Skip to main content
actguard patches google.generativeai.GenerativeModel.generate_content and its async counterpart.

​
Requirements

RequirementVersion
google-generativeai SDKany recent release
Python≥ 3.9
pip install google-generativeai

​
What gets patched

google.generativeai.GenerativeModel.generate_content        → actguard wrapper (sync)
google.generativeai.GenerativeModel.generate_content_async  → actguard wrapper (async)

​
Non-streaming

import google.generativeai as genai
from actguard import BudgetGuard

genai.configure(api_key="YOUR_API_KEY")
model = genai.GenerativeModel("gemini-1.5-pro")

with BudgetGuard(user_id="alice", usd_limit=0.10) as guard:
    response = model.generate_content("Explain quantum computing.")
    print(response.text)

print(f"${guard.usd_used:.6f}  ({guard.tokens_used} tokens)")

​
Streaming

For streaming, actguard reads usage_metadata from the first chunk that carries it: 
with BudgetGuard(user_id="alice", usd_limit=0.10) as guard:
    for chunk in model.generate_content("Write a poem.", stream=True):
        print(chunk.text, end="", flush=True)

print(f"\n${guard.usd_used:.6f}")

​
Async client

import asyncio
import google.generativeai as genai
from actguard import BudgetGuard

genai.configure(api_key="YOUR_API_KEY")
model = genai.GenerativeModel("gemini-1.5-pro")

async def main():
    async with BudgetGuard(user_id="alice", usd_limit=0.10) as guard:
        response = await model.generate_content_async("Hello!")
    print(f"${guard.usd_used:.6f}")

asyncio.run(main())

​
Model name normalisation

The Google SDK prefixes model names with models/ (e.g. models/gemini-1.5-pro). actguard strips this prefix before looking up the pricing table, so gemini-1.5-pro is the key used for cost calculation.