Skip to content

checkpoint==2.0.9

Choose a tag to compare

@github-actions github-actions released this 12 Dec 20:10
a200027

langgraph-checkpoint 2.0.9

Summary of Changes

  • Added safety checks to prevent deadlocks when mixing async and sync store operations (#2673)
  • Improved documentation for storage APIs, especially around vector search and embeddings (#2616, #2623, #2628, #2630)
  • Enhanced Item class with better string representation for debugging (#2655)
  • Fixed msgpack unpacking in serialization to handle non-string dict keys (#2744)
  • Improved performance of async methods in InMemorySaver by removing unnecessary thread pool usage

Detailed Changes

langgraph.store.base.batch.AsyncBatchedBaseStore

  • Added _check_loop decorator to detect and prevent deadlocks when synchronous methods are called from the event loop (#2673)
  • Implemented proper synchronous methods that safely call their async counterparts using asyncio.run_coroutine_threadsafe (#2673)
  • Fixed proper initialization with super().__init__() to ensure parent class initialization (#2673)

langgraph.store.base.Item

  • Added helpful __repr__ method for better debugging and display of items (#2655)
  • Standardized the output order of fields in the dict() method to be more consistent (#2655)

langgraph.store.base.IndexConfig

  • Improved documentation for embedding dimensions, showing the correct model names (e.g., openai:text-embedding-3-small) (#2624)
  • Enhanced documentation for the fields parameter to better explain JSON path syntax and usage patterns (#2628)
  • Added extensive examples for embedding configuration, including OpenAI SDK direct usage (#2616)
  • Added clarification about when indexing is ignored if vector search is not enabled (#2630)

langgraph.store.memory.InMemoryStore

  • Enhanced documentation with clear examples of basic and vector search usage (#2616)
  • Added note clarifying that semantic search requires explicit configuration (#2630)
  • Updated example to use modern LangChain embedding initialization with init_embeddings() (#2616)

langgraph.checkpoint.memory.InMemorySaver

  • Improved performance by removing unnecessary thread pool execution for in-memory operations

langgraph.checkpoint.serde.jsonplus.JsonPlusSerializer

  • Fixed message unpacking by adding strict_map_key=False to handle non-string dictionary keys in msgpack deserialization