Bỏ qua tới nội dung
TakoAPI
Quay lại kỹ năng

guava-memory

bởi koatora20

Notes & PKM
—Chưa có đánh giá
Đánh giá của bạn

Tóm tắt

Structured episodic memory with Q-value scoring.

Mô tả

Kỹ năng này làm gì

GuavaMemory adds structured episodic memory to OpenClaw agents. It records task outcomes with Q-value scores, searches past episodes by semantic similarity, and promotes high-success patterns into reusable skill procedures. Anti-patterns from failures are tracked to prevent repeating the same mistakes.

Unlike static notes or documentation, Q-value scoring continuously reweights past approaches by actual outcome, so the most reliable procedures surface first rather than the most recent ones.

Khi nào nên dùng

  • Recalling which debugging approach worked on a similar error last week
  • Avoiding a deployment sequence that previously caused an outage
  • Promoting a repeated file-parsing workflow into a reusable skill procedure
  • Surfacing the highest-Q-value episode before starting a new refactor task
  • Logging a failed API integration attempt as an anti-pattern for future reference

Quy trình ví dụ

ĐẦU VÀO

User asks: debug the failing authentication test suite

AGENT
  1. 1

    Run memory_search with intent 'debug failing tests' to find related episodes

  2. 2

    Filter results by Q-value, select top 3 active episodes

  3. 3

    Apply the success pattern from the highest-scoring episode

  4. 4

    Complete the task, then write ep_YYYYMMDD_NNN.md with intent, context, outcome, and Q-value

  5. 5

    Update memory/episodes/index.json to keep the index in sync

ĐẦU RA

Test suite fixed using a recalled pattern; new episode recorded with reward score and feel tag for future retrieval

Cài đặt

clawhub install koatora20/guava-memory

Chạy lệnh CLI, hoặc dán prompt vào Claude Code, Cursor, Codex hay coding agent bạn đang dùng.

Liên kết