feedy

Developer blog feed aggregator with AI-powered summaries.

feedy scrapes developer blogs from platforms you care about, stores new posts in a local SQLite database (deduplicated by URL), and turns them into a concise daily digest — each item summarized in two sentences plus a "why it matters for devs" line, courtesy of the Anthropic API.

Sources

Source	Name in config	Site
X (Twitter)	x	docs.x.com/changelog
Telegram	telegram	core.telegram.org/blog
TikTok	tiktok	developers.tiktok.com/blog
Meta	meta	developers.facebook.com/blog
OpenAI	openai	openai.com/news
Anthropic	anthropic	anthropic.com/news
Hacker News	hackernews	news.ycombinator.com

All sources are enabled by default. Adding a new one takes about ten lines — see CONTRIBUTING.md.

Install

Requires Python 3.11+.

git clone https://github.com/<your-account>/feedy.git
cd feedy
pip install -e ".[dev]"

This installs the feedy command.

Usage

feedy fetch                      # pull latest entries from configured sources
feedy list                       # show saved entries as a table
feedy digest                     # print today's AI digest

Filtering

feedy list --source hackernews            # only one source
feedy list --since 2026-05-01             # on or after a date (YYYY-MM-DD)
feedy digest --since 2026-05-01           # digest over a date range
feedy digest --source meta                # digest for one source

Export the digest to a file

feedy digest --output report.md           # write Markdown to a file
feedy digest -o report.md                 # short flag

When --output is given, the digest is written to the file instead of being printed to the terminal.

Deliver the digest

feedy digest --slack                       # post to a Slack incoming webhook
feedy digest --email                       # send over SMTP to the recipient

--slack needs a webhook URL ([slack] webhook_url or FEEDY_SLACK_WEBHOOK). --email needs the [email] SMTP settings (password via [email] password or FEEDY_SMTP_PASSWORD). See Configuration.

Serve the digest as JSON

feedy serve                                # http://127.0.0.1:8000
feedy serve --host 0.0.0.0 --port 9000

Exposes GET / (health) and GET /digest?since=YYYY-MM-DD&source=NAME, which returns today's saved entries as JSON.

Configuration

feedy reads an optional TOML config from ~/.feedy/config.toml. If the file is missing, sensible defaults are used (all sources enabled, Markdown output).

# ~/.feedy/config.toml

# Which sources to fetch and include in the digest.
sources = ["x", "telegram", "tiktok", "meta", "openai", "anthropic", "hackernews"]

# Digest format: "markdown" (default) or "plain".
output_format = "markdown"

[anthropic]
# API key for AI summaries. The ANTHROPIC_API_KEY env var takes precedence.
api_key = "sk-ant-..."

[slack]
# Incoming webhook URL for `feedy digest --slack`.
# FEEDY_SLACK_WEBHOOK env var takes precedence.
webhook_url = "https://hooks.slack.com/services/..."

[email]
# SMTP settings for `feedy digest --email`.
# FEEDY_SMTP_PASSWORD env var takes precedence over `password`.
host = "smtp.gmail.com"
port = 587
username = "you@example.com"
password = "app-password"
sender = "you@example.com"
recipient = "you@example.com"

API key

AI summaries need an Anthropic API key. feedy looks for it in this order:

1. ANTHROPIC_API_KEY environment variable
2. anthropic.api_key in ~/.feedy/config.toml

export ANTHROPIC_API_KEY="sk-ant-..."
feedy digest

Without a key, fetch and list still work; digest falls back to showing titles without summaries.

Automating daily fetches

The repo ships a GitHub Actions workflow (.github/workflows/fetch.yml) that runs feedy fetch once a day at 06:00 UTC and caches the database between runs so deduplication keeps working. Trigger it manually from the Actions tab via Run workflow.

Storage

Entries live in a SQLite database at ~/.feedy/feedy.db. Posts are deduplicated by URL, so re-running fetch only saves what's new.

Development

pip install -e ".[dev]"
pytest

Roadmap

See ROADMAP.md for the full feature plan.