Struct TotalReclawMemory 

pub struct TotalReclawMemory { /* private fields */ }

TotalReclaw memory backend for ZeroClaw.

Implements the ZeroClaw Memory trait with full E2E encryption.

Implementations

impl TotalReclawMemory

pub async fn new(config: TotalReclawConfig) -> Result<Self>

Create a new TotalReclaw memory backend.

This initializes all crypto keys, derives the EOA and Smart Account, sets up the LSH hasher, the embedding provider, and registers with the relay.

pub fn wallet_address(&self) -> &str

Get the wallet address.

pub fn relay(&self) -> &RelayClient

Get a reference to the relay client.

pub fn keys(&self) -> &DerivedKeys

Get a reference to the derived keys.

pub fn private_key(&self) -> &[u8; 32]

Get a reference to the private key.

pub fn name(&self) -> &str

Backend name.

pub async fn store( &self, _key: &str, content: &str, category: MemoryCategory, _session_id: Option<&str>, ) -> Result<()>

Store a memory entry using native UserOp.

Importance defaults to 10 (max) if not specified. Per the client-consistency spec, decayScore = importance / 10, so importance=10 -> decayScore=1.0.

Clears the hot cache after storing to avoid stale results.

pub async fn store_with_importance( &self, _key: &str, content: &str, category: MemoryCategory, _session_id: Option<&str>, importance: f64, ) -> Result<()>

Store a memory entry with explicit importance (1-10 scale).

Per the client-consistency spec, decayScore = importance / 10.

pub async fn store_batch(&self, facts: &[(&str, &str)]) -> Result<Vec<String>>

Store multiple memory entries as a single batched UserOp.

Gas savings: ~64% vs individual submissions for batch of 5.

pub async fn recall( &self, query: &str, limit: usize, _session_id: Option<&str>, ) -> Result<Vec<MemoryEntry>>

Recall memories matching a query.

Uses a hot cache to skip remote queries when a semantically similar query (cosine >= 0.85) was recently answered.

pub async fn auto_recall(&self, query: &str) -> Result<Vec<MemoryEntry>>

Auto-recall: search with the spec-mandated top_k=8.

Per the client-consistency spec, auto-recall at session start uses the raw user message as query and returns top 8 after reranking.

pub async fn get(&self, key: &str) -> Result<Option<MemoryEntry>>

Get a specific memory entry by key/ID.

pub async fn list( &self, _category: Option<&MemoryCategory>, _session_id: Option<&str>, ) -> Result<Vec<MemoryEntry>>

List all memories (paginated export).

Parses each decrypted envelope to extract the correct category (previously hardcoded to Core).

pub async fn forget(&self, key: &str) -> Result<bool>

Forget (soft-delete) a memory entry using native UserOp.

Emits a Memory Taxonomy v1 tombstone (outer protobuf version = 4) so the subgraph can distinguish v1 deletes from legacy v3 deletes.

pub async fn store_v1(&self, input: &V1StoreInput) -> Result<String>

Store a Memory Taxonomy v1 claim (explicit v1 write path).

Use this when the caller has full v1 context (type, source, scope, volatility, optional reasoning). Produces a v4 protobuf envelope with the canonical MemoryClaimV1 JSON inner blob.

For the ZeroClaw Memory trait’s simpler (key, content, category) shape, store_with_importance() remains the primary entry point and continues to emit v3 envelopes during the v0→v1 migration window. Switch your agent to store_v1() when you want v1 provenance-weighted reranking at recall time.

pub async fn pin(&self, _memory_id: &str) -> Result<()>

Pin a memory claim — supersede it with a new v1 blob whose volatility: stable signals “never auto-supersede”.

ZeroClaw implements pin as a supersede operation: fetch the fact, re-store it as a v1 claim with volatility: stable, then tombstone the prior version. The new claim’s superseded_by field tracks the previous id so a future unpin can reverse the chain.

Not-yet-implemented: this method returns an error advising the caller to run the MCP server’s totalreclaw_pin tool instead. See ZeroClaw 2.0 Known Gaps in CLAUDE.md.

pub async fn retype( &self, _memory_id: &str, _new_type: MemorySource, ) -> Result<()>

Retype a memory claim — change its v1 type (claim → directive, etc.) via supersede.

Not-yet-implemented in ZeroClaw’s native trait. The MCP server’s totalreclaw_retype tool is the canonical path.

pub async fn set_scope( &self, _memory_id: &str, _new_scope: MemoryScope, ) -> Result<()>

Set or change the v1 scope of a memory claim via supersede.

Not-yet-implemented in ZeroClaw’s native trait. The MCP server’s totalreclaw_set_scope tool is the canonical path.

pub async fn count(&self) -> Result<usize>

Count active memories.

pub async fn health_check(&self) -> bool

Health check.

pub async fn status(&self) -> Result<BillingStatus>

Billing status – tier, usage, limits. Also updates the billing cache.

pub async fn billing_cache(&self) -> Result<BillingCache>

Fetch billing cache (from disk or relay, with 2h TTL).

Returns a cached billing status with parsed feature flags.

pub async fn quota_warning(&self) -> Option<String>

Check for quota warnings (>80% usage).

Returns a human-readable warning message or None if usage is below 80%. Call at session start (before_agent_start equivalent).

pub async fn debrief(&self, items: &[DebriefItem]) -> Result<usize>

Store debrief items from a session.

ZeroClaw calls this from its consolidation phase or session-end handler. The caller is responsible for running the LLM and passing parsed results.

Each item is stored via the existing store pipeline with source: "zeroclaw_debrief".

pub async fn export(&self) -> Result<Vec<MemoryEntry>>

Export all memories as plaintext (decrypted).

pub async fn upgrade(&self) -> Result<String>

Upgrade to Pro tier – returns Stripe checkout URL.