Sideeye

A self-hosted news reader and social media agent with AI commentary and Bluesky posting.

Aggregates RSS feeds, scores articles against your interests, generates witty one-liner comments via a hosted LLM, and lets you post directly to Bluesky — all from a minimal dark-theme web UI.

• RSS aggregation — polls configurable feeds every 5 minutes; live push via Server-Sent Events
• Interest scoring — ranks articles by keyword match against your topic list (0–10 per topic)
• Content filtering — blacklist keywords to suppress unwanted categories entirely
• AI commentary — per-article witty comments via DeepInfra (Llama 3.1 8B); auto-comments articles scoring above 40; regenerate anytime
• Bluesky posting — one-click post to Bluesky with the AI comment attached
• Settings UI — two-tab settings page: tune the AI system prompt and temperature (AI tab), manage RSS sources (Sources tab)
• First-run setup banner — new users get a dismissable guide on first visits
• 46 automated tests — covering RSS parsing, content filtering, interest scoring, LLM client, and Bluesky client

git clone https://github.com/yourname/sideeye.git
cd sideeye
python -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -r requirements.txt

cp .env.example .env

Edit .env with your own values:

DEEPINFRA_API_KEY=your_deepinfra_api_key_here
BLUESKY_HANDLE=yourhandle.bsky.social
BLUESKY_APP_PASSWORD=xxxx-xxxx-xxxx-xxxx

• DeepInfra API key — sign up at deepinfra.com
• Bluesky App Password — create one at bsky.app/settings/app-passwords — do not use your main password

python main.py

Open http://localhost:8000. On first run, default config files are created automatically and a setup guide is shown in the UI.

All three files are created with sensible defaults on first run. Edit them via the web UI or directly.

See feeds.example.json, interests.example.json, and settings.example.json for the expected format.

• FastAPI + uvicorn
• feedparser for RSS parsing
• httpx for async HTTP
• atproto for Bluesky AT Protocol
• python-dotenv for secrets

python -m pytest tests/ -v

MIT

Aggregates RSS feeds, scores articles against your interests, generates witty AI comments via DeepInfra, and lets you post directly to Bluesky — all from a clean single-page interface.

• RSS aggregation — polls BBC, DW, Al Jazeera, The Guardian, CNN (configurable)
• Interest scoring — ranks articles by keyword match against your topic list
• Content filtering — blacklist keywords to suppress categories entirely
• AI commentary — per-article witty comments via DeepInfra (Llama 3.1 8B); auto-comments articles with score > 40
• Bluesky posting — one-click post to Bluesky with the AI comment attached
• Settings page — edit the AI system prompt and temperature from the UI
• Live updates — Server-Sent Events push new articles without a page reload

git clone https://github.com/yourname/sideeye.git
cd sideeye
python -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -r requirements.txt

cp .env.example .env

Edit .env with your own values:

DEEPINFRA_API_KEY=your_deepinfra_api_key_here
BLUESKY_HANDLE=yourhandle.bsky.social
BLUESKY_APP_PASSWORD=xxxx-xxxx-xxxx-xxxx

• DeepInfra API key — sign up at deepinfra.com
• Bluesky App Password — create one at bsky.app/settings/app-passwords (do not use your main password)

python main.py

Open http://localhost:8000.

All three files are created with defaults on first run if missing.

• FastAPI + uvicorn
• feedparser for RSS parsing
• httpx for async HTTP
• atproto for Bluesky AT Protocol
• python-dotenv for secrets

python -m pytest tests/ -v

46 tests covering RSS parsing, content filtering, interest scoring, LLM client, and Bluesky client.

MIT