How to set a default model for a particular project?

[yashasvigirdhar]

For one of my project, I want to use gemini-2.0-flash-lite by default.

Launching gemini --model gemini-2.0-flash-lite works for me but I want to just launch it using gemini. I don't see a parameter in settings.json here: https://github.com/google-gemini/gemini-cli/blob/main/docs/cli/configuration.md to do it.

[Manamama]

A hack: use an alias: alias gemini=gemini --model gemini-2.0-flash-lite etc.

[yashasvigirdhar]

That's a good idea but I think I want it to be for one project, not all.. so would be great to have it as part of project configuration.

[rycore]

I don't know if you ever got a solution to your problem, but I thought I'd share an idea. In each project folder, place a file containing the model you want to use:

echo "gemini-2.0-flash-lite" > ~/project/name/gemini.model

Then save the following as a local function or script gemai (tested on Debian WSL):

ai_model="$(pwd)/gemini.model"
ai_command=()

[[ -f "${ai_model}" ]] && ai_command+=("--model" "$(<"${ai_model}")")

ai_command+=("$@")
gemini "${ai_command[@]}"

If you cd into your project directory, and run your function/script gemai -p "hello world" it will now use the model you set.

Also, if you want to keep the name gemini, you can replace the command within the function/script to /usr/local/bin/gemini, just use which gemini to get your path and replace that part of the code /usr/local/bin/gemini "${command[@]}", now you can name your script/function gemini.

This method allows you to use multiple AIs and their models on as many projects as you want, claude.model, gemini.model, gpt.model alternatively, you could store all the models in one file .ai-models and pull each model for each command using while IFS.

# Not Tested 
ai_file="$(pwd)/.ai-models"
ai_type="gemini"
ai_command=()

if [[ -f "${ai_file}" ]]; then
  while IFS='=' read -r key value; do
      if [[ "$key" == "$ai_type" ]]; then
          ai_command+=("--model" "$value")
          break
      fi
  done < "$ai_file"
fi

ai_command+=("$@")
/usr/local/bin/gemini "${ai_command[@]}"

I hope this helps.

[HunDun0Ben]

In the latest version, You can configure the model in your project by referring to the documentation. https://geminicli.com/docs/get-started/configuration/#configuration-layers
In your project’s .gemini/settings.json, add the following:

{
  "model": {
    "name": "gemini-2.5-flash"
  }
}