Module memory 

Agent memory system.

Provides persistent memory for agents across sessions. Memory entries are stored as JSON files via StateStore. Supports embedding-based vector search using TF-IDF + cosine similarity.

Re-exports

pub use embedding_cache::CacheStats;

pub use embedding_cache::EmbeddingCache;

pub use normalizer::cosine_similarity_f32;

pub use normalizer::l2_normalize_f32;

pub use normalizer::l2_normalize_f64;

Modules

auto_memory_bridge

Auto-memory bridge: synchronization between external memory systems and Oxios.

embedding_cache

Embedding cache for reducing API calls.

flash_attention

Flash Attention — block-wise attention for O(N) memory usage.

hyperbolic

Hyperbolic embeddings using the Poincaré ball model.

normalizer

Embedding vector normalization utilities.

Structs

ChunkConfig

Configuration for text chunking.

CurationCandidate

A single candidate for removal during curation.

CurationReport

Report from a curation run.

HnswIndex

HNSW approximate nearest neighbor index.

HnswMemoryIndex

HNSW index manager for memory entries.

MemoryBudget

Budget for memory curation — limits per type.

MemoryEntry

A single memory entry.

MemoryGraph

A memory link graph for computing PageRank-style importance.

MemoryManager

Agent memory manager.

SemanticHit

Result of a semantic search hit.

TextChunk

A text chunk with metadata.

TextVector

Simple TF-IDF vector for text similarity.

Enums

MemoryType

Memory entry type.

Functions

chunk_fixed

Split text into fixed-size overlapping chunks.

chunk_paragraphs

Split text into paragraphs, then group paragraphs into chunks that don’t exceed max_chunk_size.

content_hash

Compute a stable hash of content for deduplication.