Your personal operating system for career transition.

Track token usage and cost per model across async calls

Cost visibility is the first step to cost control. This exercise builds a thread-safe cost tracker that can be used as a decorator around any async LLM call.

What you are building

Implement a CostTracker class and a track decorator:

CostTracker:

• record(model: str, input_tokens: int, output_tokens: int) -> None — Record a single call (thread-safe)
• total_cost_usd(model: str | None = None) -> float — Total cost across all models, or for a specific model
• summary() -> dict — Per-model breakdown: {model: {"calls": int, "input_tokens": int, "output_tokens": int, "cost_usd": float}}

Pricing (per million tokens):

"claude-haiku-4-5":  input=$0.80, output=$4.00
"claude-sonnet-4-5": input=$3.00, output=$15.00
"gpt-4o-mini":       input=$0.15, output=$0.60
"gpt-4o":            input=$2.50, output=$10.00

For unknown models, use $3.00/$15.00 as the default rate.

track decorator:

Implement track(tracker: CostTracker, model: str) as a decorator factory. It wraps an async function that returns a dict with input_tokens and output_tokens keys. After each successful call, it records the usage in the tracker.

Example

tracker = CostTracker()

@track(tracker, model="gpt-4o-mini")
async def call_openai(prompt: str) -> dict:
    return {"text": "response", "input_tokens": 100, "output_tokens": 50}

asyncio.run(call_openai("hello"))
print(tracker.summary())
# {"gpt-4o-mini": {"calls": 1, "input_tokens": 100, "output_tokens": 50, "cost_usd": 0.0000450}}