Summarize any article with one click using your favorite AI provider.

• One-click summaries — A "Summarize" button appears in every article. Click it to get an AI-generated summary streamed in real time below the title.

• Real-time streaming — Summaries are delivered via Server-Sent Events (SSE), so you see text appear as the AI generates it, with a live typing cursor.

• Auto-fetch full articles — When the RSS feed only contains a short excerpt, the extension automatically fetches and parses the full article from the original URL.

• 4 AI providers — Choose the one that fits your setup:

  Provider	Default Model	API Key Required
  OpenAI (ChatGPT)	gpt-4o-mini	Yes
  Anthropic (Claude)	claude-sonnet-4-6	Yes
  Google (Gemini)	gemini-2.5-flash	Yes
  Ollama	llama3.2	No

• Custom prompts — Override the default summarization prompt. Use {content}, {title}, and {language} placeholders in your template.

• Language override — Choose the summary output language independently of your FreshRSS UI language.

• Toggle summaries — Click again to show/hide a cached summary without re-fetching.

• Markdown formatting — AI responses are rendered with support for headers, bold, italic, inline code, lists, and horizontal rules.

• Theme-aware styling — Adapts to your FreshRSS theme's colors with a gradient border accent.

• 14 languages — cs, de, en, es, fr, it, ja, ko, nl, pl, pt-br, ru, tr, zh-cn.

• Secure — API keys stay server-side. All requests are proxied through PHP.

cd /path/to/FreshRSS/extensions
git clone https://github.com/deimosfr/xExtension-AiSummary.git

1. Download the latest release or the repository ZIP.
2. Extract it into your FreshRSS extensions/ directory.
3. Rename the folder to xExtension-AiSummary if needed.

1. In FreshRSS, go to Settings > Extensions.
2. Enable AI Summary.
3. Click the gear icon to configure your provider, API key, and model.

User clicks "Summarize"
        |
        v
  JS sends AJAX POST ──> AiSummaryController (PHP)
                              |
                              ├── Fetches article from database
                              ├── If content is too short, fetches full article from URL
                              ├── Strips HTML, truncates to 12k chars
                              ├── Builds prompt (system + article content)
                              └── Calls AI provider API via cURL
                                      |
                                      v
                              Streams response via SSE
                              (status → chunks → done)
                                      |
                                      v
                          JS renders formatted summary
                          in real time below the article title

No build step required. PHP 8.1+ with the curl and mbstring extensions.

# Install test dependencies
composer install

# Run tests
vendor/bin/phpunit

# Run tests with verbose output
vendor/bin/phpunit --testdox

# Run a single test
vendor/bin/phpunit --filter testMethodName

# Lint PHP files
php -l extension.php

xExtension-AiSummary/
├── extension.php              # Extension entrypoint
├── configure.phtml            # Settings form
├── metadata.json              # Extension metadata
├── Controllers/
│   └── AiSummaryController.php  # AI provider API proxy
├── static/
│   ├── script.js              # Frontend logic
│   └── style.css              # Styling (light + dark)
├── i18n/                      # Translations (14 languages)
│   ├── en/ext.php
│   ├── fr/ext.php
│   └── ...
├── tests/                     # PHPUnit test suite
│   ├── stubs/                 # FreshRSS framework stubs
│   ├── AiSummaryControllerTest.php
│   ├── AiSummaryExtensionTest.php
│   ├── I18nTest.php
│   └── MetadataTest.php
└── .github/workflows/ci.yml  # CI pipeline

Contributions are welcome! Please:

1. Fork the repository.
2. Create a feature branch.
3. Ensure tests pass: vendor/bin/phpunit
4. Submit a pull request.

Create i18n/{language-code}/ext.php with all keys from i18n/en/ext.php. The test suite will automatically validate it.

This project is licensed under the GNU General Public License v3.0.