feat(import): note-vault importers — obsidian, joplin, notes, keep, markdown folders

[plind-junior]

What you're trying to do

people arrive with years of notes already written — obsidian vaults, joplin
archives, apple notes and google keep exports, plain markdown folders — and
vouch has nothing to say to them. a fresh KB starts empty and stays empty until
someone does enough work in it.

ditto imports all of the above plus chatgpt / claude / gemini / perplexity
exports (heyditto.ai/docs/importing-your-memory), deduping on stable source ids
so re-importing the same export is safe, up to 200 MiB.

vouch has chatgpt_import.py, codex_rollout.py, vouch ingest and the inbox
folder drop. the gap is note vaults — and note vaults are where vouch has an
advantage ditto structurally cannot match: every note is a file with byte
offsets, so imported knowledge can be receipt-backed and citable rather than
paraphrased into an embedding.

What you've tried

• vouch ingest <file> → works per-file; no vault semantics (frontmatter,
  wikilinks, folder structure), no idempotent re-run.
• inbox folder drop → registers and proposes per file, but dropping a
  10,000-note vault into it is not a migration story.
• vouch import-chatgpt → the shape to copy, one source only.

Suggested shape

vouch import obsidian <vault>
vouch import joplin <path|.jex>
vouch import notes|keep <export>
vouch import md <folder>

• one source per note; note title and path preserved; frontmatter carried into
  source metadata; wikilinks become relation proposals where the target
  resolves.
• stable source ids derived from the origin's own identifier, so re-running
  an import updates rather than duplicates — the property that makes this safe
  to run repeatedly, and the one ditto calls out explicitly.
• claims proposed through the existing selection knob (feat(extract): ingest selection knob + keep dotted numbers intact #541) so a large vault
  does not produce ten thousand pending claims; --max-claims applies.
• everything lands PENDING. an import is a proposal firehose, not a write.

Compatibility considerations

additive; new CLI subcommands, no kb.* method, no schema change beyond a
stable-id field on sources (nullable, back-compatible).

scale is the real question: a big vault needs a batched, resumable import and a
sane review story on the other side — a triage view over "everything from this
import" is probably a precondition for it being usable, not a nice-to-have.

sibling of #431 (conversation-export importers) — same vouch import <kind>
surface, different sources. worth landing the surface once.

Alternatives

• document "convert your vault to markdown and use the inbox" — technically
  true, and nobody will.
• an obsidian plugin that writes into vouch — pushes the gate to the wrong side
  of the boundary.

---

part of the ditto-style track — see .superpowers/DITTO-STYLE-PLAN.md (T4.1).