Skip to main content

Automatically Register Review Findings to Lightning Review

Overview​

This sample demonstrates how to perform reviews of Word design documents using a workflow that follows predefined review points.

The sample reads the design document to be reviewed and the review points, then repeatedly runs agents for each combination of review point and document chapter. Extracted review findings are output to Lightning Review.

In this sample, the workflow executes the following steps in order:

  1. Retrieve the document to be reviewed
  2. Retrieve the review points
  3. Repeatedly run the review
  4. Output to Lightning Review

This allows you to observe the process of reviewing a Word document by review point while combining tool integrations and AI agents.

What this sample enables​

  • Integrate with Lightning Review and register extracted findings to Lightning Review
  • Read a review points list file and perform reviews using those points
  • Decompose a Word design document into chapters and review each chapter
  • Execute comprehensive repeated reviews for each chapter × review point combination

Intended users​

This sample is intended primarily for users who:

  • Are considering integrating Lightning Review with AI
  • Want to streamline and standardize design document reviews using AI
  • Want to experience repeated execution using workflows

Included files​

The sample includes the following example data.

tip

You can run this sample as-is to verify its behavior, and then replace the input/output files with your own artifacts to try it in a scenario closer to your actual work. If you replace the files, change the targets as described in "Procedure for replacing with your own artifacts" and follow the steps in the "Procedure" section.

TypeContentProcedure for replacing with your own artifacts
Workflow definition章と観点ごとにWord設計書をレビューしてLRに指摘を出力.yamlNot editable
Agent definitionWord設計書レビュー_LR指摘JSON生成.yamlNot editable
Input: Review points fileinputData/reviewPointList.json1. Place your review points file in JSON format in the inputData folder.
2. Specify the path in step 4 of the Procedure.
Input: Word design documentinputData/APPレイヤコンポーネント設計書.docx1. Place your design document in Word format in the inputData folder.
2. Specify the path in step 4 of the Procedure.
DocumentationThis READMENot editable

Using the sample data​

Prerequisites​

  • The structure of your review points must match the input format expected by this sample
  • The Lightning Review file to which review findings will be output must be open

Input / Output​

  • Input: Review points list file, design document in Word format
  • Output: A list of all review findings (Lightning Review format)

Procedure​

  1. Open the Lightning Review file for outputting review results.
  2. In the DC Agentiqs workflow screen, click "章と観点ごとにWord設計書をレビューしてLRに指摘を出力".
  3. On the "章と観点ごとにWord設計書をレビューしてLRに指摘を出力" screen, click "Run".
  4. A dialog will be displayed. Enter the "Review points list file path" and the "Word design document file path", then click "Run".
  5. The findings will be output to the Lightning Review file.

Applicable components​

  • Workflow: 章と観点ごとにWord設計書をレビューしてLRに指摘を出力
  • Agent: Word設計書レビュー_LR指摘JSON生成

Limitations​

  • Because this sample inputs data to the agent on a per-chapter basis, accuracy may decrease if a chapter is very large. In that case, customize the input scope to the agent (for example, use smaller sections).
  • Review points must be usable in the format expected by the workflow and agent. If the structure or contents differ, changes to the workflow will be required.

Reference: Workflow execution flow​

This sample decomposes the process into stages using a workflow, allowing you to verify a configuration that combines tools and AI agents to perform the review process in a sequence.

  1. Read the review points list file and shape it into a structure that can be looped over by point.
  2. Read the Word design document and shape it into a structure that can be looped over by chapter.
  3. For each chapter and each review point, input the data to the review agent in sequence and extract review findings
  4. For each chapter and each review point, register the review findings to Lightning Review
Last updated on