edit-brain

Editing this brain

This repo is a brain (a context graph built with Open Index). You can edit it two equivalent ways — both land in the same validated store, so pick whichever fits:

• Over MCP (server open-index): create_doc_type, put_entity. Preferred when the server is connected.
• By editing files then running open-index index and open-index validate.

Always start here

Use the navigation guide pre-injected by the MCP host, when available. Otherwise call navigation_guidelines() (MCP) or read doc_types/*.yaml. It tells you which doc_types exist, their fields, and the relationship vocabulary already in use. Reuse existing doc_types and relationship meanings instead of inventing near-duplicates.

Add / update an entity

• Id must be <doc_type>:<slug> (e.g. service:checkout). put_entity is an upsert.
• Fill the doc_type's schema fields. Correlate with related_to: { "target": "<id>", "relationship_edge_meaning": "<meaning>" }. Prefer a declared relationship meaning for the doc_type (see the guide); it's validated.
• File-backed doc_types → the entity is written to entities/<doc_type>/<slug>.json. Index-backed → it lives only in the DB. You don't choose per-write; the doc_type does.

Define a new doc_type (only if none fits)

create_doc_type, or write doc_types/<name>.yaml:

doc_type: <name>
description: <what it is>
storage: index            # index = DB-owned (generated/high-volume) · file = git-tracked (curated)
display: { label_field: name, color: "#6b7280" }
schema:
  fields:
    - { name: name, type: string, search: syntactic, boost: 6 }   # boost = search weight
    - { name: description, type: text, search: semantic }
relationships:            # declare the edges this type uses, so they're discoverable
  - { name: "<meaning>", target_doc_type: <other_type> }

After editing files

Run open-index index (loads file-backed entities) then open-index validate. Fix any reported errors (bad ids, wrong relationship target types) before finishing.