Skip to main content
Open In ColabOpen on GitHub

ADS4GPTs

Integrate AI native advertising into your Agentic application.

Overview

This notebook outlines how to use the ADS4GPTs Tools and Toolkit in LangChain directly. In your LangGraph application though you will most likely use our prebuilt LangGraph agents.

Setup

Install ADS4GPTs Package

Install the ADS4GPTs package using pip.

# Install ADS4GPTs Package
# Install the ADS4GPTs package using pip
!pip install ads4gpts-langchain

Set up the environment variables for API authentication (Obtain API Key).

# Setup Environment Variables
# Prompt the user to enter their ADS4GPTs API key securely
if not os.environ.get("ADS4GPTS_API_KEY"):
os.environ["ADS4GPTS_API_KEY"] = getpass("Enter your ADS4GPTS API key: ")

Instantiation

Import the necessary libraries, including ADS4GPTs tools and toolkit.

Initialize the ADS4GPTs tools such as Ads4gptsInlineSponsoredResponseTool. We are going to work with one tool because the process is the same for every other tool we provide.

# Import Required Libraries

import os
from getpass import getpass

from ads4gpts_langchain import Ads4gptsInlineSponsoredResponseTool, Ads4gptsToolkit
# Initialize ADS4GPTs Tools
# Initialize the Ads4gptsInlineSponsoredResponseTool
inline_sponsored_response_tool = Ads4gptsInlineSponsoredResponseTool(
ads4gpts_api_key=os.environ["ADS4GPTS_API_KEY"],
)

Toolkit Instantiation

Initialize the Ads4gptsToolkit with the required parameters.

# Toolkit Initialization
# Initialize the Ads4gptsToolkit with the required parameters
toolkit = Ads4gptsToolkit(
ads4gpts_api_key=os.environ["ADS4GPTS_API_KEY"],
)

# Retrieve tools from the toolkit
tools = toolkit.get_tools()

# Print the initialized tools
for tool in tools:
print(f"Initialized tool: {tool.__class__.__name__}")
Initialized tool: Ads4gptsInlineSponsoredResponseTool
Initialized tool: Ads4gptsSuggestedPromptTool

Invocation

Run the ADS4GPTs tools with sample inputs and display the results.

# Run ADS4GPTs Tools
# Sample input data for the tools
sample_input = {
"id": "test_id",
"user_gender": "female",
"user_age": "25-34",
"user_persona": "test_persona",
"ad_recommendation": "test_recommendation",
"undesired_ads": "test_undesired_ads",
"context": "test_context",
"num_ads": 1,
"style": "neutral",
}

# Run Ads4gptsInlineSponsoredResponseTool
inline_sponsored_response_result = inline_sponsored_response_tool._run(
**sample_input, ad_format="INLINE_SPONSORED_RESPONSE"
)
print("Inline Sponsored Response Result:", inline_sponsored_response_result)
Inline Sponsored Response Result: {'ad_text': '<- Promoted Content ->\n\nLearn the sartorial ways and get your handmade tailored suit by the masters themselves with Bespoke Tailors. [Subscribe now](https://youtube.com/@bespoketailorsdubai?si=9iH587ujoWKkueFa)\n\n<->'}

Async Run ADS4GPTs Tools

Run the ADS4GPTs tools asynchronously with sample inputs and display the results.

import asyncio


# Define an async function to run the tools asynchronously
async def run_ads4gpts_tools_async():
# Run Ads4gptsInlineSponsoredResponseTool asynchronously
inline_sponsored_response_result = await inline_sponsored_response_tool._arun(
**sample_input, ad_format="INLINE_SPONSORED_RESPONSE"
)
print("Async Inline Sponsored Response Result:", inline_sponsored_response_result)
Async Inline Sponsored Response Result: {'ad_text': '<- Promoted Content ->\n\nGet the best tailoring content from Jonathan Farley. Learn to tie 100 knots and more! [Subscribe now](https://www.youtube.com/channel/UCx5hk4LN3p02jcUt3j_cexQ)\n\n<->'}

Toolkit Invocation

Use the Ads4gptsToolkit to get and run tools.

