gemini-cli - 💡(How to fix) Fix Local Model Routing logs do not show routing decisions described in documentation

What happened?

I enabled the local model routing feature using gemini gemma setup and confirmed that the status was "Active".

According to the documentation (gemini gemma — Automated Local Model Routing Setup), I expected to be able to observe routing behavior in the server logs in real time.

However, I observed the following behavior:

1. When asking natural language questions in the CLI, no logs were generated in litert-server.log.
2. Logs only appeared when manually running commands such as gemini gemma status, which directly query the local LiteRT server.
3. Even when logs were generated, they only showed that a request was handled (for example, Handling generateContent for model: gemma3-1b-gpu-custom) and did not indicate which model (Gemini Flash or Gemini Pro) the request was ultimately routed to.

Additional context

• Project: Local Model Routing with Gemma 3 1B
• Logs were checked using:
  • gemini gemma logs
  • direct inspection of ~/.gemini/tmp/litert-server.log

What did you expect to happen?

As described in the documentation, I expected the server logs to display routing activity for every user request in real time.

Specifically, I expected the logs to clearly indicate:

• when a routing request was triggered from a normal CLI prompt
• which target model was selected by the local Gemma routing model
  • e.g. Gemini Flash or Gemini Pro
• the routing decision process for each request

Client information

• CLI Version: 0.42.0
• Git Commit: 68e2196d5
• Operating System: darwin v24.11.1
• Sandbox Environment: no sandbox
• Model Version: gemini-3-flash-preview
• Auth Type: oauth-personal
• Memory Usage: 237.8 MB
• Terminal Name: Unknown
• Terminal Background: #141414
• Kitty Keyboard Protocol: Unsupported

Login information

No response

Anything else we need to know?

No response