Financial_Event_Analysis.md

Financial Event Analysis

Case Description

This case study is based on PeerPlanner and presents a multi-agent collaborative example focused on analyzing financial events. Specifically, regarding the topic of "Buffett's 2023 Reduction in BYD Shares"， it demonstrates how to utilize the PEER multi-agent collaboration model within the agentUniverse framework. Additionally, it provides detailed configurations and output examples for each agent in PEER. In this case study, we utilizes the GPT-4o model by OPENAI. Prior to its usage, you need to configure the OPENAI_API_KEY in your environment variables.

Agents

Planning Agent

Reference the original code files:

• Configuration file
• Prompt file

The Planning Agent is tasked with breaking down the initial financial problem into multiple sub-problems that can be independently solved and then provided to the subsequent Executing Agent. In this particular scenario, the overarching question "Analyze the reasons for Buffett's reduction in BYD shares" can be dissected into several sub-questions, as illustrated in the diagram below.

Executing Agent

Reference the original code files:

• Configuration file
• Prompt file

In this Agent, we provide a tool namedgoogle_search_tool for searching information on Google. To utilize this tool, you should configure SERPER_API_KEY in your environment variables.

The Executing Agent is responsible for solving the sub-problems that have been broken down by the Planning Agent. In this particular case, the execution results of the Executing Agent are as follows:

Expressing Agent

Reference the original code files:

• Configuration file
• Prompt file

The Expressing Agent is responsible for summarizing all the results outputted by the Executing Agent and formulating them into an answer to the original question, adhering to the requirements specified in the prompt file. In thisinstance, the output result of the Expressing Agent is as follows:

Reviewing Agent

Reference the original code files:

• Configuration file

The Reviewing Agent is responsible for evaluating whether the answer produced by the Expressing Agent effectively addresses the original question. In this particular case, the Reviewing Agent accepted the answer provided by the Expressing Agent:

PEER Agent

info:
  name: 'demo_peer_agent'
  description: 'demo peer agent'
plan:
  planner:
    name: 'peer_planner'
    eval_threshold: 60
    retry_count: 2
    planning: 'demo_planning_agent'
    executing: 'demo_executing_agent'
    expressing: 'demo_expressing_agent'
    reviewing: 'demo_reviewing_agent'
metadata:
  type: 'AGENT'
  module: 'agentuniverse.agent.default.peer_agent.peer_agent'
  class: 'PeerAgent'

Users can configure the four Agents mentioned above into a complete PEER Agent within the agentuniverse through the peer_planner collaboration model. The configurations include:

• name: Set as peer_planner, indicating the utilization of the PEER multi-agent collaboration model.
• eval_threshold: The minimum score threshold that the Reviewing Agent must achieve to accept an answer.
• retry_count: The number of retry attempts for the PEER Agent if the Reviewing Agent rejects the answer.
• planning：The Agent responsible for the Plan part.
• executing：The Agent responsible for the Execute part.
• expressing：The Agent responsible for the Express part.
• reviewing：The Agent responsible for the Review part.

You can run the complete case in the example file.