Inclusive Code Review Language Analyzer

A Chrome/Edge browser extension that analyzes text in real-time on GitHub and Azure DevOps to suggest more inclusive language in code reviews. Uses both rule-based suggestions and a custom ONNX ML model for sentiment analysis, with optional OpenAI integration for enhanced suggestions.

How This Extension Works

The extension operates through a multi-layer processing pipeline:

1. **Content Scripts** inject into GitHub and Azure DevOps pages and capture text input

2. **Background Service Worker** orchestrates validation using imported modules

3. **Rule-Based Suggestions** check text against a curated dictionary of inclusive language replacements

4. **ONNX ML Model** runs sentiment analysis entirely in-browser (no server calls)

5. **Optional OpenAI Integration** provides enhanced rewrite suggestions via Azure endpoint

Architecture Overview

Core Components

**Packed Modules** (`src-packed/`)

**Extension Code** (`src/`)

**ML Model** (`assets/model.onnx`)

Setup Instructions

Initial Setup

1. Install dependencies:

```bash

npm install

```

2. Create `src-packed/secrets.js` for API keys:

```javascript

// For local development without OpenAI, use empty strings

export const OPENAI_KEY = "";

export const OPENAI_ENDPOINT = "";

export const APP_INSIGHTS_KEY = "";

```

3. Build the extension:

```bash

npx xt-build

```

4. Install in browser:

- Navigate to `edge://extensions/` or `chrome://extensions/`

- Enable "Developer Mode"

- Drag `release.zip` into the extensions page

Development Commands

Adding New Inclusive Language Suggestions

Step 1: Update Suggestions Dictionary

Edit `src-packed/suggestions.js` to add new replacement patterns:

```javascript

const allReplacements = {

// Socially charged terms

"blacklist": ["denylist", "exclusion list", "block list"],

"whitelist": ["allowlist", "inclusion list", "safe list"],

// Gendered terms

"guys": ["folks", "friends", "people", "everyone"],

"man-hours": ["person-hours", "work-hours", "engineering time"],

// Ableist terms

"sanity check": ["confidence check", "coherence check", "rationality check"],

"cripple": ["impair", "disable", "hinder"],

// Add your new terms here:

"your-term": ["alternative1", "alternative2", "alternative3"]

};

```

**Categories:**

Step 2: Add Tests

Create corresponding tests in `test/suggestions.js`:

```javascript

it(`Replace "your-term"`, async function() {

const text = 'This sentence contains your-term in it.';

const matches = [];

await getMatches(ort, text, matches);

expect(matches.length).equal(1);

expect(matches[0].shortMessage).equal('Avoid "your-term"');

expect(matches[0].replacements).deep.equal([

{value: 'alternative1'},

{value: 'alternative2'},

{value: 'alternative3'}

]);

});

```

Step 3: Rebuild and Test

```bash

npx xt-test # Run tests to verify

npx xt-build # Build updated extension

```

Step 4: Manual Testing

1. Reload the extension in your browser

2. Navigate to GitHub or Azure DevOps

3. Start typing in a comment or review text area

4. Verify that your new suggestion appears when the term is used

Updating the ONNX ML Model

Prerequisites

Step 1: Replace Model File

```bash

Backup existing model

cp assets/model.onnx assets/model.onnx.backup

Add your new model

cp /path/to/your/new-model.onnx assets/model.onnx

```

Step 2: Update Inference Logic (if needed)

Edit `src-packed/textAnalytics.js` if your model has different:

```javascript

// Example: Adjusting sentiment threshold

const sentimentScore = results.sentiment_scores[0];

const isNegative = sentimentScore > 0.6; // Adjust threshold here

```

Step 3: Test Model Integration

Create test cases in `test/textAnalytics.js`:

```javascript

it('New model correctly identifies negative sentiment', async function() {

const text = 'This is an extremely harsh and critical review.';

const score = await analyzeSentiment(ort, text);

expect(score).to.be.above(0.6);

});

it('New model correctly identifies positive sentiment', async function() {

const text = 'This is excellent work, well done!';

const score = await analyzeSentiment(ort, text);

expect(score).to.be.below(0.6);

});

```

Step 4: Validate with Various Text Samples

```bash

npx xt-test # Run all tests

```

Test with diverse inputs:

Step 5: Performance Testing

Monitor model performance:

Debugging Guide

Debugging Content Scripts

1. Navigate to target page (GitHub/Azure DevOps)

2. Open browser DevTools (F12)

3. Console logs from content scripts appear in page context

4. Use "Sources" tab to set breakpoints in injected scripts

Debugging Background Service Worker

1. Navigate to `edge://extensions/` or `chrome://extensions/`

2. Find your extension

3. Click "service worker" link

4. DevTools opens with access to background script logs

Debugging with VS Code

1. Set breakpoints in test files or packed modules

2. Run tests via VS Code debugger:

- Open Debug panel

- Select "Mocha Tests" configuration

- Press F5

Common Issues

**ONNX model not loading:**

**Suggestions not appearing:**

**OpenAI integration failing:**

Integration with GitHub Copilot

This extension complements GitHub Copilot by:

1. **Post-generation review** - Analyzes code review comments after Copilot generates them

2. **Real-time feedback** - Provides immediate suggestions as you type reviews

3. **Team consistency** - Enforces inclusive language standards across all review authors

4. **Educational tool** - Teaches developers about inclusive terminology

Workflow Integration

1. Copilot suggests code or review comments

2. Developer types or accepts suggestions

3. Extension analyzes text in real-time

4. Inclusive language alternatives appear as inline suggestions

5. Developer can accept suggestions with one click

Packaging for Distribution

Build Release Package

```bash

npx xt-build

```

This creates `release.zip` containing:

Side-Loading for Testing

1. Open `edge://extensions/` or `chrome://extensions/`

2. Enable "Developer Mode" toggle

3. Drag `release.zip` onto the page

4. Extension auto-installs

Publishing to Store

For Chrome Web Store or Edge Add-ons:

1. Prepare store assets (screenshots, descriptions)

2. Submit `release.zip` via developer dashboard

3. Complete store-specific review process

Technical Constraints

**Runtime Requirements:**

**Dependencies:**

**Limitations:**

Privacy & Security

License & Attribution

Built on LanguageTool foundation (see LICENSE file in repository). Extension code uses Manifest V3 architecture with custom ONNX integration.