Three kinds of context
Compaction does not depend on durable memory. A long chat can retain continuity even when durable memory is disabled or temporarily unavailable.
Durable-memory settings
Durable memory is opt-in at the tenant level:
A knowledge base can narrow the tenant policy further with
kbMemoryMode:
Scope by caller
See Tenancy for the corresponding chat-ownership checks.
Recall lifecycle
1
Prepare chat context
Long histories are compacted when needed. Recent turns remain verbatim and the older summary preserves decisions and prior context.
2
Resolve the memory scope
AutoSage applies the tenant policy, knowledge-base mode, caller type, and
external_subject_id before any recall.3
Recall relevant facts
Relevant durable memories are selected from the resolved scope and added as context for the current message.
4
Generate the answer
Memory context is combined with recent history, the new message, and any knowledge-base or tool results.
5
Retain new facts
When retention permits it, useful facts from the turn can be stored for future conversations.
memory_used boolean rather than exposing recalled memory contents.
Explicit memory
When durable memory is enabled and retention isexplicit_only or automatic, users can store a fact directly with phrasing such as:
Management endpoints
Memory management requires the corresponding tenant or knowledge-base access. Restricted-KB endpoints return isolated data only when that knowledge base uses
kb_restricted mode.
See the Tenant API, Knowledge-base API, and Memory product docs for configuration and management details.
Next steps
Durable memory
Choose retention and knowledge-base memory modes.
Prompt assembly
See where memory and compacted history enter a run.