AG2 Perplexity Search Agent

Integration of Perplexity AI real-time web search within the AG2 (AutoGen) multi-agent framework. Enables AI agents to perform live web searches and incorporate current information into collaborative workflows.

Overview

The AG2 framework (formerly AutoGen) provides a PerplexitySearchTool that enables AI agents to perform real-time web searches within multi-agent conversations. This allows agents to access current information during collaborative problem-solving.

How It Works

1. Create an AssistantAgent with LLM configuration
2. Create a UserProxyAgent for tool execution
3. Instantiate PerplexitySearchTool with API key and token limits
4. Register tool for LLM recommendations (assistant) and execution (user proxy)
5. Agents automatically invoke search during conversations when needed

Key Features

• Real-time web search within multi-agent workflows
• Automatic tool invocation based on conversation context
• Configurable max_tokens per search query
• Integration with AG2's multi-turn conversation system
• Compatible with any LLM backend supported by AG2

Setup

pip install -U "ag2[openai]"
export PERPLEXITY_API_KEY=your_key

Code Example

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

tool = PerplexitySearchTool(api_key="pplx-...", max_tokens=1024)
tool.register_for_llm(assistant)
tool.register_for_execution(user_proxy)

Ideal for research agents, fact-checking workflows, and data-gathering pipelines.