lustro 0.2.1

Survey and illuminate the AI/tech landscape.

lustro

Survey and illuminate the AI/tech landscape.

lustro is a command-line tool that fetches AI/tech news from RSS, web pages, and X accounts, logs new items, and generates monthly thematic digests.

Installation

pip install lustro

uv tool install lustro

Install optional digest dependencies:

pip install "lustro[digest]"
# or
uv tool install "lustro[digest]"

Quickstart

lustro init          # Create config dirs and starter sources
lustro fetch         # Fetch sources, dedupe, append to log
lustro log           # Show recent log lines
lustro status        # Show config paths, state, cache stats
lustro sources       # List all configured sources
lustro check         # Health-check configured sources
lustro breaking      # Check for breaking AI news
lustro discover      # Find new X handles from For You feed
lustro digest        # Generate monthly thematic digest

Commands

Command	Description
fetch [--no-archive]	Fetch all sources, dedupe against log, append new items. --no-archive skips full-text caching.
check	HTTP health-check all configured sources.
log [-n LINES]	Tail the news log (default 50 lines).
status	Show config paths, state file ages, and article cache stats.
sources [--tier N]	List configured sources with type, tier, and cadence. Filter by tier.
breaking [--dry-run]	Poll tier-1 sources for breaking news (entity + action keyword match). Sends Telegram alerts with rate limiting (3/day, 60min cooldown).
discover [--count N]	Scan X/Twitter For You feed for AI-relevant tweets from untracked accounts. Requires bird CLI.
digest [--month M] [--dry-run] [--themes N] [--model M]	Monthly thematic digest via LLM. Clusters articles into themes, synthesizes evidence briefs.
init	Create config/cache/data directories and starter sources config.
--version	Print version and exit.

Configuration

lustro uses XDG paths by default:

• config: ~/.config/lustro
• cache: ~/.cache/lustro
• data: ~/.local/share/lustro

You can override base directories with:

• LUSTRO_CONFIG_DIR
• LUSTRO_CACHE_DIR
• LUSTRO_DATA_DIR

sources.yaml

lustro init writes ~/.config/lustro/sources.yaml if missing.

Example:

web_sources:
  - name: OpenAI News
    tier: 1
    cadence: daily
    rss: https://openai.com/news/rss.xml
    url: https://openai.com/news/
  - name: Anthropic News
    tier: 2
    cadence: weekly
    rss: https://www.anthropic.com/news/rss.xml
    url: https://www.anthropic.com/news

x_accounts:
  - name: OpenAI on X
    handle: openai
    tier: 2
    cadence: daily

x_discovery:
  enabled: true
  cadence: weekly
  count: 50
  keywords:
    - "\\bAI\\b"
    - "\\bLLM"
    - "\\bGPT"

config.yaml

Create ~/.config/lustro/config.yaml to override defaults:

log_path: ~/.local/share/lustro/news.md
digest_model: google/gemini-3-flash-preview
max_log_lines: 500
digest_output_dir: ~/.local/share/lustro/digests
bird_path: /usr/local/bin/bird
tg_notify_path: /usr/local/bin/tg-notify.sh

Supported options:

• log_path: markdown news log path.
• digest_model: OpenRouter model ID.
• max_log_lines: log rotation threshold.
• digest_output_dir: digest markdown output directory.
• bird_path: absolute path to bird CLI (default: auto-detect from PATH).
• tg_notify_path: absolute path to tg-notify.sh (default: auto-detect from PATH).

Source Format

Top-level keys are grouped lists; each source item is a mapping.

• web_sources: supports name, tier, cadence, and either rss or url (or both).
• x_accounts: supports name, handle, tier, cadence.
• x_discovery: keywords (regex patterns), count, cadence.

tier controls archival behavior during fetch (tier: 1 enables full-text archive attempts) and breaking news monitoring (tier-1 sources only).

X / Twitter Support

X account fetching, checking, and discovery require the bird CLI available on PATH.

If bird is missing, lustro skips X operations gracefully and continues with web/RSS sources.

Breaking News

lustro breaking polls tier-1 RSS/web sources and matches headlines against keyword patterns:

• Entities: major AI companies, regulators, model names
• Actions: launch, release, announce, open-source, acquire, ban
• Negative filter: partnership, hiring, podcast, funding

Alerts are sent via tg-notify.sh (Telegram) with rate limiting (3 alerts/day, 60-minute cooldown). Use --dry-run to preview matches without sending.

Digest Support

Digest generation uses OpenRouter and requires:

• LUSTRO_API_KEY or OPENROUTER_API_KEY
• optional deps installed (lustro[digest])

Without these, lustro digest exits with a clear error.

License

MIT