Version localisee en cours de revue

Cette page n est pas encore disponible comme norme localisee revue. Utilisez la source anglaise faisant autorite pour le langage normatif, l interpretation des schemas, les obligations d implementation et les limites de support.

Dossier UAIX-DOC-3323
Chemin /fr-fr/guides/hierarchical-goal-memory/
Utilisation Dossier public canonique

Document status

Public standards page Published on UAIX as part of the current public standards record

Code: UAIX-DOC-3323
Surface: Version localisee en cours de revue
Accès: Public et accessible par lien

Comment utiliser cette page

Use this guide to keep multi-session goal work compact with source-linked subgoals, Folded Records, bounded evidence retrieval, Fresh-Context Review, and review-gated promotion.

Use beside

Long-Running Goal Execution Context Budget Guide Assistant de paquet de mémoire IA Project Handoff Knowledge Graphs

Source-Linked Goal Memory

Fold completed subgoals without turning summaries into authority

Use this guide when a long-running goal needs subgoals, Folded Records, bounded evidence retrieval, Context Packs, Fresh-Context Review, and review-gated promotion while source files remain canonical.

Source

Source-file canonical

Folded Records and Derived Memory Hierarchies are navigation aids until reviewed results are written back to canonical files or public records.

Retrieval

Bounded evidence only

A Context Pack should carry the smallest source slice needed for the next subgoal, plus evidence links and skipped-check risk.

Packets

Reuse existing UAI-1 records

Task-status, blocker, handoff, memory-proposal, and final-report packets cover the first checkpoint path before any new schema is considered.

Use beside

Long-Running Goal ExecutionRuntime goal overlays and checkpoints.Context Budget GuideHot/cold memory split.Assistant de paquet de mémoire IADefault-off local export for this strategy.Project Handoff Knowledge GraphsDerived projection boundaries.

CheckpointFresh-context checkpoint shape

Current subgoal: <subgoal id and title>
UTC timestamp: <current UTC ISO-8601>
Files changed: <paths or no-change evidence>
Completed work: <accepted outcome>
Folded Record: <compact source-linked summary>
Remaining work: <next bounded work items>
Blockers: <none or exact blocker>
Tests run/skipped: <commands, skipped reasons, residual risk>
Exact next action: <one action>
Proposed memory updates: <review-gated targets>

Use this shape after each bounded slice so the next session can resume from canonical evidence instead of private chat memory.