heaventree adds a full Supabase migration file upstream had never shipped
heaventree/mike lands a 389-line SQL migration covering the complete Mike data model - something the upstream willchen96/mike never published. If you're deploying on Supabase, this is the only authored migration in the fork ecosystem that gets you from zero to a working schema.
The migration (supabase/migrations/20260519201745_initial_schema.sql) defines 16 tables spanning the full application surface: user_profiles, user_api_keys with AES-GCM columns (encrypted_key, iv, auth_tag) for per-user Claude/Gemini/OpenAI keys, projects and project_subfolders, documents with document_versions and document_edits, workflows with sharing and hidden-list tables, chats and chat_messages, and the tabular review hierarchy (tabular_reviews, tabular_cells, tabular_review_chats, tabular_review_chat_messages). Direct grants to anon and authenticated roles are revoked; all writes go through the backend service role. An on_auth_user_created trigger auto-inserts a user_profiles row on Supabase signup, with an exception-swallowing handler so a seed failure never blocks the auth flow.
Two things are worth scrutinising before importing. First, projects.user_id is typed as text, not uuid references auth.users(id) - a deliberate looseness that could let bad data slip in if your backend doesn't enforce the constraint. Second, the tabular_model default on user_profiles is gemini-3-flash-preview; check that against whatever model slug your own backend resolves as default.
The same commit also sets "license": "AGPL-3.0-only" in frontend/package-lock.json and swaps the favicon for a 1150-byte custom one. Neither affects the schema, but they signal heaventree is treating this as their own product build.
Spotted something wrong? Or know the PR text has fresher detail than the writeup above?