OpenCode Go Provider

VS Code extension to use OpenCode Go models in Copilot Chat with your own OpenCode Go subscription.

Requirements

• VS Code 1.104.0 or later
• GitHub Copilot extension installed and active
• An OpenCode Go API key (get one here)

Installation

From Source

1. Clone this repository.
2. Run bun install --ignore-scripts && bun run compile.
3. Press F5 in VS Code to launch the Extension Development Host.

From VSIX

1. Run bun install --ignore-scripts && bun run package:vsix.
2. Install the generated .vsix file via the Extensions view (Install from VSIX...).

Setup

1. Open Copilot Chat (Cmd/Ctrl + Alt + I) and open the model picker.
2. Choose Manage Models and add or configure OpenCode Go.
3. Enter your OpenCode Go API key when prompted.
4. If needed, you can still run OpenCode Go: Manage OpenCode Go API Key from the Command Palette.
5. Select OpenCode Go in Copilot Chat and choose a model.

Supported Models

The extension uses the source-controlled FALLBACK_MODELS list bundled in this repository. When OpenCode Go adds new models, this extension must be updated and republished. Current bundled models include:

• GLM-5, GLM-5.1
• DeepSeek V4 Pro, DeepSeek V4 Flash
• Kimi K2.5, Kimi K2.6
• MiMo-V2-Pro, MiMo-V2-Omni, MiMo-V2.5-Pro, MiMo-V2.5
• MiniMax M2.5, MiniMax M2.7
• Qwen3.5 Plus, Qwen3.6 Plus

Usage

1. Open Copilot Chat (Cmd/Ctrl + Alt + I).
2. Select OpenCode Go from the provider selector.
3. Choose a model (e.g., Kimi K2.6) and start chatting.

Development

bun install --ignore-scripts
bun run compile
bun run lint
bun run test -- --runInBand

Press F5 in VS Code to launch the Extension Development Host.

Available Scripts

• bun run compile – Compile TypeScript
• bun run watch – Compile with file watching
• bun run test – Run tests
• bun run lint – Lint check with ESLint
• bun run lint:fix – Auto-fix with ESLint
• bun run format – Format with Prettier
• bun run package:vsix – Create VSIX package
• bun run repro:deepseek -- "What model are you?" – Check DeepSeek upstream response without the extension
• bun run repro:compare -- "What model are you?" – Compare DeepSeek and reference model responses side by side
• bun run repro:compare:json -- "What model are you?" – Output comparison results as JSON
• bun run measure:deepseek – Measure deepseek-v4-flash:max / deepseek-v4-pro:max directly against /chat/completions
• bun run measure:deepseek:json – Output the same DeepSeek measurement results as JSON

Troubleshooting

DeepSeek Claiming a Different Model Name

To isolate whether the model identity issue is in the extension, run:

export OPENCODE_GO_API_KEY="your-api-key"
bun run repro:deepseek -- "What model are you?"

This script sends a /messages request directly to deepseek-v4-flash without going through the extension. If it still returns Claude or another model name, the cause is on the OpenCode Go side or upstream routing, not this extension.

To compare with a reference model at the same time, run:

export OPENCODE_GO_API_KEY="your-api-key"
bun run repro:compare -- "What model are you?"

To compare with a different combination, use --models with comma-separated IDs:

bun run repro:deepseek -- --models deepseek-v4-flash,glm-5,qwen3.6-plus "What model are you?"

To save results as JSON, add --json and redirect:

export OPENCODE_GO_API_KEY="your-api-key"
bun run repro:compare:json -- "What model are you?" > deepseek-compare.json

The same works for any combination of models:

bun run repro:deepseek -- --json --models deepseek-v4-flash,glm-5,qwen3.6-plus "What model are you?" > compare.json

Measuring DeepSeek Empty-Response Behavior Directly

To measure whether deepseek-v4-flash:max or deepseek-v4-pro:max are returning visible output directly from the OpenCode Go backend, run:

export OPENCODE_GO_API_KEY="your-api-key"
bun run measure:deepseek

For machine-readable results:

export OPENCODE_GO_API_KEY="your-api-key"
bun run measure:deepseek:json > deepseek-measure.json

Marketplace Packaging

bun run package:vsix

The command above produces a .vsix that can be uploaded in the VS Code Marketplace publisher portal.

Privacy

• Your API key is stored securely in VS Code and synced with the extension's SecretStorage compatibility path when needed.
• Chat requests are sent to https://opencode.ai/zen/go/v1.