shekel 0.1.1

LLM cost tracking and budget enforcement. Zero config — with budget(max_usd=1.00): run_agent(). Works with LangGraph, CrewAI, raw OpenAI/Anthropic.

with budget(max_usd=1.00):
    run_my_agent()  # raises BudgetExceededError if spend exceeds $1.00

I spent $47 debugging a LangGraph retry loop. The agent kept failing, LangGraph kept retrying, and OpenAI kept charging — all while I slept. I built shekel so you don't have to learn that lesson yourself.

---

Install

pip install shekel[openai]      # OpenAI only
pip install shekel[anthropic]   # Anthropic only
pip install shekel[all]         # Both SDKs

Advanced: pip install shekel installs the core with no SDK deps (track-only mode).

---

Quick start

from shekel import budget, BudgetExceededError
import openai

client = openai.OpenAI()

try:
    with budget(max_usd=1.00, warn_at=0.8) as b:
        response = client.chat.completions.create(
            model="gpt-4o",
            messages=[{"role": "user", "content": "Hello!"}],
        )
    print(f"Spent: ${b.spent:.4f} of ${b.limit:.2f}")
except BudgetExceededError as e:
    print(e)

---

Track-only mode

No max_usd = track spend without enforcing a limit. Great for profiling agents.

with budget() as b:
    run_my_agent()

print(f"That run cost: ${b.spent:.4f}")

---

How it works

• Monkey-patching: When you enter a budget() context, shekel wraps openai.ChatCompletions.create and anthropic.Messages.create at the class level. Your code calls the real SDK — shekel intercepts the response, reads the token counts, and calculates cost. On context exit, original methods are restored.
• ContextVar isolation: Each budget() context tracks its own spend using Python's contextvars.ContextVar. Two concurrent agent runs never share a budget counter, even in async or multi-threaded code.
• Zero config: No API keys, no external services, no config files. pip install + with budget(...) is all you need.

---

Model support

Model	Input (per 1k)	Output (per 1k)
gpt-4o	$0.00250	$0.01000
gpt-4o-mini	$0.000150	$0.000600
o1	$0.01500	$0.06000
o1-mini	$0.00300	$0.01200
gpt-3.5-turbo	$0.000500	$0.001500
claude-3-5-sonnet-20241022	$0.00300	$0.01500
claude-3-haiku-20240307	$0.000250	$0.001250
claude-3-opus-20240229	$0.01500	$0.07500
gemini-1.5-flash	$0.0000750	$0.000300
gemini-1.5-pro	$0.00125	$0.00500

Model not listed? Pass a price override:

with budget(max_usd=1.00, price_per_1k_tokens={"input": 0.001, "output": 0.003}):
    run_my_agent()

---

Works with LangGraph, CrewAI, and any framework

shekel is framework-agnostic. It intercepts at the SDK level, so it works with anything that calls OpenAI or Anthropic under the hood:

# LangGraph
with budget(max_usd=2.00, warn_at=0.8) as b:
    result = langgraph_app.invoke({"input": "..."})
print(f"Graph run cost: ${b.spent:.4f}")

# CrewAI
with budget(max_usd=5.00) as b:
    crew.kickoff()

# Raw SDK
with budget(max_usd=0.50) as b:
    for _ in range(100):
        client.chat.completions.create(...)  # stops when budget hit

---

License

MIT