memorize

memorize stores a learning, decision, pattern, or fix to persistent memory (Continuous Claude PostgreSQL) for future sessions.


Learning types

TypeUse when
WORKING_SOLUTIONA fix or approach that worked
FAILED_APPROACHSomething that didn't work (avoid repeating)
CODEBASE_PATTERNA pattern discovered in the codebase
ARCHITECTURAL_DECISIONA design choice and its rationale
USER_PREFERENCEBruno's preferred approach or style
ERROR_FIXHow a specific error was resolved
OPEN_THREADUnfinished work to resume later

Process — mandatory order

  1. Resolve content
    • With argument → that IS the learning. Store verbatim intent; never paraphrase away the detail
    • No argument → ASK Bruno. Never infer from conversation history. Never guess what "should" be memorable
  2. Classify — pick exactly one type from the table. None fit → ask, don't invent
  3. Tags — extract 2-5 specific tags. No generic ("misc", "stuff", "general")
  4. Store via ~/Continuous-Claude-v3/opc/scripts/core/store_learning.py with --session-id, --type, --content, --context, --tags, --confidence high
  5. Update persistent memory affordance (e.g. MEMORY.md in project's memory folder) — concise, grouped by topic: Key Patterns, Critical Rules, API Quirks, Workflows, Error Solutions

Recall

cd ~/Continuous-Claude-v3/opc && PYTHONPATH=. uv run python scripts/core/recall_learnings.py \
  --query "<search terms>" --k 5

Output

Confirm: type, content stored, tags applied, memory ID returned.


  • forget — remove a specific learning
  • distill — process pending events into CLAUDE.md
  • memory — show librarian stats