Trust and scoring
Methodology
The site is built from a local SQLite seed exported to static JSON. Ranking is intentionally gated until records are manually reviewed.
Data pipeline
`npm run export:data` reads `raw_research/ai_builder_resource_seed.sqlite`, normalizes it into `src/data/generated/resources.json`, and Astro pre-renders pages from that static file.
Review status
Most records are currently `needs_review` or `auto_enriched`. Those pages are useful locally but should not be treated as reviewed public rankings.
Scoring weights
| Signal | Weight | Meaning |
|---|---|---|
| topical fit | 25 | Fit for the target use case, category, and user intent. |
| popularity | 15 | Adoption and attention signals such as stars and forks, using scaling rather than raw counts. |
| maintenance | 15 | Signals that the resource is maintained, including commits, releases, issue activity, and deprecation status. |
| freshness | 10 | How current the directory metadata and source evidence are. |
| documentation quality | 10 | Quality of README, docs, examples, installation guidance, and API references. |
| trust source quality | 10 | Reliability of source refs, ownership, official registry/site evidence, and licensing clarity. |
| beginner friendliness | 5 | Ease of first successful use for a newer builder. |
| open source self hosted value | 5 | Value from open-source licensing, local/self-hosted use, and reduced lock-in. |
| commercial pricing clarity | 5 | Transparency of pricing, free tier, and commercial limitations. |
What is not done yet
Manual verification, reviewed scoring, alternative/comparison pages, domain selection, deployment, and production indexing are still pending.