Support for multiple git worktrees

[dcnoaoh]

I know there is work around improving indexing and avoiding duplicate directories for git worktrees. What I am not sure of is if multiple worktrees, let's say worktree a and worktree b, both would have to have cce init run for them, or if there is already information sharing between worktrees.

[rajkumarsakthivel]

Good question. As of v0.4.23, each worktree needs its own cce init and gets its own storage directory. The storage path collision fix we just shipped uses a <basename>-<6hex> slug derived from the absolute path, so two worktrees of the same repo (e.g. ~/work/api and ~/work/api-feature) resolve to separate storage dirs with separate indexes.

This is intentional for the index layer: each worktree can be on a different branch with different file contents, so sharing chunk data would produce stale results. The index is fast enough that re-indexing per worktree is not a bottleneck.

For memory (decisions, session history, code areas), sharing across worktrees of the same repo would make sense since those are project-level, not branch-level. That's not implemented yet but is on the radar. The workaround today is that decisions made in one worktree won't surface in another unless you manually re-record them.

Summary of current behavior:

• Run cce init in each worktree separately
• Each worktree gets its own index (correct, since file content differs)
• Each worktree gets its own memory.db (not ideal, could be shared in a future version)
• No collision between worktrees that share the same directory basename