v0.12.0
omnifeed v0.12.0
Self-hosted web search (SearXNG) + LLM-friendly crawling with a dedicated Reddit engine — MCP server, Open WebUI compatible.
Changelog
🎉 Features
- 5e8b9d3 feat(hackernews): add a Hacker News engine via the Algolia API
- a74dbcb feat(reddit): render subreddit listing URLs (/r/{sub})
- 20c567a feat: make crawl4ai content prune configurable - observability track http attempt
🐛 Bug fixes
- 3db8775 fix(crawl4ai): keep hyperlink anchor text and stop retrying anti-bot blocks
- ab286f2 fix(hackernews): correct feed recency, comment permalinks, and code fidelity
- 39e7c49 fix(reddit): correct subreddit-listing URL handling and limiter key
- 073c639 fix: address Tier-2 review findings in the HN and Reddit engines
♻️ Refactor
- d7398bb refactor: Tier-3 cleanups — shared anti-bot retry predicate, dedup, rune-safe truncate
📄 Documentation
- 5849809 docs: add web-stack benchmark results and omnifeed improvement plan
- 647c844 docs: document the HN engine + subreddit listings; correct the crawl4ai gotcha
Other changes
- 0f41579 chore(release): v0.12.0 [skip ci]
Docker
docker pull kinorai/omnifeed:0.12.0
Verify
The checksums.txt file contains SHA-256 hashes for every artifact in this release.