Commit Graph
3338 Commits
Author SHA1 Message Date
Nuno Campos efb572f15b Add missing branch 2024-09-18 13:18:37 -07:00
Nuno CamposandGitHub d46fd258bf Merge pull request #1753 from Etelis/bugfix-retry-nodes-exception-handling
[Bug Fix] retry_on handling and add support for lists in RetryPolicy
2024-09-18 13:01:38 -07:00
Nuno CamposandGitHub e40aabb1c1 Update retry.py 2024-09-18 10:36:40 -07:00
Nuno CamposandGitHub 0bfeb7f4e1 Update types.py 2024-09-18 10:24:28 -07:00
Nuno CamposandGitHub fd0b52f4ff Update retry.py 2024-09-18 10:23:52 -07:00
Nuno CamposandGitHub f4962479d8 Merge pull request #1757 from langchain-ai/nc/18sep/unit-test-cancelled
Add one more unit test for exception propagation
2024-09-18 09:01:31 -07:00
Nuno CamposandGitHub 94fa06e5f7 Merge pull request #1749 from langchain-ai/nc/17sep/stream-subgraph-sync
Stream output from subgraphs while in-progress for sync stream
2024-09-18 09:01:15 -07:00
Nuno Campos 680fa3c226 Add one more unit test for exception propagation
- we already have plenty of tests for this, but adding one more with a slight variation
2024-09-18 08:53:47 -07:00
Itay Etelis adf5cb0ff1 Fix retry_on handling in run_with_retry
- Correctly distinguish between exception classes, lists/tuples of exception classes, and callables.
- Add support for lists in `retry_on`, alongside tuples.
- Prevent exception classes from being incorrectly treated as callables.
- Raise a `TypeError` if `retry_on` is of an unsupported type.
2024-09-18 13:41:22 +03:00
Itay Etelis 6873229bbf Support lists in retry_on for RetryPolicy
Previously, `retry_on` in `RetryPolicy` accepted only exception classes, tuples of exception classes, or callables.
Update the `retry_on` type annotation to include `List[Type[Exception]]`

Changes:
- Updated `retry_on` in `RetryPolicy` to accept `List[Type[Exception]]`.
2024-09-18 13:39:04 +03:00
Nuno Campos 2a769fb39e Less fun 2024-09-17 17:51:18 -07:00
Nuno Campos c0d8493db5 Only look at chain runs which are nodes 2024-09-17 17:32:14 -07:00
Nuno Campos 01bb6cbd4d Add stream_mode=messages
- yields messages token-by-token from LLMs, and any messages returned from a node
2024-09-17 17:27:37 -07:00
Isaac FranciscoandGitHub 7d3389b6f1 docs: fix links (#1748)
* fix links

* branching link

* readme compatibility
2024-09-17 13:05:37 -07:00
Nuno Campos 3a2a4e44b2 Stream output from subgraphs while in-progress for sync stream
- previously implemented only for async stream
2024-09-17 12:50:29 -07:00
Nuno CamposandGitHub 36c757e758 Merge pull request #1740 from langchain-ai/nc/17sep/deprecation-warnings
Fix some deprecation warnings in tests
2024-09-17 11:57:22 -07:00
Nuno CamposandGitHub a520dc6106 Merge pull request #1744 from langchain-ai/nc/17sep/sync-benchmarks
Add sync benchmarks
2024-09-17 11:57:14 -07:00
Nuno Campos da901b01d8 Fix flaky test 2024-09-17 10:52:39 -07:00
Nuno CamposandGitHub f03dc2f006 Merge pull request #1743 from langchain-ai/nc/17sep/runner-fast-path
perf: Implement fast path in Runner for single-task steps
2024-09-17 10:44:35 -07:00
Nuno Campos 21d60b0d9c Add sync benchmarks 2024-09-17 10:38:49 -07:00
Jacob LeeandGitHub 47a68c16d5 feat(js-sdk): Support CJS (#1742)
* feat(js-sdk): Support CJS

* Bump version
2024-09-17 10:28:23 -07:00
Nuno Campos 9644525b59 Fix some deprecation warnings in tests 2024-09-17 10:19:28 -07:00
Nuno Campos 76ff7c3f1a perf: Implement fast path in Runner for single-task steps
- if the current step has a single task then we can execute it inline, instead of moving it to asyncio task / bg thread
2024-09-17 10:19:19 -07:00
Nuno CamposandGitHub c4d4d61a43 Merge pull request #1735 from langchain-ai/nc/16sep/stream-subgraph-in-progress
Stream subgraph output while it executes
2024-09-17 10:14:26 -07:00
Nuno Campos f59435a892 Add more comments 2024-09-17 09:31:48 -07:00
Nuno Campos bd2ecba622 Fix for py 3.9 2024-09-17 09:29:16 -07:00
Nuno CamposandGitHub 0bf4bf8b6b Merge pull request #1732 from langchain-ai/nc/16sep/stream-selected
perf: In PregelLoop, only emit stream values requested by caller
2024-09-16 21:26:30 -07:00
Nuno Campos c30119824c Add comment 2024-09-16 18:06:06 -07:00
Nuno Campos 54207f4d89 Add test 2024-09-16 18:02:57 -07:00
Nuno Campos d163938a59 Stream subgraph output while it executes 2024-09-16 17:36:53 -07:00
William FHandGitHub 182dcea526 Docstring type (#1733) 2024-09-16 15:49:49 -07:00
Nuno Campos f505afe066 perf: In PregelLoop, only emit stream values requested by caller
- this avoids creating stream values that will be ignored
2024-09-16 15:47:17 -07:00
Nuno Campos 4d0c12117b postgres 1.0.7 2024-09-16 14:10:19 -07:00
Nuno Campos 70af3bc3cb checkpoint 1.0.10 2024-09-16 14:10:03 -07:00
Nuno CamposandGitHub 3b05279e32 Implement serialization with msgpack library (#1716)
* Implement serialization with msgpack library

