Herald v2: Local-First News Intelligence for AI Agents

I wanted my AI agent to know what's happening in tech - without cloud APIs, paid tiers, or data leaving my machine. So I built Herald: a Claude Code plugin that collects RSS and Hacker News, clusters related articles into stories, scores them, and generates a ranked Markdown brief. v2 is a complete rewrite. Here's how it works. The pipeline Four stages, each a standalone Python module: RSS/Atom feeds ─┐ ├─→ articles ─→ stories (clustered) ─→ scored brief HN Algolia API ─┘ Collect fetches from adapters (RSS, HN Algolia, optional Tavily). Each source is isolated - one failure doesn't stop others. Ingest deduplicates via URL canonicalization (strip tracking params, normalize hosts, sort query params), UPSERTs into SQLite, tracks cross-source mentions, and assigns topics by keyword matching. Cluster groups articles into stories using title similarity. This is the interesting part - more below. Project generates a Markdown brief with YAML frontmatter, stories grouped by type (release, resea