Store System For Cross Thread Persistent Memory

via “store system for cross-thread persistent memory and knowledge bases”

Graph-based framework for stateful multi-agent LLM applications with cycles and persistence.

Unique: Pluggable BaseStore interface for cross-thread persistent storage, enabling agents to build and access shared knowledge bases independent of execution checkpoints

vs others: More flexible than in-memory state, but less queryable than full databases; requires custom key-value patterns

via “store api for cross-thread persistent memory and knowledge bases”

Build resilient language agents as graphs.

Unique: Provides a pluggable Store API for cross-thread persistent memory, separate from checkpoint-based thread state. This two-level memory architecture (short-term channels + long-term store) enables agents to maintain both execution state and persistent knowledge without coupling them.

vs others: Separates short-term execution state from long-term memory, enabling cleaner architecture than frameworks storing all context in a single state structure. Provides better scalability for multi-agent systems than thread-local storage.

via “store system for cross-thread persistent memory”

Building stateful, multi-actor applications with LLMs

Unique: Implements a pluggable store system (BaseStore interface) for cross-thread persistent memory, complementing per-thread channel state. Store operations are integrated into the execution model with checkpoint metadata, enabling reproducible multi-agent workflows with shared knowledge bases.

vs others: More flexible than implicit global state (explicit store operations are visible) while remaining simpler than distributed consensus systems, enabling agents to share knowledge without complex coordination.