The Perplexity AI search integration allows users to perform real-time web searches within the AG2 framework. Follow these steps to integrate PerplexitySearchTool with AG2 Agents.

​
Configuring Your Perplexity API Key

  1. Create a Perplexity Account:

    • Visit Perplexity AI
    • Click Sign Up and create an account
    • Select Free Tier during registration

  2. Get API Key:

    • Navigate to API Settings
    • Add payment method (required)
    • Generate API key under API Keys

  3. Set Environment Variable:

    export PERPLEXITY_API_KEY="your_api_key_here"

​
Package Installation

To get started with the Perplexity Search integration in AG2, follow these steps:

Install AG2 with openai since we use OpenAI’s LLMs in our example:

pip install -U "ag2[openai]"

Note: If you have been using autogen or pyautogen, all you need to do is upgrade it using:

pip install -U "autogen[openai]"

or

pip install -U "pyautogen[openai]"

as pyautogen, autogen, and ag2 are aliases for the same PyPI package.

You’re all set! Now you can start using Perplexity AI Search with AG2.

​
Implementation

​
Imports

import os
import autogen
from autogen import AssistantAgent, UserProxyAgent, LLMConfig
from autogen.tools.experimental import PerplexitySearchTool

​
Agent Configuration

Configure an assistant agent and user proxy to be used for LLM recommendation and execution respectively.

llm_config = LLMConfig(api_type="openai", model="gpt-4o-mini")

assistant = AssistantAgent(
    name="assistant",
    llm_config=llm_config,
)

user_proxy = UserProxyAgent(
    name="user_proxy",
    human_input_mode="NEVER"
)

​
Tool Setup

perplexity_search_tool = PerplexitySearchTool(
    api_key=os.getenv("PERPLEXITY_API_KEY"),
    max_tokens=1000
)

# Register the tool for LLM recommendation and execution.
perplexity_search_tool.register_for_llm(assistant)
perplexity_search_tool.register_for_execution(user_proxy)

​
Usage Example

With the setup complete, you can now use the assistant to fetch live web search results.

response = user_proxy.initiate_chat(
    recipient=assistant,
    message="What is AG2?",
    max_turns=2,
)

​
Output

user_proxy (to assistant):

What is AG2?

--------------------------------------------------------------------------------
assistant (to user_proxy):

***** Suggested tool call (call_kOW9AP9YmrCJbxhnnqiNjXVz): perplexity-search *****
Arguments:
{"query":"What is AG2?"}
**********************************************************************************

--------------------------------------------------------------------------------

>>>>>>>> EXECUTING FUNCTION perplexity-search...
Call ID: call_kOW9AP9YmrCJbxhnnqiNjXVz
Input arguments: {'query': 'What is AG2?'}
user_proxy (to assistant):

***** Response from calling tool (call_kOW9AP9YmrCJbxhnnqiNjXVz) *****
{"content":"AG2 is an open-source framework for building AI-powered agents, originally a community-driven fork of the AutoGen project. It is led by Dr. Chi Wang of Google DeepMind and Dr. Qingyun Wu of Penn State University. AG2 focuses on **multi-agent systems**, enabling autonomous agents to collaborate and solve complex problems across various domains such as education, enterprise workflows, and healthcare[1][3][5]. It provides tools for enhanced Large Language Model (LLM) inference and optimization, facilitating the creation of diverse applications[3][4].","citations":["https://www.forwardfuture.ai/p/agents-in-action-the-rise-of-ag2-and-the-future-of-intelligent-agents","https://www.youtube.com/watch?v=UItthyY71UU","https://docs.ag2.ai/docs/home/home","https://ag2.airt.ai/latest/docs/home/home/","https://cloud.google.com/vertex-ai/generative-ai/docs/agent-engine/develop/ag2","https://atmosphericg2.com/homepage/products/ag2data/","https://atmosphericg2.com/ag2trader/"]}
**********************************************************************

--------------------------------------------------------------------------------
assistant (to user_proxy):

AG2 is an open-source framework for building AI-powered agents, initially a community-driven fork of the AutoGen project. It is led by researchers Dr. Chi Wang of Google DeepMind and Dr. Qingyun Wu of Penn State University. AG2 emphasizes **multi-agent systems**, allowing autonomous agents to collaborate and tackle complex problems in various fields, including education, enterprise workflows, and healthcare.

The framework provides tools for improved Large Language Model (LLM) inference and optimization, facilitating the development of diverse applications.

For more detailed information, you can explore the following resources:
- [Forward Future Article](https://www.forwardfuture.ai/p/agents-in-action-the-rise-of-ag2-and-the-future-of-intelligent-agents)
- [Official Documentation](https://docs.ag2.ai/docs/home/home)

TERMINATE

--------------------------------------------------------------------------------

>>>>>>>> TERMINATING RUN (2ab104c4-cfe6-4ce7-88e3-fb83b9623a7a): Maximum turns (2) reached

Process finished with exit code 0

​
Key Considerations

​
Cost Management

Free Tier Limits:

​
Performance Optimizations

Best Practices:

  • Start with smaller response lengths by setting the max_tokens parameter.
  • Filter search domains with search_domain_filter parameter to restrict the search results to specific domains.
optimized_tool = PerplexitySearchTool(
    api_key=os.getenv("PERPLEXITY_API_KEY"),
    max_tokens=1000,
    search_domain_filter=["arxiv.org", "towardsdatascience.com"],
)
Edit this page