Skip to content

Running the Agent

⚠️ Before you begin:
Make sure you’ve completed the Prework and Setup sections.
Your .env must include a Tavily API key, your redis stack must be running, and you either need an OpenAI API key or your local Ollama model must be running.

Step 1: Verify Environment

  1. Navigate to the working directory:

    cd beeai-workshop/beeai_fw_tavily_redis

  2. Sync dependencies:

    uv sync

Step 2: Prepare to Run the Agent

  1. Navigate to the following file:

    src/agent.py

  2. This file contains fill-in-the-blank sections. Follow the in-line comments to complete the logic.

    • Uncomment the language model you intend to use:
      • Use OpenAI if your API key is in .env
      • Use Ollama if you're running granite3.3:8b locally

    If you are using Ollama, run this command in a new terminal:

    ollama run granite3.3:8b

    If you are using OpenAI, your .env file must include your API key:

    OPENAI_API_KEY=your-key-here
    • Be sure to add the tools you implemented earlier (Tavily MCP Tool and Redis Retriever).
    • You can read more about tool usage in the BeeAI Framework here:
      👉 Tool Usage Documentation

Step 3: Build the Agent

  1. Read the instructions, role and notes variables in the file. They all fit into the system prompt to guide the agent's behavior. It defines how the assistant should behave and respond to users.

  2. Add the custom tools that we built in the previous sections to the agent to enable it with internet search and RAG capabilities.

  3. Explore the conditional_requirements section — this allows you to control how and when the agent should use a specific tool giving you more reliability.

  4. Uncomment the middlewares parameter and experiment with what is excluded. This allows you to see what the agent is doing with detailed output (maybe too much).

💡 To learn more about how the BeeAI Requirement Agent works: Requirement Agent Documentation

👉 To View the complete argument reference for conditional requirements: Conditional Requirements Reference

Step 4: Run the Agent

Execute or debug the agent script:

uv run src/agent.py

Remember to save the file with your changes before running it!

You’ll be prompted to enter a question in your terminal! You can ask your own or try one of these examples:

Sample Questions

  • Tell me about the new pilot. Who is the target audience?
  • What options do we have for vegetarians?
  • I haven't gotten my order of french fries for 3 days and we are running low. This will start to affect operations. What do I do?
  • Who do I call if there is an event in my store and I think it might turn into a PR emergency?

💡 If asking your own questions, remember:
The agent only knows what’s in its vector store documents and what the prompt instructs it to do. If you'd like to expand the agent scope to include more documents in the RAG search or more custom tools, use this workshop as a starting point and go wild!