Five-Type Memory (Superseded)

Ported from legacy upstream docs/memory/5-TYPE-MEMORY-GUIDE.md. The upstream page exists only as a redirect to current memory docs; the redirect is preserved here for amplihack-rs users who follow links from older READMEs or external references.

See also: Memory Backend Architecture · Memory Tree · Agent Memory Quickstart

This older guide described an automatic 5-type memory flow with hook behavior and backend assumptions that are not the current source of truth for amplihack-rs.

Use These Docs Instead

• Agent Memory Quickstart
• Memory Backend Architecture
• Memory Backend Reference
• LadybugDB Reference
• Memory Tree

Why This Page Was Retired

The current docs separate two concerns that the 5-type guide conflated:

• The in-repo CLI memory backend — exposed by the amplihack memory subcommands (tree, export, import, clean) and backed by either SQLite or LadybugDB. See Memory Backend Architecture.
• The generated amplihack new --enable-memory scaffold — a standalone goal-agent package that ships its own local ./memory/ directory and its own backend selection. See Agent Memory Quickstart.

The older automatic-hook narrative and backend performance claims from the previous version of this page should not be treated as the current user guide.

About the name. "Five-type memory" refers to the original taxonomy of conversation, decision, pattern, context, and learning. A sixth type, artifact, was added later for binary or large-payload entries; the page name is preserved for compatibility with older external links. All six values are accepted by amplihack memory tree --type for backwards compatibility — see Memory Tree for the supported list.