Releases: langchain-ai/langgraph
Releases · langchain-ai/langgraph
checkpoint==1.0.10
0.2.21
Changes since 0.2.20
- langgraph 0.2.21
- Remove grandalf from dev deps
0.2.20
Changes since 0.2.19
- langgraph: release 0.2.20 (#1714)
- More performance improvements in checkpointing and channels (#1685)
- Add 2 more benchmark graphs (#1707)
- Nc/12sep/perf (#1702)
- Performance improvements in core library (#1683)
- langgraph: update snapshots for 0.3.x core tests (#1692)
- langgraph: update core & snapshots (#1689)
- update bench job
- Merge pull request #1609 from langchain-ai/nc/4sep/benchmark
- Remove files
- Rename command
- Fix output dir
- Add benchmark/profile commands
- refactor: Type narrowing for ToolNode.name (#1678)
- Add benchmark script
- Implement subgraph delegation for distributed arch
- Use thread_id for partition key, ignore tasks for stale checkpoints
- Fix subgraph test
- Add drain_topics test helper
- Remove is_resuming arg
- In update state skip nodes not found in instance
- Remove double shield usage
- WIP Subgraphs
- Rename compose projects
- Catch errors thrown by nodes run by Executor
- Implement LangGraph Scheduler for Kafka
- adding langchain academy link (#1643)
- langgraph: add tests for MessagesState/AnyMessage (#1637)
- Merge
- update ye ol' snapshots
- icanspell
- Assert input/output channels aren't managed
- Exclude shared
- Update test
- Merge branch 'main' into wfh/output_schema
- Optional types for dataclass state defs (#1557)
- Support optional output schema
0.2.19
0.2.18
Changes since 0.2.17
- Fix copy
- langgraph 0.2.18
- Add back Pregel.copy()
0.2.17
Changes since 0.2.16
- langgraph: update mermaid snapshots (#1627)
- langgraph 0.2.17
- langgraph: changes for compatibility pydantic v2 / langchain-core==0.3 (#1594)
- Remove legacy id
- Split out prepare_single_task from prepare_next_tasks
- Merge pull request #1610 from langchain-ai/an/04sept/metadata-interrupts
- Add before and after values to __interrupt metadata.
- Lint
- Speed improvements to local_read, emit output
- Fix ids in debug task events
- Add test to confirm only one root run is produced
- Fix
- Try to fix test
- Avoid using json encoding when generating task ids
- Run loop tick in bg thread
- Run branch.reader in bg thread
- Store loop in async bg exec init
- Reduce from 9s to 4s on benchmark graph
- Reduce cpu time spent on langchain-core utilities
- Fix test
- Move runner code to standalone class, use in Pregel.stream/astream
- Add large parallel graph test
- Avoid deepcopy
- x
- Merge branch 'main' into isaac/nodelogging
- Update
- Update
- Use a lock for all operations on sqlite checkpointer
- Remove pydantic base class from Pregel/PregelNode
- Merge pull request #1549 from langchain-ai/wfh/optional_types
- checkpoint: Implement sync methods in async checkpointers
- Disable test in old versions of py
- test: Ensure that async checkpointers in tests are setup in same event loop as they are used in
- Handle py3.9
- fmt
- x
- x
- Merge branch 'main' into isaac/nodelogging
- Merge branch 'main' into wfh/optional_types
- Return a default value instead
- Update snapshot
- Handle total=false
- Support optional types from typeddict
- wip
checkpoint==1.0.9
checkpointsqlite==1.0.3
Changes since checkpointsqlite==1.0.2
- sqlite 1.0.3
- Use a lock for all operations on sqlite checkpointer
- checkpoint-sqlite: close connection in from_conn_string (#1589)
checkpointsqlite==1.0.2
Changes since checkpointsqlite==1.0.1
- sqlite 1.0.2
- Require latest version of langgraph-checkpoint
- checkpoint: Implement sync methods in async checkpointers
checkpointpostgres==1.0.6
Changes since checkpointpostgres==1.0.5
- postgres 1.0.6
- Require latest version of langgraph-checkpoint
- Lint
- checkpoint: Implement sync methods in async checkpointers