- encode custom python objects with a msgpack extension type, with constructor path string, and args encoded as nested msgpack doc

* Smaller msgpack extension types

* Update lock files

* lock

* Don't delegate to pydantic json

* Fix kafka serde

- should use our serializer to load, as inputs to subgraphs are serialized using it
2024-09-16 14:06:42 -07:00
Nuno Campos afc64c4f62 langgraph 0.2.22 2024-09-16 13:45:14 -07:00
BagaturandGitHub 29932bb59b docs: clarify course free (#1730)
* docs: clarify course free

* sync
2024-09-16 20:07:16 +00:00
0b1d0eb40d Retry subgraph starting at failing node (#1695)
* Failing test

* Ensure retried subgraphs resume from current point (if any)

* Lint

* Cleanup Test

---------

Co-authored-by: Nuno Campos <nuno@boringbits.io>
2024-09-16 12:53:43 -07:00
fa792e9f74 Update core in poetry lock (#1717)
* Update core in poetry lock

* ci: update langgraph test job for core 0.3 release (#1715)

---------

Co-authored-by: Vadym Barda <vadym@langchain.dev>
2024-09-16 12:36:14 -07:00
Vadym BardaandGitHub ab8aeea129 langgraph: use create_model_v2 and fallback on create_model (#1708)
* langgraph: use create_model_v2 and fallback on create_model

* fix

* code review

* fix
2024-09-16 10:44:33 -07:00
Isaac FranciscoandGitHub b4e62492a7 add deprecation warning for no overall state (#1698)
* deprecation warning for no overall state

* lance comments
2024-09-16 15:09:23 +00:00
Nuno CamposandGitHub e8cb6565ee langgraph 0.2.21 2024-09-13 12:18:04 -07:00
Nuno Campos 4cd9540427 Remove grandalf from dev deps
- this was never included in published package, given it's a dev dep, but some people get confused, so easier to remove
2024-09-13 11:13:10 -07:00
Vadym BardaandGitHub 33f9ec6c04 langgraph: release 0.2.20 (#1714) 2024-09-13 10:36:05 -04:00
c3b4acf3f1 checkpoint postgres: allow passing custom serde (#1699)
* Use correct serde in postgres checkpoint

* Revert metadata serde

* Revert

---------

Co-authored-by: Vadym Barda <vadym@langchain.dev>
2024-09-12 21:09:23 -07:00
Nuno CamposandGitHub 66fc7c96aa More performance improvements in checkpointing and channels (#1685)
* Performance improvements in checkpointer libs

- Use sha1 instead of md5 for hashing (faster in python 3.x)
- Use orjson instead of json for json dumping (sadly can't use for json loading)

* Update tests

* Update

* Use random number instead of hash for get_version_number

* Avoid saving writes for the last task to complete in each step

- only when possible, exceptions for ERROR, INTERRUPT, SEND

* Make Channel.from_checkpoint a regular function

- context manager no longer needed since Context became a managed value

* Use __slots__ for Channels

* Fix for kafka
2024-09-12 21:08:30 -07:00
Nuno CamposandGitHub 8a80b1d3b1 Add 2 more benchmark graphs (#1707) 2024-09-12 16:31:16 -07:00
Nuno CamposandGitHub 95b97d69ec Nc/12sep/perf (#1702)
* Remove unused fil;e

* Add benchmark-fast command for running locally

* Small improvements to jsonplus serializer

* Don't use PregelNode.mapper when schema is a typed dict

- All it would do is create a new copy of same dict

* Avoid copying checkpoint when fetching at beginning of loop

* Fix needs array

* Update tests
2024-09-12 12:50:53 -07:00
Nuno CamposandGitHub 8f87cd1c91 Performance improvements in core library (#1683)
* Performance improvements in core library

- Avoid creating new callback manager when received one as arg
- Avoid looking for config when already received one as arg
- Avoid copies of values in ensure_config/merge_configs
- Implement version of ensure_config that accepts multiple configs (avoids calling merge_configs first)
- Avoid calling merge_configs when we only need to attach extra tags/metadata

* Fix

* Fix

* Try again

* Debug ci job

* Fix

* Try again

* Try again

* Try again

* Some more variations

* Attach annotation to first changed file

* Fix

* Re-enable benchmarks
2024-09-12 11:01:37 -07:00
Nuno CamposandGitHub 73d1051034 Merge pull request #1693 from langchain-ai/nc/11sep/kafka-finally-send
kafka: Expose finally_send as public api
2024-09-11 16:48:41 -07:00