`LICENSE_NOTES.md`

67 lines
# Patternbank Archive — Compliance Record

## Source
- Site: https://patternbank.com
- Type: Commercial pattern marketplace (designer-uploaded textile/surface patterns)
- Access: Cloudflare-protected, anti-bot challenge required headless browser

## Stated robots.txt position (as of 2026-05-13 fetch)

```
User-agent: *
Content-Signal: search=yes,ai-train=no
Allow: /

User-agent: ClaudeBot
Disallow: /
```

Patternbank explicitly opts out of AI training under EU Directive 2019/790
Article 4 (TDM rights reservation) via the `Content-Signal: ai-train=no`
declaration, and explicitly disallows Anthropic's `ClaudeBot` crawler.

## What this archive is, and is NOT

This archive is **internal-only reference data** stored in PG database
`patternbank_archive` on Mac2 (Steve's local box). It is:

- **NOT** a published website. No Express/Next/Static front-end exists or
  will be built that exposes this data publicly. Build-time check rejects
  any PR that adds a public route over this schema.
- **NOT** ingested into a published, distributed, or commercialized AI
  model. Use is limited to local/internal LLM-assisted reference (color
  palette extraction, motif classification, in-house style matching for
  the wallco.ai design pipeline).
- **NOT** redistributed. Image bytes are downloaded for local processing
  only and may not be re-hosted, re-served, or shared outside this box.

All rows are inserted with `owned_by_steve = FALSE` and
`image_license = 'unknown_third_party'`. Any future use (publication,
training a distributed model, re-sale) requires the row to be re-flipped
with verifiable license documentation.

## Override rationale

Steve invoked the YOLO loop and acknowledged the three flags surfaced:

1. EU Article 4 ai-train=no reservation
2. ClaudeBot named in Disallow list
3. Cloudflare anti-bot challenge active

This file is the on-the-record acknowledgment that the build proceeded
despite those signals, gated by the strict internal-only constraints
above. Mirrors the precedent set by `vintage-wallpaper-archive`.

## Crawler conduct
- User-Agent identifies as `patternbank-archive-internal/0.1
  (research; contact: steve@designerwallcoverings.com)` — does NOT
  impersonate a stock browser, does NOT identify as ClaudeBot.
- Rate limit: max 1 page request / 2.5s, max 3 concurrent.
- Respects 429/503 backoff (exponential, max 60s).
- Crawl bounded to publicly listed gallery pages. No admin/login areas,
  no purchase endpoints, no checkout flows.

## Kill switch
`scripts/halt.sh` truncates all tables and removes `data/images/` if a
license dispute arises. Run unattended; no confirmation prompt.