litellm-cost-tracker 0.1.0

A lightweight Python library for tracking LLM API costs via litellm's callback system

LLM Cost Tracker

A lightweight Python library for tracking LLM API costs via litellm's callback system, with budget alerts and spending limits.

Installation

pip install litellm-cost-tracker

Quick Start

import litellm
from llm_cost import CostTracker, BudgetExceededError

# Create a tracker with a $5 budget
def alert(tracker):
    print(f"Budget exceeded! Spent ${tracker.total_cost:.2f}")

tracker = CostTracker(
    budget=5.00,
    on_budget_exceeded=alert,
    raise_on_budget=True
)

# Register with litellm
litellm.callbacks = [tracker]

# Make LLM calls as usual
try:
    response = litellm.completion(
        model="gpt-4",
        messages=[{"role": "user", "content": "Hello!"}]
    )
except BudgetExceededError as e:
    print(f"Stopped at ${e.total_cost:.2f} (budget: ${e.budget:.2f})")

# Check usage
print(f"Total: ${tracker.total_cost:.4f} across {tracker.request_count} requests")

Features

• Real-time cost tracking during LLM calls
• Budget alerts via callback and/or exception
• In-memory storage (no persistence needed)
• Uses litellm's pricing data for accurate costs

API Reference

CostTracker

CostTracker(
    budget: float | None = None,           # Spending limit in USD
    on_budget_exceeded: Callable | None = None,  # Callback when exceeded
    raise_on_budget: bool = False          # Raise exception when exceeded
)

Properties:

• total_cost: float — Running total in USD
• request_count: int — Number of successful requests
• history: list[dict] — All logged requests
• budget: float | None — Configured budget
• budget_exceeded: bool — Whether budget has been exceeded

Methods:

• reset() — Clear all tracked data

BudgetExceededError

Raised when budget is exceeded (if raise_on_budget=True).

class BudgetExceededError(Exception):
    budget: float       # Configured budget
    total_cost: float   # Actual spend when exceeded

History Entry Format

Each request is logged with:

{
    "model": "gpt-4",
    "prompt_tokens": 150,
    "completion_tokens": 50,
    "cost": 0.0123,
    "timestamp": "2026-02-22T10:30:00Z"
}

Development

git clone https://github.com/Paawan13/litellm-cost-tracker.git
cd litellm-cost-tracker
pip install -e ".[dev]"
pytest

License

MIT