`f6aba6c`

Bootstrap manifest on first train of a hand-authored .dlm

Trainer's load_manifest call expected the file present, but a .dlm authored without 'dlm init' (LSP / VSCode flow, or hand-edited) only has the on-disk store after ensure_layout(). Drop the just_scaffolded gate so any never-init'd doc gets a fresh manifest before the load.

Authored by mfwolffe <wolffemf@dukes.jmu.edu> 2 weeks ago

SHA: f6aba6cfd119f82e60e99b8fcda92d9446472538
Parents: be9fc89
Tree: d2f5104

1 changed file

Status	File	+	-
M	`src/dlm/cli/commands.py`	9	5

src/dlm/cli/commands.pymodified

      store = for_dlm(parsed.frontmatter.dlm_id)
      store.ensure_layout()
 -    # `dlm init` writes a manifest as part of store provisioning. When
 -    # we just scaffolded a fresh `.dlm`, mirror that manifest write
 -    # here too. Guarded by exists() so --rescaffold (same dlm_id,
 -    # prior store) preserves training history.
 -    if just_scaffolded and not store.manifest.exists():
 +    # `dlm init` writes a manifest as part of store provisioning. Mirror
 +    # that manifest write here when the store layout exists but has no
 +    # manifest yet — covers two flows:
 +    #   - auto-scaffold via `dlm train <dir>` (just_scaffolded path)
 +    #   - hand-authored .dlm with a fresh ULID that never went through
 +    #     `dlm init` (e.g. authored via the LSP / VSCode extension)
 +    # License acceptance has already been validated upstream by this
 +    # point, so we just record it.
 +    if not store.manifest.exists():
          from dlm.base_models import is_gated
          from dlm.base_models.license import require_acceptance
          from dlm.store.manifest import Manifest, save_manifest