mattrguerra/better-memory-pipelinemainPrivate on GitHub, published here in full

README.md

Language Markdown1.4 KB27 lines

distiller/README.md

  1. # The distiller
  2. Stage 2 of better-memory. `gate.py` has already walled off the private material and
  3. staged only the notes that are new or changed since the last run. The distiller reads
  4. that staging directory and upserts one Notion page per entity into `better-memory`
  5. ([schema](notion-schema.md)).
  6. Distillation needs a model, not a script, so this stage is a Claude skill run as a
  7. scheduled Cowork task: [`better-memory-nightly-distill/SKILL.md`](better-memory-nightly-distill/SKILL.md).
  8. Each run starts with no memory of the last one, so the skill carries everything it
  9. needs: the read boundary, the schema, idempotency by Title, and the two light
  10. judgments (memory vs raw material, no private-by-nature specifics).
  11. ## Schedule and binding
  12. - Nightly, **after** the gate: gate at 20:00 (launchd), distiller at 20:30 (Cowork).
  13. - Bound to `~/cowork-staging/staging` **only**. The task is never given a path to any
  14. vault, so the walled folders are physically absent from everything it can read: the
  15. wall is enforced by the filesystem, not by the skill's instructions.
  16. - Notion access is the Claude connector's own authorization. No token lives in a
  17. vault, a script or this repo. Fill the database URL and data source ID into the
  18. skill before scheduling it.
  19. ## The invariant
  20. **Only this task writes the database.** See
  21. [the invariants](../README.md#the-invariants-that-make-it-safe) in the top-level README.