Generates SVG charts (bar, line, scatter) for weekly reports and landscape figures without relying on external plotting libraries like matplotlib or plotly. Demonstrates non-obvious data engineering skill by manually constructing SVG elements from data, making it reusable for any project that needs lightweight, dependency-free chart generation.

Full refresh pipeline that fetches GitHub repository metadata (stars, license, language) and ClickHouse OpenRank metrics concurrently, with retry logic and rate-limit handling. The pattern of combining multiple data sources with idempotent updates and quality checks is reusable for any project that needs to maintain a curated dataset of open-source projects.

Implements a multi-stage candidate discovery pipeline using GitHub search queries, relevance scoring, and agent signal counting. The methodology of filtering from thousands of candidates to a curated shortlist with quantitative and qualitative criteria is a reusable pattern for ecosystem scanning and landscape curation.

Central orchestration script that contains the keyword taxonomy engine (word-boundary regex with tiered keywords) and the recommendation scoring algorithm (OpenRank growth, trend slope, participants, novelty bonus). These components are self-contained and can be extracted for domain-specific classification and ranking tasks in other ecosystem analyses.