GoonBuilding a Company Research Agent
This article provides a step-by-step guide on building and evaluating a company research agent. The agent takes in company information and schema to research and uses Tavily to do the research.
Introduction
The agent is designed to perform open-ended research on companies and output a structured schema. The research process involves generating search queries, performing web research, and extracting relevant information. The agent uses a default schema if none is provided by the user.
The Research Process
The research process starts with the user providing a company name and optionally a schema. The agent generates search queries based on the company name and performs web research using Tavily. The results are then used to extract relevant information and populate the schema.
Evaluation
The agent's performance is evaluated using a test suite and lsmith. The evaluation process involves defining a Target function that takes in inputs and outputs a score. The score is then used to determine the agent's performance.
Example Use Case
An example use case is provided to demonstrate how the agent can be used to research a company. The agent is able to correctly identify the company's description, CEO, and funding information.
Images
This is the image at 11 seconds
This is the image at 91 seconds
This is the image at 713 seconds
This is the image at 841 seconds
This is the image at 940 seconds
Conclusion
The company research agent is a powerful tool for performing open-ended research on companies. The agent's performance can be evaluated using a test suite and lsmith, and the results can be used to improve the agent's performance. With the ability to generate search queries, perform web research, and extract relevant information, the company research agent is a valuable resource for anyone looking to research companies.