Issue with tokens that crewai generates

[adharshctr]

Description

I have a crew application running and getting the usage metrics as follows:

Here prompt_tokens(input) is greater the completion_tokens(output)

But my input and output are follows

INPUT

OUTPUT

Steps to Reproduce

1. Build a crewai application
2. collect the traces of this application and see the usage metrics

Expected behavior

Completion token should be greater than prompt tokens in my case, but getting prompt tokens as big

Screenshots/Code snippets

Operating System

macOS Sonoma

Python Version

3.10

crewAI Version

0.83.0

crewAI Tools Version

0.17.0

Virtual Environment

Venv

Evidence

Possible Solution

1. How the prompt and completion tokens are collected?
2. Is there is any way to get the prompt and completion token count of each llm used in agent than count of prompt and completion tokens from a crew ?

Additional context

Nil