v2.41.0 — full Clef for spam checks and a bring-your-own-key decisions provider

Highlights:

• feat(spam): new spam_ai_model setting (admin Settings page or env) picks clef-flash (default, unchanged) or full clef for SPAM_PROVIDER=workers-ai.
• feat(spam): new SPAM_PROVIDER=decisions sends the spam question to any OpenAI-style /decisions endpoint with your own key (OpenRouter's alpha route by default). New env vars: SPAM_DECISIONS_API_KEY (secret), SPAM_DECISIONS_URL (optional, must be https://), SPAM_DECISIONS_MODEL (also the spam_decisions_model admin setting). It shares spam_ai_threshold with workers-ai.
• feat(admin): the dashboard spam summary flags a decisions provider with no key.

Notes:

• Drop-in upgrade: no migrations. All new env vars are optional; existing providers behave as before.
• With decisions, comment text leaves Cloudflare (no email, IP or user agent). Update your privacy policy; see docs/compliance/subprocessors.md and docs/ANTISPAM.md.