Docs assistant

Searches these docs locally; a configured answer service can generate responses.

Ask about this page or the wider Carina docs. Try: “How do I roll back a patch?”

llms.txt

Skip to content
CarinaCarinaCarina

Markdown

Memory overview
Governed local memory, MemoryWrite capability, and optional HMS projection.
Memory

Carina’s local long-term memory belongs to the control plane, not the prompt builder. Memory mutation is a capability-mediated side effect (MemoryWrite).


Source: docs/architecture.md (Governed memory), docs/rpc-api.md (Memory API).


Targets
Section titled “Targets”



















TargetPurpose
memoryProject / agent notes
userProfile facts (scoped by Nebutra identity when available)


The daemon stores bounded entries under its state directory. Each agent run receives a frozen memory snapshot in the prompt — writes during that run persist for future work without changing the current run’s stable prefix.


Write path
Section titled “Write path”


    

  1. Daemon builds a resource string from target, scope, action, operation count, and content hash
    2. 

  2. Requests MemoryWrite from the Rust kernel
    3. 

  3. Built-in policy defaults to requires_approval
    4. 

  4. On allow: apply add/replace/remove/batch atomically after local content scanning and size checks
    5. 

  5. Audit records decision + hash metadata — not raw memory text
    6. 



# conceptual RPC
memory.list / memory.context / memory.status / memory.write


If approval is required, governance.action.approve applies the pending write; governance.action.deny discards it.


Identity scoping (target=user)
Section titled “Identity scoping (target=user)”


Resolution order:


    

  1. CARINA_NEBUTRA_IDENTITY_JSON
    2. 

  2. Claims in CARINA_NEBUTRA_TOKEN
    3. 

  3. CARINA_NEBUTRA_USER_ID
    4. 

  4. Local fallback profile
    5. 



Token claims choose a local memory scope only — they do not grant Gateway, kernel, or filesystem authority.


External recall (HMS)
Section titled “External recall (HMS)”


Optional HMS provider modes: off, hms-shadow, hms-hybrid.


    

  • Shadow evidence does not enter prompts
    • 

  • Hybrid evidence is frozen into the task checkpoint as a fenced, low-trust tool observation
    • 

  • HMS never changes local write/delete authority
    • 



Projection
Section titled “Projection”


HMS projection is a separate, disabled-by-default setting. After local commit, a durable outbox reconciles replace updates and delete tombstones. Projection independently requires NetworkAccess and MemoryExternalize — MemoryWrite approval is not sufficient.


    

  • memory.projection.authorize / retry / reseed
    • 

  • Projection failure never rolls back canonical local memory
    • 



See docs/integrations/hms-memory.md.


Layers at a glance
Section titled “Layers at a glance”
































LayerLifetimeNotes
Turn contextSingle turnModel-visible working set
Session historySessionConversation + tool outcomes
Governed memoryDurable localCapability-gated
Audit storeDurableHash-chained, not model-writable



Source of truth
Section titled “Source of truth”



Next
Section titled “Next”


Was this page helpful?
One vote per page is recorded during this session.
Docs for agents
API
Product