llm-cost-guardian 0.1.1

Real-time cost monitoring and budget enforcement for LLM API calls

LLM Cost Guardian

New here? Start with the Getting Started Guide.

Real-time cost monitoring and budget enforcement for LLM API calls.

---

Why?

LLM API costs can spiral out of control fast - a single runaway loop can burn through hundreds of dollars in minutes. LLM Cost Guardian wraps your existing clients with transparent tracking and automatic budget enforcement so you never get a surprise bill again.

Features

• 📊 Real-time cost tracking - automatic per-call cost calculation from token usage
• 🛡️ Budget enforcement - hard caps, soft warnings, and sliding window policies
• 🔌 Drop-in wrappers - wrap OpenAI and Anthropic clients with one line of code
• 📈 Prometheus export - expose metrics for your monitoring stack
• 💾 JSON & CSV export - save usage reports for analysis
• 🖥️ CLI tool - estimate costs and view reports from the terminal
• 🧩 Extensible - add custom models, policies, and exporters
• 🔒 Thread-safe - safe for concurrent use in async applications

Quick Start

pip install llm-cost-guardian

from llm_cost_guardian import CostTracker, HardCapPolicy, BudgetManager

tracker = CostTracker()
budget = BudgetManager().add(HardCapPolicy(limit_usd=5.00))

# Track a call (or use the wrapper for automatic tracking)
tracker.record("gpt-4o", input_tokens=1500, output_tokens=800)
budget.enforce(tracker)  # raises BudgetError if over limit
print(f"Cost so far: ${tracker.total_cost:.4f}")

Architecture

┌──────────────┐     ┌─────────────────────────────────────┐     ┌──────────────┐
│              │     │        LLM Cost Guardian            │     │              │
│  Your Code   │────>│  ┌───────────┐   ┌──────────────┐   │────>│   LLM API    │
│              │     │  │  Tracker  │   │   Budget     │   │     │  (OpenAI /   │
│              │<────│  │  (costs)  │   │  (policies)  │   │<────│  Anthropic / │
│              │     │  └───────────┘   └──────────────┘   │     │   Google)    │
└──────────────┘     │  ┌───────────┐   ┌──────────────┐   │     └──────────────┘
                     │  │ Exporters │   │     CLI      │   │
                     │  │ (JSON/CSV/│   │              │   │
                     │  │Prometheus)│   │              │   │
                     │  └───────────┘   └──────────────┘   │
                     └─────────────────────────────────────┘

Usage

Basic Cost Tracking

from llm_cost_guardian import CostTracker

tracker = CostTracker()

# Record API calls manually
tracker.record("gpt-4o", input_tokens=1500, output_tokens=800)
tracker.record("claude-3-5-haiku-20241022", input_tokens=2000, output_tokens=600)

print(f"Total: ${tracker.total_cost:.6f}")
print(f"Tokens: {tracker.total_tokens:,}")
print(tracker.cost_by_model())

Drop-in Client Wrappers

Wrap your existing client - zero code changes needed:

from openai import OpenAI
from llm_cost_guardian import CostTracker, TrackedOpenAI

tracker = CostTracker()
client = TrackedOpenAI(OpenAI(), tracker)

# Use exactly like the normal client - costs tracked automatically
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello!"}],
)

print(f"This call cost: ${tracker.total_cost:.6f}")

Works the same way with Anthropic:

from anthropic import Anthropic
from llm_cost_guardian import CostTracker, TrackedAnthropic

tracker = CostTracker()
client = TrackedAnthropic(Anthropic(), tracker)

response = client.messages.create(
    model="claude-sonnet-4-20250514",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello!"}],
)

Budget Policies

Stack multiple policies for layered protection:

from llm_cost_guardian import (
    BudgetManager,
    HardCapPolicy,
    SoftWarningPolicy,
    SlidingWindowPolicy,
    CostTracker,
    TrackedOpenAI,
)

tracker = CostTracker()
budget = BudgetManager(
    on_warn=lambda result: print(f"WARNING: {result.message}")
)
budget.add(SoftWarningPolicy(warning_usd=1.00))       # warn at $1
budget.add(HardCapPolicy(limit_usd=5.00))              # block at $5
budget.add(SlidingWindowPolicy(                         # $0.50/hour max
    limit_usd=0.50,
    window_seconds=3600,
))

# Attach to a client
client = TrackedOpenAI(OpenAI(), tracker, budget)
# Budget is enforced automatically before each API call

Exporting Data

from llm_cost_guardian import to_json, to_csv, to_prometheus, save_json

# JSON string
print(to_json(tracker))

# CSV string
print(to_csv(tracker))

# Prometheus metrics
print(to_prometheus(tracker))

# Save to file
save_json(tracker, "usage_report.json")

CLI Usage

# List supported models and pricing
llm-cost-guardian models
llm-cost-guardian models --provider openai --json-output

# Estimate cost for a specific call
llm-cost-guardian estimate gpt-4o --input-tokens 10000 --output-tokens 5000

# View a saved report
llm-cost-guardian report usage_report.json

Prometheus Export

Expose a /metrics endpoint for your monitoring stack:

from flask import Flask, Response
from llm_cost_guardian import CostTracker, to_prometheus

app = Flask(__name__)
tracker = CostTracker()  # shared instance

@app.route("/metrics")
def metrics():
    return Response(to_prometheus(tracker), content_type="text/plain")

Output format:

# HELP llm_cost_guardian_total_cost_usd Total cost in USD
# TYPE llm_cost_guardian_total_cost_usd gauge
llm_cost_guardian_total_cost_usd 0.01234500
# HELP llm_cost_guardian_cost_by_model_usd Cost per model in USD
# TYPE llm_cost_guardian_cost_by_model_usd gauge
llm_cost_guardian_cost_by_model_usd{model="gpt-4o"} 0.00750000

Configuration

LLM Cost Guardian supports YAML configuration files:

# llm_cost_guardian.yml
budget:
  hard_cap_usd: 10.00
  soft_warning_usd: 5.00
  sliding_window:
    limit_usd: 2.00
    window_seconds: 3600

export:
  format: json
  path: ./reports/usage.json

# Override or add custom model pricing
models:
  my-fine-tuned-model:
    provider: openai
    input_cost_per_1m: 5.00
    output_cost_per_1m: 15.00

Supported Models

Model	Provider	Input / 1M tokens	Output / 1M tokens
gpt-4o	OpenAI	$2.50	$10.00
gpt-4o-mini	OpenAI	$0.15	$0.60
gpt-4-turbo	OpenAI	$10.00	$30.00
gpt-4	OpenAI	$30.00	$60.00
gpt-3.5-turbo	OpenAI	$0.50	$1.50
o1	OpenAI	$15.00	$60.00
o1-mini	OpenAI	$3.00	$12.00
o3-mini	OpenAI	$1.10	$4.40
claude-opus-4-20250514	Anthropic	$15.00	$75.00
claude-sonnet-4-20250514	Anthropic	$3.00	$15.00
claude-3-5-sonnet-20241022	Anthropic	$3.00	$15.00
claude-3-5-haiku-20241022	Anthropic	$0.80	$4.00
claude-3-opus-20240229	Anthropic	$15.00	$75.00
claude-3-haiku-20240307	Anthropic	$0.25	$1.25
gemini-2.0-flash	Google	$0.10	$0.40
gemini-1.5-pro	Google	$1.25	$5.00
gemini-1.5-flash	Google	$0.075	$0.30

Contributing

Contributions are welcome! See CONTRIBUTING.md for guidelines.

License

MIT License - see LICENSE for details.