Skip to content

REMEMBER_BOUNDARY_DECISION.md

Latest commit

 

History

History
84 lines (54 loc) · 2.73 KB

REMEMBER_BOUNDARY_DECISION.md

File metadata and controls

84 lines (54 loc) · 2.73 KB

Remember Boundary Decision

Status

Accepted

Decision

remember will remain a lightweight searchable cache, not a rebuild-guaranteed long-term archive path.

The durable closure path in Evo-Lite remains:

active_context -> context track -> archive

remember may still be useful for:

  • short-lived searchable hints
  • workaround breadcrumbs
  • temporary recall aids that are helpful during active development

remember is not the primary place for:

  • closed-loop task completion records
  • architecture decisions that must survive rebuilds
  • bug postmortems that should be preserved as structured assets
  • anything that needs raw_memory/-based recovery guarantees

Why

Evo-Lite is aimed at keeping AI-assisted project work usable for people who are not necessarily full-time software engineers. That means the main closure path must stay:

  • stable
  • low-ambiguity
  • easy for both humans and AI agents to follow

If remember is upgraded into a second long-term asset path too early, the system becomes harder to reason about:

  • users must guess whether to use remember or archive
  • AI can start treating lightweight notes as durable knowledge assets
  • rebuild semantics become less clear
  • the product boundary between searchable cache and structured archive becomes blurry again

Keeping remember intentionally lightweight preserves a simpler model:

  • active_context tracks the live state
  • track/archive preserve durable project knowledge
  • remember helps with lightweight recall without competing with the main closure path

Consequences

Positive

  • The main flow stays explicit and easier to teach.
  • Rebuild semantics remain centered on structured archives.
  • AI agents are less likely to confuse convenience notes with durable memory assets.

Trade-offs

  • Some useful hints stored only through remember may not survive archive-based rebuild.
  • Users need to consciously promote important knowledge into track/archive.

Operational Guidance

Use remember when:

  • the note is useful to recall later
  • the cost of losing it during rebuild is acceptable
  • it does not need to become part of the project's durable handoff story

Use track/archive when:

  • the work item is closed-loop
  • the knowledge should survive rebuild and migration
  • the result should be inspectable as a structured project asset

Revisit Conditions

This decision should be revisited only if one of these becomes clearly valuable:

  1. remember gains an optional promotion mode that writes both to the database and a structured archive path.
  2. Users repeatedly lose important knowledge because they naturally reach for remember first.
  3. The extra complexity of dual durable paths can be justified by a clear gain in daily usability.