• Click the status bar item to refresh usage.
• Hover the status bar item to view detailed usage information.

Hover details include:

• Current spend
• Budget limit, when returned by the gateway
• Usage percentage
• Text progress bar
• Display mode
• Key name
• Key alias
• Parsed user name
• Last updated time

The hover also includes quick actions for refreshing usage, opening settings, and toggling the display mode.

Shared API Key

Chat requests and usage queries use the same API key setting:

• gptslForCopilotChat.apiKey

The legacy setting gptsl.apiKey is still read as a fallback.

Settings Command

Use GPTSL: Open Settings or the status bar prompt to set the API key and edit model configuration in the native VS Code settings UI.

Settings

• gptslForCopilotChat.apiKey: Bearer token used to call the configured gateway.
• gptslForCopilotChat.baseUrl: global gateway base URL used by all models.
• gptslForCopilotChat.models: model configuration used by the Copilot Chat provider.
• gptslForCopilotChat.usageDisplayMode: status bar usage display mode, either percentage or amount.

Commands

• GPTSL: Open Settings: open GPTSL settings and prompt for an API key if one is missing.
• GPTSL: Refresh Usage: refresh status bar usage information.
• GPTSL: Toggle Usage Display Mode: switch the status bar between percentage and amount display.

Model Configuration

gptslForCopilotChat.models is an array of model objects.

Supported fields:

• id: model identifier sent to the gateway.
• apiMode: request format. Use openai-responses or anthropic.
• owned_by: optional owner label.
• configId: optional internal config identifier.
• context_length: maximum input context tokens.
• max_tokens: maximum output tokens.
• vision: whether the model supports image input.
• temperature: optional sampling temperature.

Example:

[
  {
    "id": "GPT5.5",
    "context_length": 1000000,
    "max_tokens": 128000,
    "vision": true,
    "apiMode": "openai-responses",
    "temperature": 0
  },
  {
    "id": "claude-sonnet-4-6",
    "context_length": 200000,
    "max_tokens": 40960,
    "vision": true,
    "apiMode": "anthropic"
  }
]

Usage Query

Usage is fetched from gptslForCopilotChat.baseUrl plus /key/info. If the configured base URL ends with a version suffix such as /v1, the usage client removes that version suffix before building the usage URL.

The extension sends the configured API key as both the key query parameter and the api-key request header, matching the GPTSL usage endpoint behavior.

If the endpoint does not return a global budget, the extension reads model-level budget limits and uses the largest returned budget limit to calculate the status bar percentage.

Development

npm install
npm run compile

Press F5 in VS Code to open an Extension Development Host.