Commit Graph
4906 Commits
Author SHA1 Message Date
Nuno CamposandGitHub 4fb2aeacc7 Add checkpoint_during arg (#4169)
- This provides a new mode of execution where only the last checkpoint
is saved
- We save the last checkpoint no matter how the agent run is terminated
(success, error, interrupt, etc)
- This cuts down on cpu time spent on checkpointing, while not losing
any resilience benefits, given individual task writes are still saved
- If an error occurs and the run is retried, any tasks that completed
successfully before will be skipped (as currently)
- checkpoint_during=True is useful when you want to time-travel to inner
steps of a run
- The default value will remain the current behavior, ie.
checkpoint_during=True
2025-04-08 15:03:06 -07:00
David DuongandGitHub 8252668bcc release(sdk-js): 0.0.64 (#4207) 2025-04-08 23:46:23 +02:00
Tat Dat Duong 27e4b0fcfe release(sdk-js): 0.0.64 2025-04-08 23:45:03 +02:00
David DuongandGitHub ba388e25b3 feat(sdk-js): add auth types (#4199) 2025-04-08 23:44:33 +02:00
Nuno Campos 947a233fc5 Fix 2025-04-08 14:20:29 -07:00
Nuno Campos b76dc8ae0a Fix 2025-04-08 14:15:13 -07:00
Nuno Campos cbbfaba1fd Add comments 2025-04-08 14:07:08 -07:00
Nuno Campos e9aec77893 Add more tests 2025-04-08 14:07:02 -07:00
Nuno CamposandGitHub 067c4dd246 Implement simpler filtering of config keys in RemoteGraph (#4205) 2025-04-08 13:51:11 -07:00
Nuno Campos ccc21974e0 Implement simpler filtering of config keys in RemoteGraph 2025-04-08 13:44:31 -07:00
Nuno CamposandGitHub a6e66746f7 Make compatible with langchain-core 0.1 by conditionally importing _StreamingCallbackHandler (#4203) 2025-04-08 13:28:01 -07:00
Vadym BardaandGitHub 3a17df6106 langgraph: release 0.3.26 (#4204) 0.3.26 2025-04-08 14:53:10 -04:00
Nuno Campos cee6a450dc Lint 2025-04-08 10:52:28 -07:00
Nuno Campos 0b3bf37a55 Fix the rest 2025-04-08 10:47:55 -07:00
Nuno CamposandGitHub 305a676675 langgraph: raise GraphInterrupt only if used as a subgraph (#4202) 2025-04-08 10:45:45 -07:00
Nuno Campos 5b73e38c38 Make compatible with langchain-core 0.1 by conditionally importing _StreamingCallbackHandler 2025-04-08 10:44:12 -07:00
vbarda 41fb5ec77c Revert "add warning"
This reverts commit cff349e22e.
2025-04-08 13:38:49 -04:00
vbarda cff349e22e add warning 2025-04-08 13:34:47 -04:00
vbarda 8f32fc4819 update tests 2025-04-08 13:28:35 -04:00
vbarda 5690555394 langgraph: raise GraphInterrupt only if used as a subgraph 2025-04-08 12:51:36 -04:00
Tat Dat Duong 7d5621a84f Default type for TExtra 2025-04-08 17:10:43 +02:00
Tat Dat Duong 5f1213a1c7 Remove extra 2025-04-08 17:07:04 +02:00
Tat Dat Duong a98f9542fa Add unused extra generic for future typing of metadata 2025-04-08 17:06:51 +02:00
William FHandGitHub c5b118a672 Add admonitions about managed checkpointers (#4197)
If you're deploying with langgraph API, you don't need to manually
define a checkpointer. For folks who already know they'll be developing
with the api server, I'd like to save everyone time by making this more
clear in the docs on checkpointing.