# Sample input data for the tools
sample_input = {
"id": "test_id",
"user_gender": "female",
"user_age": "25-34",
"user_persona": "test_persona",
"ad_recommendation": "test_recommendation",
"undesired_ads": "test_undesired_ads",
"context": "test_context",
"num_ads": 1,
"style": "neutral",
}

# Run one tool and print the result
tool = tools[0]
result = tool._run(**sample_input)
print(f"Result from {tool.__class__.__name__}:", result)


# Define an async function to run the tools asynchronously
async def run_toolkit_tools_async():
result = await tool._arun(**sample_input)
print(f"Async result from {tool.__class__.__name__}:", result)


# Execute the async function
await run_toolkit_tools_async()
Result from Ads4gptsInlineSponsoredResponseTool: {'ad_text': '<- Promoted Content ->\n\nLearn the sartorial ways and get your handmade tailored suit by the masters themselves with Bespoke Tailors. [Subscribe now](https://youtube.com/@bespoketailorsdubai?si=9iH587ujoWKkueFa)\n\n<->'}
Async result from Ads4gptsInlineSponsoredResponseTool: {'ad_text': '<- Promoted Content ->\n\nGet the best tailoring content from Jonathan Farley. Learn to tie 100 knots and more! [Subscribe now](https://www.youtube.com/channel/UCx5hk4LN3p02jcUt3j_cexQ)\n\n<->'}

Chaining

if not os.environ.get("OPENAI_API_KEY"):
os.environ["OPENAI_API_KEY"] = getpass("Enter your OPENAI_API_KEY API key: ")
import os

from langchain_openai import ChatOpenAI

openai_model = ChatOpenAI(model="gpt-4o", openai_api_key=os.environ["OPENAI_API_KEY"])
model = openai_model.bind_tools(tools)
model_response = model.invoke(
"Get me an ad for clothing. I am a young man looking to go out with friends."
)
print("Tool call:", model_response)
API Reference:ChatOpenAI
Tool call: content='' additional_kwargs={'tool_calls': [{'id': 'call_XLR5UjF8JhylVHvrk9mTjhj8', 'function': {'arguments': '{"id":"unique_user_id_001","user_gender":"male","user_age":"18-24","ad_recommendation":"Stylish and trendy clothing suitable for young men going out with friends.","undesired_ads":"formal wear, women\'s clothing, children\'s clothing","context":"A young man looking for clothing to go out with friends","num_ads":1,"style":"youthful and trendy","ad_format":"INLINE_SPONSORED_RESPONSE"}', 'name': 'ads4gpts_inline_sponsored_response'}, 'type': 'function'}], 'refusal': None} response_metadata={'token_usage': {'completion_tokens': 106, 'prompt_tokens': 1070, 'total_tokens': 1176, 'completion_tokens_details': {'accepted_prediction_tokens': 0, 'audio_tokens': 0, 'reasoning_tokens': 0, 'rejected_prediction_tokens': 0}, 'prompt_tokens_details': {'audio_tokens': 0, 'cached_tokens': 1024}}, 'model_name': 'gpt-4o-2024-08-06', 'system_fingerprint': 'fp_eb9dce56a8', 'finish_reason': 'tool_calls', 'logprobs': None} id='run-e3e64b4b-4505-4a71-bf02-a8d77bb68eee-0' tool_calls=[{'name': 'ads4gpts_inline_sponsored_response', 'args': {'id': 'unique_user_id_001', 'user_gender': 'male', 'user_age': '18-24', 'ad_recommendation': 'Stylish and trendy clothing suitable for young men going out with friends.', 'undesired_ads': "formal wear, women's clothing, children's clothing", 'context': 'A young man looking for clothing to go out with friends', 'num_ads': 1, 'style': 'youthful and trendy', 'ad_format': 'INLINE_SPONSORED_RESPONSE'}, 'id': 'call_XLR5UjF8JhylVHvrk9mTjhj8', 'type': 'tool_call'}] usage_metadata={'input_tokens': 1070, 'output_tokens': 106, 'total_tokens': 1176, 'input_token_details': {'audio': 0, 'cache_read': 1024}, 'output_token_details': {'audio': 0, 'reasoning': 0}}

API reference

You can learn more about ADS4GPTs and the tools at our GitHub


Was this page helpful?