ChromeFactCheck

ChromeFactCheck is a Chrome extension plus ASP.NET Core backend for fast claim verification. It lets users select text on any page, request a fact check, view a short in-page summary, and open a full report on demand.

What It Does

• Select text on a page and trigger fact checking.
• Show a short 2-3 sentence summary in a popover.
• Show loading state with spinner while analysis is running.
• Open detailed output only when user clicks Read more....
• Support configurable LLM provider, endpoint, model, API key, and answer language.

Repository Structure

• extension/ Chrome Manifest V3 extension
• src/ChromeFactCheck.Api/ ASP.NET Core backend (net9.0)
• ChromeFactCheck.sln solution file

Prerequisites

• .NET SDK 9.0+
• Google Chrome (or Chromium with extension support)
• An LLM API key for configured provider

Quick Start

1) Run backend

From repo root:

dotnet run --project src/ChromeFactCheck.Api --urls http://localhost:5053

Health check:

curl http://localhost:5053/api/health

2) Load extension

1. Open chrome://extensions
2. Enable Developer mode
3. Click Load unpacked
4. Select the extension folder

3) Configure settings

Use the extension popup (toolbar icon) or options page:

• Backend URL: http://localhost:5053 for local development
• Provider: openai, azure_openai, or custom
• Endpoint: required for azure_openai and custom
• Model: for example gpt-4.1-mini
• API Key: stored in chrome.storage.local
• Answer Language: auto or explicit locale (en-US, sv-SE, etc.)

User Flow

1. User highlights text on a web page.
2. Extension opens the small fact-check prompt/popover.
3. Backend receives selection + metadata and runs analysis.
4. Extension shows short summary with estimated true/false probability.
5. User can click Read more... to open full report page.

API Reference

Endpoints

• GET /api/health
• POST /api/fact-check/selection

Request contract (POST /api/fact-check/selection)

{
  "selectedText": "string",
  "pageUrl": "string",
  "pageTitle": "string",
  "locale": "en-US",
  "userPreferences": {
    "provider": "openai|azure_openai|custom",
    "endpoint": "string",
    "model": "string",
    "apiKeyPresent": true,
    "strictness": "low|medium|high",
    "answerLanguage": "auto|en-US|sv-SE",
    "maxSources": 5,
    "trustedDomains": ["wikipedia.org"],
    "blockedDomains": ["example.com"]
  }
}

Provider behavior

• openai: endpoint optional, defaults to https://api.openai.com
• azure_openai: full chat-completions endpoint required (include api-version)
• custom: OpenAI-compatible endpoint required

Secrets and auth

• Extension sends key in X-Llm-Api-Key header.
• Backend uses it per request and should not persist it.

Development Workflow

Build

dotnet build ChromeFactCheck.sln

Typical local loop

1. Edit API and/or extension files.
2. Build backend with dotnet build.
3. Reload extension in chrome://extensions.
4. Re-test by selecting text on a page.

Troubleshooting

• Could not load file 'content.js' ... UTF-8 encoded
  • Re-save extension/content.js as UTF-8 (without BOM recommended).
• Cannot find menu item with id ...
  • Reload extension and ensure context menu is created on startup/install.
• Popup not updating
  • Reload extension and check chrome://extensions service worker logs.
• Backend not reachable
  • Verify API URL and run curl http://localhost:5053/api/health.

Security Notes

• Do not commit real API keys.
• Do not log secrets in API or extension logs.
• Keep provider endpoints explicit in non-default setups.

Roadmap

• Full-page auto-check mode (beyond selected text)
• Retrieval-assisted second pass with source adjudication
• Better citation display and conflict handling between sources