-
Notifications
You must be signed in to change notification settings - Fork 2k
Description
Plugin Type
JetBrains Plugin
App Version
5.7.0
Description
I am experiencing a "Connection error" when using a custom OpenAI-compatible internal API in the JetBrains version of the plugin. The exact same configuration (Endpoint, API Key, and VPN environment) works perfectly in the VS Code version of the plugin.
Kilo Code error message:
OpenAI completion error: Connection error. You seem to be using a proxy-based provider. Make sure to check its logs and to ensure it's not rewriting Roo's requests.
The JetBrains "Check proxy settings" tool returns "Connection successful" for the target API URL, confirming that the IDE can reach the endpoint, but Kilo Code seems unable to inherit or use these network settings correctly.
Manual curl -X GET https://xxxxx.yyyyyyy/api/v1/models from the JetBrains terminal works and returns the model list correctly.
Reproduction steps
- Connect to a VPN required to access the API.
- IDE Check connection tool returns "Connection Successful".
- Configure Kilo Code with an OpenAI-compatible base URL (without /v1 suffix. Format: ttps://xxxxx.yyyyyyy/api).
- The autocomplete feature for the list of models does not return any results. However, I type the model manually.
- Attempt a completion.
- Error appears.
Troubleshooting already performed:
- Tested with and without /v1 or /api suffixes in the Base URL.
- Configured "No Proxy" in JetBrains System Settings to avoid interference.
- Added the server certificate to JetBrains "Server Certificates" (HTTPS).
- Reinstalled and restarted the plugin multiple times.
- Verified the API key and model name are correct.
Provider
OpenAI-compatible
Model
Indifferent
System Information
Node.js Version: v22.13.1
CPU Architecture: aarch64
Operating System: Mac OS X 15.7.3
IDE Version: PhpStorm 2025.2.3 (build PS-252.26830.95)
Plugin Version: 5.7.0
JCEF Support: Yes
Metadata
Metadata
Assignees
Labels
Type
Projects
Status