[recipes] Brain backup and export

[alanshurafa]

Summary

Node.js recipe that exports every Open Brain Supabase table to dated local JSON files with streaming pagination.

• Paginates through PostgREST 1,000 rows at a time; streams rows to disk so large tables never load fully into memory
• Dated filenames (<table>-YYYY-MM-DD.json) for simple daily retention
• Prints a row-count and file-size summary per table
• thoughts is required; optional companion tables (entities, edges, thought_entities, ingestion_jobs, ingestion_items) are backed up only if present. Stock Open Brain installs skip them cleanly instead of erroring.
• No dependencies beyond Node 18+; config via .env.local

Why

Open Brain users accumulate thousands of thoughts over time but Supabase-side backups require plan-tier access or manual dumps. A zero-dependency Node recipe gives anyone on the free tier a scriptable backup path that runs from cron / Task Scheduler.

The optional-table handling lands specifically so this recipe works against every Open Brain install, not just ones that have also installed the entity-extraction or smart-ingest schemas. Missing companion tables return a PostgREST 404, which the script treats as "skip with a log line" rather than an error.

Part 11 of 12 in the OB1 Alpha Milestone consolidation.

Test plan

• Run against a stock Open Brain project (only thoughts table) — verify thoughts backs up, optional tables log skipped (table not present)
• Run against a project that has the optional tables — verify all exports succeed
• Drop thoughts temporarily → verify the script throws Required table "thoughts" not found
• Kill mid-run on a large thoughts export → verify no partial/corrupt JSON left behind (next run overwrites cleanly)
• Verify dated filename rotates correctly across days
• node --check backup-brain.mjs passes
• metadata.json passes the gate schema

[alanshurafa]

Conflicts with main. Before I rebase, should this stay as-is, be reshaped, or be dropped? I can refresh it quickly if you want to keep it.

[alanshurafa]

Rebased onto main — no conflicts now. This updates the existing recipes/brain-backup/ rather than adding it.

For review context: main's copy of this recipe predates its code review. This PR is the reviewed version — terminating pagination (the loop on main checks a stale page-size value), PGRST205-only "table missing" handling, atomic writes via tmp+rename, an AbortController fetch timeout, and BOM-tolerant env parsing. An upgrade over what's on main.