ACT I
Ingest
Core Data Atlas
The ledger never forgets.
It only gets stricter.
This is the data story behind Locus: every memory has structure, every sync has order, and every rule exists to keep operations steady under pressure.
ACT II
Validate
ACT III
Persist
ACT IV
Retrieve
ACT V
Synchronize
Five scenes of persistence drama
Read top-to-bottom. Each scene shows how memory stays consistent, recoverable, and trustworthy.
Scene 1
Arrival
Memory enters with scope, time, and cognitive context.
Scene 2
Judgment
Structure and quality checks run before anything is accepted.
Scene 3
Identity
Each memory gets a stable identity so replays stay safe.
Scene 4
Consequences
The system decides whether to keep, refresh, or ignore duplicates.
Scene 5
Continuity
Sync checkpoints preserve order and make restarts predictable.
Entity constellation
Hover each block to see how responsibilities connect across the data model.
Rules theater
These guarantees keep memory trustworthy as systems scale and recover.
Idempotency
Rule
The same memory cannot create duplicate records in the same scope.
Ordering
Rule
Change streams always return in a stable, deterministic sequence.
Compatibility
Rule
Legacy data remains readable during migrations and backfills.
Validation
Rule
Invalid structure is rejected before it can affect stored memory.
Coherence
Rule
Scoring state stays coherent even when partial fields are missing.
Rekey
Rule
Scope moves happen together so related records never split apart.
Index pressure board
Key index paths that keep retrieval fast and synchronization stable.
Identity Index
Prevents duplicate writes during replay and retry.
Cursor Index
Powers reliable incremental sync pulls.
Scope Index
Keeps tenant and session retrieval fast.
Checkpoint Index
Maintains connector progress with predictable recovery.