This repository provides a tool set and a deployment guide showcasing the usage of Vertex AI / Gemini in a complete Software Development Life Cycle process, using Gitlab as the main Documentation and Source Code repository. The solution showcases the User Story document evaluation, Test Case document creation based on User Story document (the screen prototype is also evaluated), Test Script creation based on Test Case document, Bulk Test Data creation, and also Code Changes review submitted to the repository.
This repository is organized in subprojects, each of them with its own detailed documentation on the setup and deployment.
Prompts for each of the components are provided in English and Brazilian Portuguese, and may be customized as needed.

WARNING
PII: Testing data model may not generate valid PII data like SSN. (Sometimes it may even fail to generate any data at all, even invalid ones).

The organization of the components on the platform is detailed below:

As the diagram shows, the solution relies heavily on Google Cloud Run. There is also the presence of Secret Manager to hold few of the secrets necessary for integration between multiple components. Of course, we have Vertex AI APIs as well and a BigQuery to hold metrics and rating.

Each of the described components is built into its own container. Details for each of them are provided in their respective folders.
There are two main integration points to Gitlab, the wiki-interface component and the pull-request component.
The wiki-interface component is basically some HTML Form + Javascript that reads the Wiki Page list from Gitlab, and writes back to Gitlab the results. If you wish to integrate to any other system, you can use this component as an starting point.
The wiki-interface is also responsible for storing document generation metrics as well as result rating to BigQuery.
The pull-request component is a Rest service that receives Webhook calls and integrates to the Repository to read data and write feedback. As-is, this component supports Gitlab and Github.

The proposed solution works with three different personas.

• Business Analyst
  Business Analyst writes user story and submits to Gitlab.
  Business Analyst uses wiki-integration to execute an evaluation through evaluator.
  evaluator invokes Vertex AI with our prompt to get the results.
  wiki-integration gets the result from evaluator and submits to Gitlab Wiki.

• Test/QA Analyst
  Test/QA Analyst reviews User Story document and submits to tc-generator using wiki-integration. wiki-integration detects if the input document has any images and sends them to image-processor. image-processor uses Vertex AI to generate Image/Screen details. wiki-interface adjusts the input document replacing the images with its details in text. tc-generator uses Vertex AI to generate a Test Case document based on User Story document and image details.
  wiki-integration gets the results and submits to Gitlab Wiki.
  Test/QA Analyst reviews Test Case document and using wiki-integration submits it to one of our script components.
  script component invokes Vertex AI to generate a Test Script in Cypress or Playwright.
  wiki-integration gets the result from script component and submits to Gitlab Wiki.
  Test/QA Analyst reviews and adapts Test Script using Duet AI and saves final version. Test/QA Analyst submits sample CSV data to test-data using wiki-integration. test-data uses Vertex AI to generate Bulk Test Data based on sample provided.

• Software Developer
  Software Developer reviews User Story document and writes code with help of Duet AI.
  Software Developer opens a Pull/Merge Request to Gitlab.
  Gitlab invokes our Webhook to evaluate the change proposed using Vertex AI.
  Webhook comments back to the proposed review with change evaluation.

The proposed deployment uses Gitlab as its main repository for holding both, Wiki Pages and Source Code.
This solution doesn't plan to cover Gitlab setup and provisioning. There are multiple ways to achieve that. To mention a few:

• Google Cloud Marketplace
• Gitlab Official Documentation
• click-to-deploy
• Install Gitlab CE in a single Compute Engine instance

Each component of the solution has a deployment/installation guide on its subdirectory.

IMPORTANT:

• There are resources that are shared between multiple components, like Secret Manager entries.
• Pay attention not to recreate those shared components and break the solution.

SUGGESTED COMPONENTS INSTALLATION ORDER:

1. evaluator
2. image-processor
3. script-cypress
4. script-playwright
5. tc-generator
6. test-data
7. wiki-interface
8. pull-request-evaluator

Almost each component of the solution has a README.md file with details about it.
There is also a TLDR.md file on each component for those that are in a hurry.

The integration of Gitlab Wiki with your solution is done through wiki-interface component. To enable it, you must go into your project settings and enable External wiki integration.

The URL must be in the following format: https://cloud-run-url/webui/:project-id .
You can get the your Gitlab Project Id going to Settings, General:

To setup Gitlab's Webhook, go to Settings, Webhooks, and add a new Webhook:

The URL must be in the following format: https://cloud-run-url/webhook .
The Secret Token must match the token that you are using on Secret Manager.

Select Comments and Merge Request events on the checkboxes:

Note: Depending on Gitlab Version or Github Usage, if Content Type is present it should be set to application/json for the webhook call.

Check CONTRIBUTING.md file.