Pythontoken_tracker.py
Copy
1
import tiktoken
2
from openai import OpenAI
3
 
4
client = OpenAI()
5
enc = tiktoken.get_encoding("cl100k_base")
6
 
7
# Pricing (per 1M tokens as of 2025 — check current prices)
8
PRICING = {
9
    "gpt-4o": {"input": 2.50, "output": 10.00},
10
    "gpt-4o-mini": {"input": 0.15, "output": 0.60},
11
}
12
 
13
session_cost = 0.0
14
session_tokens = {"input": 0, "output": 0}
15
 
16
def count_tokens(text: str) -> int:
17
    return len(enc.encode(text))
18
 
19
def tracked_chat(messages: list[dict], model: str = "gpt-4o-mini") -> str:
20
    global session_cost, session_tokens
21
 
22
    # Estimate before call
23
    estimated = sum(count_tokens(m["content"]) for m in messages)
24
    print(f"Estimated input tokens: {estimated}")
25
 
26
    resp = client.chat.completions.create(model=model, messages=messages)
27
    usage = resp.usage
28
 
29
    # Actual usage
30
    input_tokens = usage.prompt_tokens
31
    output_tokens = usage.completion_tokens
32
    price = PRICING.get(model, {"input": 0, "output": 0})
33
    cost = (input_tokens * price["input"] + output_tokens * price["output"]) / 1_000_000
34
 
35
    session_tokens["input"] += input_tokens
36
    session_tokens["output"] += output_tokens
37
    session_cost += cost
38
 
39
    print(f"Actual: {input_tokens} in / {output_tokens} out | Cost: ${cost:.6f}")
40
    return resp.choices[0].message.content or ""
41
 
42
tracked_chat([{"role": "user", "content": "What is a transformer?"}])
43
print(f"\nSession total: {session_tokens} | Total cost: ${session_cost:.6f}")
44