What happened?
Even though logPrompts is set to false in the configuration, user prompts are still being recorded in the api_request log as a session's conversation history.
{
"jsonPayload": {
"message": "API request to gemini-2.5-flash.",
"process.runtime.name": "nodejs",
"request_text": "..., {\"role\":\"user\",\"parts\":[{\"text\":\"hello, are you ready?\"}]},{\"role\":\"model\",\"parts\":...",
"installation.id": "...",
"host.name": "...",
"prompt_id": "...",
"host.id": "...",
"event.name": "gemini_cli.api_request",
"event.timestamp": "2026-02-13T07:42:14.794Z",
"model": "gemini-2.5-flash",
"process.runtime.description": "Node.js"
},
"timestamp": "2026-02-13T07:42:14.793999910Z",
}(Some unimportant fields have been removed from sample above)
What did you expect to happen?
The request_text field should be omitted from the gemini_cli.api_request log when logPrompts is false.
Client information
Client Information
Run gemini to enter the interactive CLI, then run the /about command.
> /about
╭─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ │
│ About Gemini CLI │
│ │
│ CLI Version 0.28.2 │
│ Git Commit da5e47ae3 │
│ Model auto-gemini-2.5 │
│ Sandbox no sandbox │
│ OS darwin │
│ Auth Method Logged in with Google (my Google Account) │
│ GCP Project my-google-cloud-project-id │
│ IDE Client VS Code │
│ │
╰─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
Login information
No response
Anything else we need to know?
-
in toSemanticLog, it seems to check logPrompts flag.
-
(additionally, response_text of gemini_cli.api_response and gen_ai.output.messages of gen_ai.client.inference.operation.details also should likely also be omitted when logPrompts is false.)
If this behavior is not intended, I am happy to submit a PR to fix this by updating ApiRequestEvent.toLogRecord to align with the logic used in toSemanticLogRecord.
Thank you.
What happened?
Even though
logPromptsis set tofalsein the configuration, user prompts are still being recorded in theapi_requestlog as a session's conversation history.{ "jsonPayload": { "message": "API request to gemini-2.5-flash.", "process.runtime.name": "nodejs", "request_text": "..., {\"role\":\"user\",\"parts\":[{\"text\":\"hello, are you ready?\"}]},{\"role\":\"model\",\"parts\":...", "installation.id": "...", "host.name": "...", "prompt_id": "...", "host.id": "...", "event.name": "gemini_cli.api_request", "event.timestamp": "2026-02-13T07:42:14.794Z", "model": "gemini-2.5-flash", "process.runtime.description": "Node.js" }, "timestamp": "2026-02-13T07:42:14.793999910Z", }(Some unimportant fields have been removed from sample above)
What did you expect to happen?
The
request_textfield should be omitted from thegemini_cli.api_requestlog whenlogPromptsis false.Client information
Client Information
Run
geminito enter the interactive CLI, then run the/aboutcommand.Login information
No response
Anything else we need to know?
in
toSemanticLog, it seems to checklogPromptsflag.(additionally,
response_textofgemini_cli.api_responseandgen_ai.output.messagesofgen_ai.client.inference.operation.detailsalso should likely also be omitted when logPrompts is false.)If this behavior is not intended, I am happy to submit a PR to fix this by updating
ApiRequestEvent.toLogRecordto align with the logic used intoSemanticLogRecord.Thank you.