🐛 What happened?
The opencodego.apiBaseUrl and opencodezen.apiBaseUrl settings appear to be ignored. After configuring a custom compatible API gateway and reloading VS Code, the extension still appears to send requests to the default opencode.ai endpoints instead of the configured base URL.
🔄 Steps to reproduce
- Set
opencodego.apiBaseUrl (or opencodezen.apiBaseUrl) in User Settings to a compatible custom gateway URL.
- Reload the VS Code window.
- Use Refresh Models, Test Connection, or send a chat request with that provider.
- Check the gateway access logs or the OpenCode output channel to see which host receives the request.
✅ What did you expect?
Requests should use the configured base URL, with the appropriate route appended (for example, /models, /chat/completions, /messages, or /responses; /usage for Go where applicable).
🧾 What happened instead?
Requests appear to continue using the built-in defaults: https://opencode.ai/zen/go/v1/... for Go or https://opencode.ai/zen/v1/... for Zen. This makes a custom gateway unusable.
The installed v0.7.5 package declares and documents both settings, but its compiled provider definitions still contain these fixed endpoint URLs. The settings key and URL helper appear in the bundle, but I could not find a runtime path applying the configured setting to the provider URLs. The changelog says custom base URLs should apply to chat, model discovery, inline completions, and Go usage sync, so this may be a packaging/runtime regression.
🖥️ Environment
- Extension version: 0.7.5
- VS Code version: Please fill in from Help → About
- Provider: OpenCode Go / OpenCode Zen (please specify which you tested)
- Model: Please fill in the model used, if testing chat
- Setting tested:
opencodego.apiBaseUrl / opencodezen.apiBaseUrl (please specify)
📸 Logs or screenshots (optional)
I can provide redacted gateway access logs or the relevant OpenCode output log. Any API keys or authorization headers should be removed before sharing.
🐛 What happened?
The
opencodego.apiBaseUrlandopencodezen.apiBaseUrlsettings appear to be ignored. After configuring a custom compatible API gateway and reloading VS Code, the extension still appears to send requests to the defaultopencode.aiendpoints instead of the configured base URL.🔄 Steps to reproduce
opencodego.apiBaseUrl(oropencodezen.apiBaseUrl) in User Settings to a compatible custom gateway URL.✅ What did you expect?
Requests should use the configured base URL, with the appropriate route appended (for example,
/models,/chat/completions,/messages, or/responses;/usagefor Go where applicable).🧾 What happened instead?
Requests appear to continue using the built-in defaults:
https://opencode.ai/zen/go/v1/...for Go orhttps://opencode.ai/zen/v1/...for Zen. This makes a custom gateway unusable.The installed v0.7.5 package declares and documents both settings, but its compiled provider definitions still contain these fixed endpoint URLs. The settings key and URL helper appear in the bundle, but I could not find a runtime path applying the configured setting to the provider URLs. The changelog says custom base URLs should apply to chat, model discovery, inline completions, and Go usage sync, so this may be a packaging/runtime regression.
🖥️ Environment
opencodego.apiBaseUrl/opencodezen.apiBaseUrl(please specify)📸 Logs or screenshots (optional)
I can provide redacted gateway access logs or the relevant OpenCode output log. Any API keys or authorization headers should be removed before sharing.