Corpus Overview¶
ORCA can use a persistent corpus as an optional long-lived knowledge source.
A corpus is a configured folder path that ORCA may reference across sessions and, when allowed, across projects. This is useful when the user has a stable vault, research directory, or project knowledge base that should not need to be reintroduced every time.
This feature is intentionally conservative:
- it is off by default
- reference access and write-back access are separate
- write-back is off by default
- project settings can disable a globally enabled corpus
- missing paths should fail closed instead of silently falling back to guessed context
What A Corpus Is For¶
Use a corpus when ORCA should repeatedly ground itself in the same durable source, such as:
- an Obsidian vault
- a research corpus
- a project-doc folder
- a structured notes directory
- a long-lived product or operating archive
What It Is Not For¶
A corpus is not:
- mandatory before ORCA can do normal work
- a replacement for repo inspection when the repo is the stronger source of truth
- blanket permission to rewrite notes or docs
- a reason to surface graph or vault tooling by default
Relationship To Vault Support¶
Ad hoc vault support is lightweight and session-shaped.
Persistent corpus support is the durable configuration layer above that. It lets the user say: this folder is an approved recurring source of truth, here is what ORCA may read, and here is what ORCA may never write.
Core Safety Model¶
ORCA should treat corpus access as a permissioned capability, not as ambient context.
Default posture:
- no corpus unless configured
- no reference unless enabled
- no write-back unless separately enabled
- no destructive edits
- no silent inheritance when project settings explicitly opt out