🐷 feedhog

all your feeds, devoured.

A self-hosted RSS feed aggregator that uses AI to generate summaries of your news feeds.

 

Feedhog fetches articles from your configured RSS/Atom feeds, groups them by category, and produces concise bullet-point summaries. It runs as a small web app that refreshes on a schedule and serves a single-page dashboard.

Features

• Aggregates any number of RSS/Atom feeds, organized by category
• Per-feed control over lookback window (days), whether to include the feed in AI summaries, and optional extras (author, comments)
• AI summaries with structured JSON output and source references — powered by any LLM provider via LiteLLM (Gemini, Claude, OpenAI, and more)
• Configurable refresh interval, number of summary items per category, and model
• Single-page web UI served by FastAPI
• Docker / Docker Compose setup for easy self-hosting
• Fully customizable summarization prompt

Requirements

• Python 3.12+
• An API key for your chosen LLM provider

Quickstart

With Docker Compose (recommended)

1. Clone the repository:

   git clone https://github.com/mlux86/feedhog.git
   cd feedhog

2. Create a .env file with your API key (see LLM provider):

   echo "GEMINI_API_KEY=your_key_here" > .env

3. Copy the default config and edit it to add your feeds and settings:

   cp config.default.yaml config.yaml

4. Start the service:

   docker compose up -d

5. Open http://localhost:52987 in your browser.

Without Docker

1. Install dependencies:

   pip install .

2. Create a .env file:

   echo "GEMINI_API_KEY=your_key_here" > .env

3. Copy the default config:

   cp config.default.yaml config.yaml

4. Run the app:

   uvicorn main:app --host 0.0.0.0 --port 8000

Configuration

All configuration lives in a single config.yaml file. Copy the provided default to get started:

cp config.default.yaml config.yaml

config.yaml

refresh_interval_hours: 3          # How often to fetch feeds and regenerate summaries
summary_items_per_category: 7      # Maximum bullet points per category in the summary
language: English                  # Language for the generated summaries
model: gemini/gemini-flash-latest  # LiteLLM model string (provider/model)

categories:
  - name: Technology
    feeds:
      - url: https://hnrss.org/best
        source: Hacker News Best
        extras:
          - comments     # Show link to HN discussion thread per article

      - url: https://xkcd.com/atom.xml
        source: xkcd
        summarize: false # Include in feed view but skip AI summary
        days: 14         # Look back up to 14 days (default: 1)

Feed fields

Field	Required	Default	Description
url	yes	—	URL of the RSS or Atom feed
source	yes	—	Display name for the feed
summarize	no	true	Whether to include this feed in the AI summary
days	no	1	How many days back to include articles from
extras	no	[]	Optional extra fields to extract per article (see below)

Extras

By default no extra metadata is extracted. Add any of the following to a feed's extras list to opt in:

Value	Description
author	Show the article author next to the title
comments	Show a [c] link to the comments URL (e.g. the HN discussion thread)

LLM provider

The provider is determined by the model value in config.yaml and the corresponding API key in .env. Set whichever key matches your chosen provider:

Provider	.env key	Example model
Google Gemini	GEMINI_API_KEY	gemini/gemini-flash-latest
Anthropic Claude	ANTHROPIC_API_KEY	claude-sonnet-4-6
OpenAI	OPENAI_API_KEY	gpt-4o

See the LiteLLM docs for the full list of supported providers and model strings.

Summarization prompt (prompt.txt)

The file prompt.txt is passed directly to the LLM as the system prompt. You can edit it freely to change the language, style, or structure of the summaries. The placeholders {max_items} and {language} are replaced at runtime with values from config.yaml.

License

MIT