Gemini Code completions do not work when running over remote connections

[jshelman]

Expected Behavior

Code completions should work correctly in remote (host) ide window.

Actual Behavior

Code completions do not work when running Remote ide thru ssh connection.
Language server logs show a valid response from Gemini aicompanion endpoint, but do not make it to IDE window.

idea.log
2024-04-16 04:59:34,868 [ 121166] WARN - duet-ai-language-server - Rendering inline completion failed
com.intellij.diagnostic.PluginException: Positive width should be defined for an inline element by import org.springframework.boot.SpringApplication; [Plugin: com.google.gct.core]
at com.intellij.diagnostic.PluginProblemReporterImpl.createPluginExceptionByClass(PluginProblemReporterImpl.java:23)
at com.intellij.diagnostic.PluginException.createByClass(PluginException.java:90)
at com.intellij.openapi.editor.impl.InlineInlayImpl.doUpdate(InlineInlayImpl.java:75)
at com.intellij.openapi.editor.impl.InlayImpl.(InlayImpl.java:34)
at com.intellij.openapi.editor.impl.InlineInlayImpl.(InlineInlayImpl.java:27)
at com.intellij.openapi.editor.impl.InlayModelImpl.addInlineElement(InlayModelImpl.java:154)
at com.google.cloud.tools.intellij.languageclient.aicompanion.InlineCompletionResultProvider.renderCompletionTextInInlays(InlineCompletionResultProvider.kt:252)
at com.google.cloud.tools.intellij.languageclient.aicompanion.InlineCompletionResultProvider.renderInlineCompletionText(InlineCompletionResultProvider.kt:192)
at com.google.cloud.tools.intellij.languageclient.aicompanion.InlineCompletionResultRenderingService$sendRequestAndRender$1$1.invokeSuspend(InlineCompletionResultRenderingService.kt:170)
...

Additional Information

• IDE type (e.g. IntelliJ, Pycharm): Tested with same behavior on both Intellij (and PyCharm)
• IDE version: 2023.3.4, 2024.1
• Cloud Code version (Settings > Plugins > Cloud Code): 24.4.2-api-version-223

[ivanporty]

Thanks for the report @jshelman. Remote environments do no support the mechanism we use to render the completions - we are currently working to change this into a better implementation and will let you know once it's ready.