Skip to content

Latest commit

 

History

History
65 lines (40 loc) · 2.2 KB

STEP5_AGENT_API.md

File metadata and controls

65 lines (40 loc) · 2.2 KB

Step 5 (Optional): Calling Your Kubernetes AI Agent Using an API

In this final exercise, you will learn how to interact with your AI Agent via an API call. This enables you to programmatically send queries to your AI Agent and retrieve responses. A pre-configured Python script (agent.py) is provided in the GitHub repository to simplify this process.

Prerequisites

Before proceeding, ensure you have:

  • Python 3.7 or later installed.

  • (Optional, but highly recommended) A virtual environment like conda or venv.

  • The agent.py script, available in the repository under kubernetes-walkthrough/agent.py.

  • Installed the required Python libraries: pyjwt, openai, and httpx.

    pip install pyjwt openai httpx

  • Your AI Agent's credentials:

    • Agent Key: Found in your chatbot embed code under data-chatbot-id.
    • Agent Endpoint URL: Found in your chatbot embed code (src value in the embed script).

5.1: Update agent.py

  1. Open the agent.py file in a text editor. This Python file is located in the kubernetes-walkthrough directory in this repo.
  2. Replace the placeholders with your agent's details:
  • agent_key: Replace <your-agent-endpoint-key> with your Agent Key.
  • agent_endpoint: Replace <your-agent-endpoint-url> with your Agent Endpoint URL. Ensure it ends with /api/v1/.

5.2: Test the Script

  1. Run the script:
python agent.py

  1. The script will send a query to your Agent. The default query is:

    How do I troubleshoot a CrashLoopBackOff error in Kubernetes?

  2. The agent's response will be printed to the terminal.

How the Script Works

The API Endpoint that is deployed is OpenAI API compatible. Most LLMs these days are OpenAI compatible and these Agents aren't any different.

5.3: Customize the Script

  1. Modify the messages parameter in the script to include your desired query:
messages=[{"role": "user", "content": "Your custom query here"}]
  1. Adjust the response handling as needed.

Next Steps...

Let's wrap up here...

→ Next Up: Concluding Remarks