up-to-date

Terminal dashboard for AI, tech, and software engineering news. Pulls from RSS feeds and Hacker News on every launch, no accounts or paid APIs required. Cached stories render instantly; fresh ones stream in behind them. Point it at a local Ollama model and it summarizes the top stories into a digest.

Install

git clone https://github.com/JoeyPatricio/up-to-date
cd up-to-date
uv tool install .        # or: pipx install .
utd

For development: uv sync && uv run utd.

Desktop shortcut (Windows)

After uv sync, the repo includes UtoD.bat (launches utd with a proper window title) and utd.ico. To add a clickable icon to your desktop:

$sh = New-Object -ComObject WScript.Shell
$lnk = $sh.CreateShortcut("$env:USERPROFILE\Desktop\UtoD.lnk")
$lnk.TargetPath = "<repo path>\UtoD.bat"
$lnk.WorkingDirectory = "<repo path>"
$lnk.IconLocation = "<repo path>\utd.ico,0"
$lnk.Save()

Replace <repo path> with the folder you cloned into.

Usage

Command
utd	launch the dashboard
utd config path	print the config file location
utd config init	write a commented starter config
utd --config <path>	use an alternate config file

Keys

Key
Tab / Shift+Tab	focus next / previous panel
j / k, arrows	move selection
Enter / o	open story in browser
c	open comments (HN)
r / R	refresh all / focused panel
g / G	top / bottom of panel
?	help
q	quit

Configuration

Optional TOML file (utd config init creates it). Sections control which panels show and in what order; [[feeds]] entries replace the built-in defaults per section:

[ui]
sections = ["ai-news", "hackernews", "eng-blogs", "tech", "papers"]

[[feeds]]
section = "ai-news"
name = "Hugging Face Blog"
url = "https://huggingface.co/blog/feed.xml"

Default sources: OpenAI, Google DeepMind, and Hugging Face blogs (ai-news); the Hacker News front page; Simon Willison, The Pragmatic Engineer, and the GitHub engineering blog (eng-blogs); Ars Technica and The Verge (tech); Hugging Face Daily Papers (papers). The same story appearing in multiple sources is shown once, keeping the publisher's title and the HN score/discussion link.

Digest (optional, local LLM)

With Ollama installed and a model pulled, enable the digest in your config:

[digest]
enabled = true
model = "llama3.1:8b"

A full-width panel appears above the grid once all sources finish fetching, streaming a one-paragraph summary of the top stories (HN points first, with slots reserved for AI publisher announcements and research papers). Enter opens the first story the digest references. If Ollama isn't running, the panel simply doesn't appear and the dashboard never waits on it.

Model notes: llama3.1:8b is the recommended default (fits fully in 8 GB+ VRAM). qwen2.5:7b and mistral:7b instruct variants also work. Machines with 16 GB+ RAM can try 14B-class models for better story grouping.

Development

uv run pytest        # tests
uv run ruff check .  # lint

License

MIT