HUMAIN Node for Copilot Chat

Bring HUMAIN Node models into GitHub Copilot Chat. Every model available to your HUMAIN Node API key appears directly in the Copilot model picker — with streaming, tool calling, and vision support.

Built on VS Code's official Language Model Provider API, the same integration path Copilot uses for its own models.

Features

All your models, automatically — the model list is fetched from GET /v1/models using your key, so the picker only ever shows models you're allowed to call.
Streaming — responses stream token-by-token like native Copilot models. Models that don't support streaming are detected and called non-streamed automatically.
Tool calling — models that advertise function-calling support can use Copilot's tools (file search, terminal, etc.).
Vision — models that support image input can analyze screenshots and pasted images.
Identity override — models answer "which model are you?" with their HUMAIN Node model ID (optional, see configuration).
Secure key handling — your API key is stored in VS Code SecretStorage, never in settings, source control, or logs.

Requirements

VS Code 1.116 or newer
A HUMAIN Node account with an API key

Installation

Install from the VS Code Marketplace, or from the command line:

code --install-extension Fawaz.humain-node-copilot

Getting started

Install the extension.
Open the Command Palette (Ctrl+Shift+P) and run HUMAIN Node: Set API Key. Paste your key when prompted — it is stored securely and never written to disk.
Open Copilot Chat, click the model picker, and choose a HUMAIN Node model.
Start chatting.

Tip: run HUMAIN Node: Test Connection any time to verify your key and see how many models are available to it.

Configuration

Setting	Default	Description
`humainNode.baseUrl`	`https://api.node.humain.com/v1`	HUMAIN Node inference gateway origin (the `/v1` base URL).
`humainNode.identityInjection`	`true`	Adds a system instruction so the model identifies itself as `<model>` via HUMAIN Node when asked. Disable to let each model answer from its own training.

Commands

Command	Purpose
`HUMAIN Node: Set API Key`	Store your API key in SecretStorage
`HUMAIN Node: Clear API Key`	Remove the stored API key
`HUMAIN Node: Refresh Models`	Re-fetch the model list from HUMAIN Node
`HUMAIN Node: Test Connection`	Verify the key and list available models

How it works

The extension registers a VS Code language model chat provider (vendor: humain) via vscode.lm.registerLanguageModelChatProvider. Copilot Chat then interacts with it through two callbacks:

provideLanguageModelChatInformation — returns the models from GET https://api.node.humain.com/v1/models, mapping each model's metadata to Copilot capabilities:

HUMAIN metadata	Copilot capability
`node.supports_function_calling`	tool calling
`node.supports_images`	image input
`node.max_context_tokens`	`maxInputTokens`
`node.max_output_tokens`	`maxOutputTokens`

provideLanguageModelChatResponse — streams POST /v1/chat/completions (Server-Sent Events) and reports text, tool-call, and token-usage fragments back to Copilot.

Models whose metadata reports supports_streaming: false — or that reject a streamed request with non_stream_required — are automatically retried as a single non-streamed completion, so every model works.

Privacy & security

Your API key is stored in VS Code SecretStorage and sent only to the configured gateway (humainNode.baseUrl).
Requests go directly from VS Code to HUMAIN Node — no third-party relay.
The key is never written to settings, logs, or source control.

Development

npm install
npm run compile        # build
npm run watch          # build on change

Press F5 in VS Code to launch an Extension Development Host for testing.

Release

npm run package                    # produce humain-node-copilot-<version>.vsix
npx vsce login <publisher-id>      # once; paste your Marketplace-scoped PAT
npx vsce publish patch             # bump version and publish

License

MIT

HUMAIN Node for Copilot Chat is not affiliated with or endorsed by HUMAIN. All trademarks belong to their respective owners.