Crate post_cortex_memory 

Conversation memory orchestrator for post-cortex.

This crate is the top of the dependency tree — it pulls in post_cortex_core (domain types), post_cortex_storage (RocksDB + SurrealDB backends), post_cortex_embeddings (BERT + HNSW), and post_cortex_proto (gRPC wire types). It owns ConversationMemorySystem, the lock-free hot/warm/cold memory hierarchy that ties everything together, plus the content vectorizer pipeline, semantic query engine, query cache, graph-aware context assembly, and the relevance scoring helpers.

Phase 4 of the migration adds a canonical PostCortexService trait impl here so MCP/gRPC transports can delegate to a single internal handler per operation (TODO.md:106-117). Phase 5 adds the non- blocking pipeline work queues (TODO.md:136-145).

Re-exports

pub use error::Error;

pub use error::Result;

pub use memory_system::ConversationMemorySystem;

pub use memory_system::SystemConfig;

pub use pipeline::Pipeline;

pub use pipeline::PipelineConfig;

pub use pipeline::PipelineError;

pub use services::MemoryServiceImpl;

Modules

content_vectorizer

Content Vectorization Module

context_assembly

Graph-aware context assembly that combines semantic search results with entity-graph traversals.

error

Typed error hierarchy for post-cortex-memory.

memory_system

Conversation memory system with actor-based storage and lock-free session management.

performance

Runtime performance monitoring: operation timers, cache trackers, and health snapshots.

pipeline

Non-blocking write pipeline.

query_cache

Query caching system for semantic search results

scoring

Scoring strategies for semantic search results

semantic_query_engine

Semantic Query Engine

services

Canonical PostCortexService implementation for post-cortex-memory.

Macros

time_operation

Convenience macro for timing operations