notebook: Document an agentic AI system

[validbeck]

Pull Request Description

What and why?

sc-12466

• In accordance with our writing style guide, I renamed, edited, and restructured contents of the "AI Agent Validation with ValidMind - Banking Demo" notebook to: Document an agentic AI system
• I also added the Agentic AI template as a notebook artifact as it isn't a default out-of-the-box template: agentic_ai_template.yaml

How to test

1. Pull down this PR: gh pr checkout 472
2. Make sure you have the new added scorer for StepEfficiency in your library environment by registering a new Python kernel
3. Run notebooks/code_samples/agents/document_agentic_ai.ipynb

What needs special review?

The notebook runs end-to-end without issues in my environment, but you should check that everything looks fine to them as well.

General topic and test wording

I understand how to use the library functions, how to run tests, etc. very well at this point but I don't necessarily know why we're choosing to run the tests that we do, so please make sure that my descriptions are accurate and relevant.

Assigning AI evaluation metric scores section

Important

Please check that the following are all correct and do what we want them to do for the reasons outlined in the notebook:

• I added some information here on returning scorers related to the assignments we want to make.
• I also added the following execution layer test as it was missing as per this conversation in Fix langgraph version in the agentic demo notebook #471 (comment): StepEfficiency.py

Dependencies, breaking changes, and deployment notes

Refer to the line about StepEfficiency.py above.

Release notes

Learn how to build and document an agentic AI system with the ValidMind Library with our new notebook. Construct a LangGraph-based banking agent that selects and invokes tools in response to user requests. You'll assign AI evaluation scores to your agent, run accuracy, RAGAS, and safety tests, and log the results of your tests to the ValidMind Platform.

Document an agentic AI system

Checklist

• What and why
• Screenshots or videos (Frontend)
• How to test
• What needs special review
• Dependencies, breaking changes, and deployment notes
• Labels applied
• PR linked to Shortcut
• Unit tests added (Backend)
• Tested locally
• Documentation updated (if required)
• Environment variable additions/changes documented (if required)

[AnilSorathiya]

minor comment:
We have remove the StepEfficiency scorer from the code due to bug. Main branch doesn't have it.

Otherwise it's looks good to me. Thanks 👍

[validbeck]

@AnilSorathiya It works and passes the code quality test in the current version:

[github-actions]

PR Summary

This pull request introduces a new YAML template (agentic_ai_template.yaml) that defines comprehensive guidelines for documenting agentic AI systems. The template is structured into multiple sections including conceptual soundness, data evaluation, model evaluation, and observability and monitoring. Each section comprises detailed guidelines (with examples and hierarchical parent section references) aimed at enabling users to document features such as autonomy, reasoning, memory, risk management, regulatory compliance, and more.

Additionally, a new Jupyter Notebook (document_agentic_ai.ipynb) has been added. This notebook provides step‐by‐step instructions to build and document an agentic AI system using the ValidMind Library. It includes detailed markdown explanations, code cells for installing dependencies, initializing the ValidMind environment, building and testing agent workflows, and running validation tests. The notebook guides users to verify LLM API access via environment variable configuration, integrate banking tools, bind the tools to the agent, and finally to capture test results including AI evaluation metrics.

Significantly, a legacy notebook (langgraph_agent_simple_banking_demo.ipynb) has been removed, likely because its functionality is being replaced or superseded by the new, more comprehensive documentation template and notebook. Overall, the PR refactors the documentation and testing approach for agentic AI systems by providing a structured template and modernizing the developer guides.

Test Suggestions

• Run a YAML linter to ensure the syntax and formatting of agentic_ai_template.yaml are correct.
• Execute the new document_agentic_ai.ipynb notebook cell-by-cell to verify that all code cells run without errors.
• Perform integration tests to check that the ValidMind library correctly picks up the new template for model documentation.
• Verify that the removal of the legacy banking demo notebook does not break any external references or dependencies.