About AuraTracer

AuraTracer is an open-source engine for in-depth tracking of tech events: it aggregates multi-source public coverage, produces faithful, traceable summaries for each article, groups articles about the same event into a single event page, and reconstructs the full picture with a structured frame and an integrated timeline.

How it works

1. Fetch: pull RSS feeds / web pages declared by topic packs on tiered schedules (hot sources every 5 minutes, the rest 10-60 minutes; WeChat public accounts use a pluggable adapter, disabled by default);
2. Screen and annotate: a low-cost LLM judges relevance and outputs category, entities, importance and structured signals;
3. Same-event clustering: entity alias normalization / embedding similarity merged with deterministic rules;
4. Faithful summaries and event framing: inverted-pyramid summaries, ≤300 characters, facts only from the source text; heavily-covered events are synthesized into a structured timeline (type / key figures / stage narrative);
5. Static site: hostable on GitHub Pages / Cloudflare Pages, with self-owned data.

Compliance and copyright

• This site never republishes full articles; it only shows titles, machine-generated summaries and links to the originals, with per-item source attribution.
• Summaries are AI-generated and may contain errors or omissions — click through to the original to verify; the original wording prevails.
• Copyright remains with the original publishers. If you are a rights holder and want content taken down, reach out via a repository issue or email; we will respond within 48 hours.
• This site is an aggregation tool and does not represent any media stance; event frames and integrated timelines are factual syntheses only and never rate the quality of any source's reporting.
• Titles and summaries are rendered in both Chinese and English via on-device LLM translation for easier cross-reading; machine translation may be imperfect — the original article prevails.

Subscribe

RSS 2.0 feeds are available for the whole site and per category (titles and links only, 12-hour window). Import the OPML file to subscribe to everything at once:

• Main feed：feed.xml
• Models：feed-models.xml
• Chips & Hardware：feed-chips.xml
• Products & Apps：feed-products.xml
• Policy & Governance：feed-policy.xml
• Industry & Investment：feed-business.xml
• Cybersecurity：feed-security.xml
• Science：feed-science.xml
• Download the OPML list (all categories)

Contributing

Contributions of new topic packs (sources + taxonomy + entity aliases, one YAML file) and source adapters are welcome. Repository: GitHub (set site.repo_url in auratracer.yaml)