Implement the passing of token usage information as part of the proxy's response

[afardana]

Hello,

I’d like to request a small enhancement to the Gemini CLI OpenAI proxy: exposing token usage details in the JSON response.
When you call the official OpenAI API, each response includes a tokenUsage object, for example:

"tokenUsage": {
  "completionTokens": 745,
  "promptTokens": 4336,
  "totalTokens": 5081
}

This makes it easy to track usage for cost monitoring, analytics, and debugging.

Is it possible to access the underlying Gemini token counts (or approximate them) within our Workers runtime and pass this information to the caller?