2025-04-08 12:16:24 +00:00
Tat Dat Duong aee39605e0 Add missing types 2025-04-08 14:13:59 +02:00
lc-arjunandGitHub 72bec9161a Release js sdk 0.0.63 (#4192) 2025-04-07 18:40:59 -07:00
Nuno CamposandGitHub ae17e77522 feat: add assistant description to js sdk (#4191) 2025-04-07 18:37:49 -07:00
Arjun Natarajan a96fc75c55 add assistant description to js sdk 2025-04-07 21:06:34 -04:00
Nuno Campos d541ed90d5 Save Sends unconditionally 2025-04-07 16:45:26 -07:00
Tat Dat Duong c757247858 feat(sdk-js): add auth types 2025-04-07 20:29:14 +02:00
Nuno Campos 5a0228cb13 Add test 2025-04-04 16:00:28 -07:00
Nuno Campos 4abfc7702d Subgraphs inherit checkpoint mode 2025-04-04 16:00:22 -07:00
Nuno Campos a5495e84c8 Add another test 2025-04-04 15:42:11 -07:00
Nuno Campos 4f353dac31 Fix assignment of pending writes 2025-04-04 14:41:10 -07:00
Nuno CamposandGitHub 4c89bb39d4 Add benchmark script for typed dict version of existing wide state benchmark (#4174)
- to easily compare perf impact of using pydantic, data class, or typed
dict for same workload
2025-04-04 18:37:57 +00:00
Eugene YurtsevandGitHub 05a4fcc8bb cli: release 0.1.89 (#4173)
Release to pick up this: https://github.com/langchain-ai/langgraph/pull/4164
cli==0.1.89
2025-04-04 13:46:47 -04:00
Nuno Campos 7ebd6f5e1f Better test 2025-04-04 10:01:35 -07:00
Eugene YurtsevandGitHub adac016e33 cli: support dict format for graph specification in langgraph.json (#4164)
Allow the CLI to work with dict format for the graph specification.

```json
{
  "dependencies": ["./my_agent"],
  "graphs": {
    "agent": {
      "path": "./my_agent/agent.py:graph",
      "description": "this is my agent description"
    }
  },
  "env": ".env"
}
```

And backwards compatible with:

```json
{
  "dependencies": ["./my_agent"],
  "graphs": {
    "agent": "./my_agent/agent.py:graph",
  },
  "env": ".env"
}
```
2025-04-04 10:16:20 -04:00
Nuno Campos 0a1dd7a01a Do same thing for writes 2025-04-03 17:31:34 -07:00
Nuno Campos 7e08339335 mypy is dumb 2025-04-03 16:55:23 -07:00
Nuno Campos e1d4b5552d Add checkpoint_during arg
- This provides a new mode of execution where only the last checkpoint is saved
- We save the last checkpoint no matter how the agent run is terminated (success, error, interrupt, etc)
- This cuts down on cpu time spent on checkpointing, while not losing any resilience benefits, given individual task writes are still saved
- If an error occurs and the run is retried, any tasks that completed successfully before will be skipped (as currently)
- checkpoint_during=True is useful when you want to time-travel to inner steps of a run
- The default value will remain the current behavior, ie. checkpoint_during=True
2025-04-03 16:51:53 -07:00
David DuongandGitHub 2d13904abf release(langgraph): 0.3.25 (#4167) 0.3.25 2025-04-03 22:20:03 +02:00
Tat Dat Duong dfeb9d3b46 release(langgraph): 0.3.25 2025-04-03 22:12:16 +02:00
David DuongandGitHub 81935a73d8 feat(langgraph): Add UI messages API (#4157)
Sample usage:

```python
from typing import Annotated, Sequence, TypedDict

from langchain_core.messages import BaseMessage
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph
from langgraph.graph.message import add_messages

from langgraph.graph.ui import AnyUIMessage, ui_message_reducer, push_ui_message


class AgentState(TypedDict):  # noqa: D101
    messages: Annotated[Sequence[BaseMessage], add_messages]
    ui: Annotated[Sequence[AnyUIMessage], ui_message_reducer]


async def agent(state: AgentState):  # noqa: D103
    message = await ChatOpenAI(model="gpt-4o-mini", temperature=0).ainvoke(
        state["messages"]
    )

    # Also directly writes the message to `ui`
    push_ui_message("simple", {"hello": "123"}, message=message, state_key="ui")

    return {"messages": [message]}

```
2025-04-03 22:10:18 +02:00
Tat Dat Duong 615fc8b4ae Update naming 2025-04-03 21:58:16 +02:00
William FHandGitHub 13e6f6cbde Add algolia site verification (#4165) 2025-04-03 12:53:40 -07:00
Vadym BardaandGitHub e89633f30b prebuilt: release 0.1.8 (#4161) prebuilt==0.1.8 2025-04-03 12:01:18 -04:00
Vadym BardaandGitHub 0bbf5829e8 docs: add a how-to guide for managing message history in create_react_agent (#4149) 2025-04-03 16:00:11 +00:00
David DuongandGitHub 7ed5288f8f release(cli): 0.1.84 (#4158) cli==0.1.84 2025-04-03 15:25:12 +02:00
Tat Dat Duong cba240e70e release(cli): 0.1.84 2025-04-03 15:15:52 +02:00