mynews2.0.0

mynews parses RSS and Atom feeds, filters their stories, scores their relevance, and broadcasts a personalized news stream to the destinations you choose. You define sources and broadcast preferences in a compact configuration, with five-minute feed polling, a 24-hour story window, and stdout output as the defaults. Per-source keyword include and exclude rules can use whole-word matching, keyword-or-score matching, score thresholds, and date cutoffs, while concurrent fetching prevents a slow feed from holding up the rest.

Optional relevance scoring compares titles with your interests using local static embeddings through model2vec or simple keyword matching; minScore removes low-relevance items, and interests can be overridden for individual sources. With embedding scoring, entries lacking descriptions receive best-effort extractive summaries unless disableArticleSummaries is enabled. Stories can be sent to stdout, Telegram, Discord, Slack, or a generic webhook that receives raw story JSON, with multiple independent apps configured as separate targets. Digest mode batches the top-N highest-scoring stories for each interval, and cross-source deduplication merges the same article when tracking parameters or URL fragments differ.

OPML import turns a feed-reader export into a starting configuration, while metricsAddr exposes /healthz and /metrics for monitoring. Periodic crash-safe persistence records seen stories, and graceful shutdown preserves that state; running bare mynews uses an in-memory database and prints to stdout. Law Zava’s open-source mynews 2.0.0 runs on Linux in the System Tools category, targeting x86_64 (amd64) and requiring snapd. Its single self-contained pure-Go binary uses no CGO; keyword scoring and the rest of the program work offline, while embedding scoring downloads a small static-embedding model from HuggingFace on first use.