Using instructor from_litellm with completion_cost returns cost truncated to 1 decimal place

[JohnPeng47]

**Bug
Excecuting the following code ...

import instructor
from litellm import completion
from litellm import completion, completion_cost, cost_per_token
from pydantic import BaseModel


class User(BaseModel):
    name: str
    age: int


client = instructor.from_litellm(completion)
instructor_resp, _ = client.chat.completions.create_with_completion(
    model="claude-3-5-sonnet-20240620",
    max_tokens=1024,
    messages=[
        {
            "role": "user",
            "content": "Extract Jason is 25 years old.",
        }
    ],
    response_model=User,
)
instructor_cost = completion_cost(completion_response=instructor_resp, model="claude-3-opus-20240229")
print("Instructor cost: ", instructor_cost)

litellm_resp = completion(
    model="claude-3-5-sonnet-20240620",
    messages=[
        {"role": "user", "content": "Extract Jason is 25 years old."}
    ],
    max_tokens=1024,
)
litellm_cost = completion_cost(completion_response=litellm_resp, model="claude-3-opus-20240229")
print("Litellm cost: ", litellm_cost)

Results in:

Instructor cost:  0.0
Litellm cost:  0.00036300000000000004

Would like to like more precision in instructor cost response when supporting litellm API

[JohnPeng47]

Okay I fix:

Referencing -> BerriAI/litellm#5285

Basically LiteLLM introduced two of their own custom fields to handle prompt token calculations

[ivanleomk]

I suggest using the following snippet

import instructor
from litellm import completion
from pydantic import BaseModel


class User(BaseModel):
    name: str
    age: int


client = instructor.from_litellm(completion)
instructor_resp, raw_completion = client.chat.completions.create_with_completion(
    model="claude-3-5-sonnet-20240620",
    max_tokens=1024,
    messages=[
        {
            "role": "user",
            "content": "Extract Jason is 25 years old.",
        }
    ],
    response_model=User,
)

print(raw_completion._hidden_params["response_cost"])
#> 0.00189

Additionally, the LiteLLM call and the instructor call will not cost the same since we need to make a function call in the first one and then we're just doing a normal chat completion call in the second.

Closing this issue for now