AI Engineer Portal

Your personal operating system for career transition.

Private mode

Exercise

Build a token cost tracker

Build a Token Cost Tracker

Without explicit cost tracking, AI features develop cost problems silently. Token counts accumulate, model choices drift toward expensive options, and the first signal of a problem is a billing invoice. A cost tracker built into the application layer gives you visibility before the bill arrives.

What you are building

Build a TokenCostTracker class that:

Records UsageRecord events (model, input tokens, output tokens, feature name, timestamp)
Computes cost in USD using a configurable pricing table
Provides a daily_report() that returns per-model and per-feature breakdowns for the last 24 hours
Enforces a daily_budget_usd limit: raises BudgetExceededError when a new record would push the daily total over budget
Provides a budget_remaining() method that returns how much budget is left today
Is thread-safe (use threading.Lock)

Pricing table to support

# USD per 1M tokens
PRICING = {
    "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},
}

Why this matters

Cost tracking is operational hygiene, not optional infrastructure. Teams that do not track cost per feature cannot make informed model routing decisions, cannot detect when a prompt change makes output tokens grow, and cannot answer "why did costs jump 30% this week?" with data instead of guesses.

Llm Foundations / medium / Step 4 of 6

Practice stage

General drill

Hint

Keep the solution explicit and reviewable.

Next drill

Add prompt injection detection

Success criteria

Make the solution explicit, debuggable, and easy to explain.

Review checklist

Review where the boundary is, what gets validated, and what would be hard to debug later.

from __future__ import annotations

import threading
import time
from collections import defaultdict
from dataclasses import dataclass, field
from typing import Optional

PRICING = {
    "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},
}

class BudgetExceededError(Exception):
    """Raised when recording a usage event would exceed the daily budget."""

@dataclass
class UsageRecord:
    model: str
    input_tokens: int
    output_tokens: int
    feature: str
    timestamp: float = field(default_factory=time.time)
    cost_usd: float = 0.0  # computed on creation

class TokenCostTracker:
    """Thread-safe per-request cost tracker with daily budget enforcement."""

def __init__(self, daily_budget_usd: float = 10.0) -> None:
        # TODO: store budget, initialize records list, initialize lock
        raise NotImplementedError

def compute_cost(self, model: str, input_tokens: int, output_tokens: int) -> float:
        """Compute cost in USD for a single request."""
        # TODO: look up pricing, compute (input * rate + output * rate) / 1_000_000
        # Return 0.0 if model is not in the pricing table
        raise NotImplementedError

def record(self, model: str, input_tokens: int, output_tokens: int, feature: str = "unknown") -> UsageRecord:
        """
        Record a usage event and enforce budget.
        Raises BudgetExceededError if this record would exceed the daily budget.
        """
        # TODO: compute cost, check budget, append record, return UsageRecord
        raise NotImplementedError

def budget_remaining(self) -> float:
        """Return how much budget is left for the current 24-hour window."""
        # TODO: sum costs in last 24h, return daily_budget - total
        raise NotImplementedError

def daily_report(self) -> dict:
        """
        Return usage breakdown for the last 24 hours.
        Shape: {
            "total_cost_usd": float,
            "by_model": { model_name: {"calls": int, "input_tokens": int, "output_tokens": int, "cost_usd": float} },
            "by_feature": { feature_name: {"calls": int, "cost_usd": float} },
        }
        """
        # TODO: aggregate records from last 24h
        raise NotImplementedError

Practice

Generate a variation

Generate a new exercise variation to deepen understanding or practice a related concept.

Attempt history

Recent submissions

Before you submit, decide what a strong answer should make obvious to the reviewer.

No attempts yet.