CaliLuke declares fork identity: Luke, a local-first job-search workbench
One commit reframes the entire repo - new name, new purpose, new stack, eight-milestone migration plan, and a working Rust FFI bridge to embedded SurrealDB. This fork is diverging hard from upstream.
Commit 9470452d is the fork's thesis statement. The README is rewritten end-to-end, AGENTS.md is updated to describe a transitional architecture, and a 246-line LOCAL_JOB_WORKBENCH_MIGRATION_PLAN.md appears with eight milestones: M0 persistence spike through M7 Wails feasibility. The plan rips out Supabase Auth, Postgres, R2, and Express in favor of embedded SurrealDB/SurrealKV, Loom-generated Go services, Romancy durable workflows on SQLite, local file storage, Ollama as the default AI surface, and a Wails desktop shell targeting macOS arm64.
The M0 spike lands in the same commit. A new backend-go/ package contains a Go test that opens a file-backed SurrealKV database, writes a record, closes it, reopens the same path in-process, and reads the record back. The persistence path runs through CGO and a 134-line Rust static library (backend-go/internal/persistence/rustbridge/src/lib.rs) that links the surrealdb 3.0.5 crate with the kv-surrealkv feature. The spike document explains why: the official Go SDK is remote-only, and the C binding (surrealdb.c.go) failed the same-process reopen because the C wrapper drops its Rust runtime before SurrealKV releases the file lock. The Rust bridge works around an async-shutdown lag with a bounded retry. Cargo.lock adds ~5400 lines of transitive Rust dependencies.
One legal issue worth noting: the LICENSE file is deleted from the repository while the README retains a one-line AGPL-3.0-only reference. AGPL requires the license text to accompany distribution; a one-line mention doesn't satisfy that obligation. This appears to be an oversight rather than intent.
Spotted something wrong? Or know the PR text has fresher detail than the writeup above?