[Bug]: Faulty calculation of custom pricing

[r4881t]

Describe the bug

With the openai gpt-4o-mini, I went ahead and tried setting the price param as

model_4o = "gpt-4o-mini"
llm_config_4o = {
  "config_list": [{"model": model_4o, "api_key": os.environ["OPENAI_API_KEY"]}],
  "stream": False,
  "temperature": 0,
  "cache_seed": 20,
  "timeout": 300,
  "max_tokens": 3 * 1024,
  "price": [0.00015, 0.0006],
}

Since it's mentioned that the price should be price per 1k tokens. However soon I noticed that the price it shows is quite high. So i went deeper and found that the code commited via #2902 contains a bug. Specifically at

autogen/autogen/oai/client.py

Line 798 in 0cdbc34

return n_input_tokens * price_1k[0] + n_output_tokens * price_1k[1]

where it multiplies price per 1k with the actual number of tokens. Thus gicing wrong value. To fix this I have changed the price to be price per token in my llm config, but the code needs to be updated so that people dont face this issue.

Steps to reproduce

No response

Model Used

No response

Expected Behavior

The price should be computed as price per 1k token

Screenshots and logs

No response

Additional Information

No response

[yiranwu0]

Yes, thanks! I will submit a PR for this.