Memory and your profile

yat:memory keeps durable memory as a typed graph: concepts, learnings, mnemonics, resources, cards and envs, linked by typed relations.

What you say

You say	What happens
remember this / log this learning	Picks a type, checks for an existing node, then writes and links it
what do I know about X	Answers from the graph and cites node ids
link X to Y	Adds a relation, or proposes one if the schema lacks it
show my profile	What the tutor believes about how you learn
why do you think that about me	Walks a profile line down to the dated records it rests on

Three layers

1. Raw records: dated, append-only files such as interaction logs, grill history and journals.
2. Curated facts: the typed graph.
3. The learner profile: short lines about how you learn, shared by every env.

Higher layers cite lower ones and never copy them. A profile line needs evidence from at least two sessions, is proposed at a retro, and is saved only when you accept it. Lines you reject are remembered, so the same evidence never brings them back.

Why writes go through a script

Every write goes through harness/scripts/ontology.py, which checks it against the schema and refuses with a reason if it's invalid. The graph is never edited by hand.

Files it touches

Reads	Writes
memory/ontology/graph.jsonl	memory/ontology/graph.jsonl
harness/memory/schema.toml	memory/ontology/schema.toml (a local overlay)

Terminal: plumbing. Needs python3.

Source

harness/skills/memory/SKILL.md