ReAct Agent (Web Researcher)

Concepts: ReAct Agents, Tool Calling, Dynamic Decision Making, Conditional Logic, Web Search, Web Scraping, SERP API, BeautifulSoup

This example demonstrates how to create a ReAct agent that can perform real-time web research by integrating a tool for Google searching and a tool for scraping individual web pages. The agent dynamically decides which tools to invoke based on the user’s query and the results of previous tool calls, showcasing advanced conditional logic and dynamic decision-making capabilities.

In this example, we ask the agent:

Look up Vellum AI and tell me about their homepage

The agent responds:

1 Vellum AI's homepage presents the company as a collaborative platform for building and shipping AI products. Here are the key highlights from their homepage:
2 
3 - **Core Offering**: Vellum is designed to help teams define, evaluate, and improve agentic AI systems. It provides tools to capture real-world feedback and continuously enhance AI implementations—all in one place.
4 - **Products**: Their suite includes orchestration, SDK, prompt engineering, evaluations, retrieval, deployment, and observability tools. These are aimed at making AI development more accessible and robust for teams of all technical backgrounds.
5 - **Collaboration**: Vellum emphasizes a purpose-built GUI and SDK for AI development, allowing users to build workflows visually or in code, with seamless sync between the two. This bridges the gap between technical and non-technical team members.
6 - **Model Playground**: The platform offers a unified space to manage, compare, and collaborate on prompts and models from various providers, whether closed, open-source, or self-hosted.
7 - **Enterprise Features**: Vellum is built for scale, offering SOC 2 and HIPAA compliance, virtual private cloud options, configurable data retention, and role-based access control.
8 - **Customer Trust**: The homepage features testimonials from companies like Drata, Redfin, and Woflow, highlighting faster AI development, easier validation of ideas, and improved reliability.
9 - **Support and Resources**: Vellum provides extensive resources, including case studies, guides, model comparisons, and a leaderboard for LLMs. They also offer on-demand AI expert support.
10 - **Backed by Top Investors**: The company is supported by Y Combinator and other notable investors.
11 
12 The homepage positions Vellum as a comprehensive, enterprise-ready platform for teams looking to build, deploy, and monitor AI systems efficiently and collaboratively[1].
13 
14 Sources:
15 1. [Vellum AI Homepage](https://www.vellum.ai/)

ReAct Agent Web Researcher Example Workflow