repo:
enabled: true
lookback_hours: 24
max_commits: 20
max_issues: 20
issue_labels_preference: ["bug", "compat", "performance", "observatory", "docs"]
arxiv:
enabled: true
max_results: 12
categories: ["cs.AI", "cs.LG", "cs.SE", "cs.CL"]
# Keep the query tight; better small and relevant than “all of arXiv”
query: '(agent OR agents OR "tool use" OR "code agent" OR workflow) AND (evaluation OR benchmark OR reliability OR "tool calling")'
# Lab/company updates:
# Option A: RSS endpoints
# - OpenAI News page links an RSS endpoint at /news/rss.xml. [oai_citation:0‡OpenAI](https://openai.com/news/)
rss:
enabled: true
feeds:
- name: "OpenAI News"
url: "https://openai.com/news/rss.xml"
max_items: 10
# Option B: HTML listing sources (more brittle, but works when RSS isn't available)
# Anthropic newsroom is a good candidate for HTML-list parsing. [oai_citation:1‡Anthropic](https://www.anthropic.com/news)
#
# Selector Maintenance Guide:
# If a site stops returning signals:
# 1. Open the site in browser
# 2. Inspect a news/blog article element
# 3. Update container/title/link selectors to match current HTML
# 4. Set fallback_to_generic: true for safety
# 5. Test with: npm run agentops:daily -- --dry-run
html:
enabled: true
sources:
- name: "Anthropic Newsroom"
url: "https://www.anthropic.com/news"
max_items: 10
# Site uses client-side rendering, generic selectors work better
fallback_to_generic: true
- name: "Google Research Blog"
url: "https://research.google/blog/"
max_items: 10
# Site uses client-side rendering, generic selectors work better
fallback_to_generic: true
- name: "Google DeepMind News"
url: "https://deepmind.google/blog/"
max_items: 10
selectors:
container: "article.card-blog"
title: "h3"
link: "a"
fallback_to_generic: true
# Notes:
# The web tool had trouble fetching some RSS XML directly, but that doesn’t mean your Node fetch will fail; it’s often user-agent/header related. For OpenAI, the site itself links the RSS endpoint.
