Project Instructions Step 1: Draft Your Agent Workflow Begin by drafting a diagram illustrating the interactions and data flows between the agents in your multi-agent system. Your diagram should demonstrate the sequence of operations for handling customer inquiries, inventory management, quote generation, and order fulfillment. Some recommended agents for your multi-agent system are as follows:
An orchestrator agent for handling customer inquiries and delegating tasks to different agents Answer inventory queries accurately, including deciding when to reorder supplies Generate quotes efficiently, applying bulk discounts strategically to encourage sales Finalize sales transactions, considering inventory levels and delivery timelines These agents should also have access to tools that allow them to interact with the system database and the outside world. Some recommended tools that might help you flesh out your system are:
A tool that checks inventory for different paper types A tool that gets quote history related to a customer's request A tool that checks the timeline for delivery of an item from the supplier A tool that fulfills orders by updating the system database Note that these recommendations for agents and tools are only recommendations. You may want to design the architecture of your multi-agent system differently. To understand more about the context of this project, you may want to do some reading on inventory and sales management in companies selling physical products.
You can use diagramming tools like Diagrams.net(opens in a new tab) or Mermaid(opens in a new tab) to create your diagrams.
Step 2: Review the Starter Code Carefully examine the provided starter code (project_starter.py) in your workspace. This code includes essential functionalities such as:
Initializing and managing an SQLite database Managing inventory stock levels Generating and tracking financial transactions Utility functions to estimate supplier delivery dates and current cash balance At the bottom of the project_starter.py file, you'll find a provided code stub designed to help you evaluate your agent implementation. You can use this stub to test and refine your system effectively.
Spend at least 30 minutes reviewing the starter file. Write brief descriptions for each function provided to ensure you thoroughly understand their purpose and usage within your system.
Once you complete your review, revisit your agent flow draft from Step 1. Update the tools you initially outlined, replacing hypothetical tools with tools defined using the helper functions provided in the starter code based on your newfound understanding.
Step 3: Select Your Agent Framework Decide on the agent orchestration framework you will use for this project. Your options include:
smolagents: Link to documentation(opens in a new tab) pydantic-ai: Link to documentation(opens in a new tab) npcsh: Link to documentation(opens in a new tab) Make sure you are comfortable with your chosen framework and that it aligns well with the requirements and your intended agent interactions.
Step 4: Implement Your System Using the framework you have selected, implement the agents based on the updated flow you refined in Step 2. Ensure your system follows the agent workflow diagram you drafted in step 1. Feel free to update the diagram based on changes to your approach if any hiccups arise during the implementation.
You can get started by creating different agents which accomplish each of these tasks and an orchestration agent that orchestrates the flow of control and data between the different agents. Use the helper functions provided in the starter file to aid the definition of tools for the different worker agents. Additionally, refer to the project rubric to thoroughly understand the requirements and criteria for successful completion of your system.
Step 5: Test and Evaluate Your Implementation Test your multi-agent system thoroughly using the provided dataset (quote_requests_sample.csv). Ensure:
Your agents correctly handle various customer inquiries and orders Orders are accommodated effectively to optimize inventory use and profitability The quoting agent consistently provides competitive and attractive pricing Refer to the project rubric at the end of this step to evaluate the results documented in the test_results.csv file.
Step 6: Reflect and Document After evaluating your system, prepare a clear and concise report detailing:
A comprehensive explanation of your multi-agent system Evaluation results (test_results.csv generated by the evaluation code) highlighting strengths and areas for improvement Suggestions for further improvements to the system based on the areas of improvement. Your final submission should include:
Your updated agent flow diagram Your completed implementation script Your reflective report with evaluations Ensure that you check the project rubric to understand the requirements from the different parts of the project before submitting it. You submission will be reviewed against the project rubric.