WEBDIENSTE/langgraph
1
0
Fork 0
mirror of https://github.com/langchain-ai/langgraph.git synced 2026-08-24 16:42:24 +02:00
Code Issues Packages Projects Releases Wiki Activity
6,682 Commits 322 Branches 571 Tags
2d338664475f46d508d4d0976eacfa86d1094de9
Commit Graph
3338 Commits
Author SHA1 Message Date
Nuno Campos 2cb81201b2 0.3.24 2025-04-02 15:45:06 -07:00
Nuno Campos 494cd4f6ad checkpoint 2.0.24 2025-04-02 15:44:58 -07:00
Nuno CamposandGitHub 9226b42150 Add checkpoint migrations for state graph internal channels (#4125) 2025-04-02 15:42:28 -07:00
Eugene YurtsevandGitHub 5e0a843423 sdk-py: release 0.1.61 (#4148) 2025-04-02 15:34:36 -07:00
Eugene YurtsevGitHubWilliam FH
 		6e54f74fa9 sdk-py: Add option to set description via client sdk (#4147) 
Co-authored-by: William FH <13333726+hinthornw@users.noreply.github.com>
 2025-04-02 21:26:14 +00:00
Nuno Campos ffb8400b19 Disable tests in 3.9 2025-04-02 14:05:15 -07:00
Nuno Campos b6b40149a9 Fix 2025-04-02 13:56:29 -07:00
Nuno Campos 5eecd2ac3a Update 2025-04-02 13:31:36 -07:00
Nuno Campos c38f7be89c Migrate quadratic channels too 2025-04-02 13:24:20 -07:00
Nuno Campos 2d406f03a2 Lint 2025-04-02 12:56:09 -07:00
Nuno Campos 1c48ec0aba Lint 2025-04-02 12:48:10 -07:00
Nuno Campos 4dda404da3 Implement checkpoint migration 
- Migrate start:{node} channels to branch:to:{node}
- Migrate {node} channels to branch:to:{node}
 2025-04-02 12:44:59 -07:00
Vadym BardaandGitHub 67e8f8fc11 prebuilt: add optional pre-model hook that runs before calling LLM in create_react_agent (#4059) 
Example:

```python
from typing import Any
from langchain_openai import ChatOpenAI
from langchain_community.tools.tavily_search import TavilySearchResults
from langchain_core.messages import AnyMessage
from langchain_core.messages.utils import count_tokens_approximately

from langgraph.graph import MessagesState
from langgraph.prebuilt.chat_agent_executor import create_react_agent, AgentState
from langgraph.checkpoint.memory import InMemorySaver

from langmem.short_term import SummarizationNode, RunningSummary


class State(MessagesState):
    context: dict[str, Any]

search = TavilySearchResults(max_results=3)
tools = [search]

model = ChatOpenAI(model="gpt-4o")
summarization_model = model.bind(max_tokens=256)

summarization_node = SummarizationNode(
    token_counter=count_tokens_approximately,
    model=summarization_model,
    max_tokens=2048,
    max_summary_tokens=256,
    output_messages_key="messages"
    # output_messages_key="llm_input_messages"
)

checkpointer = InMemorySaver()


class State(AgentState):
    user_language: str

    # summarization-related keys
    context: dict[str, Any]


def prompt(state):
    language = state["user_language"]
    system_msg = f"Always respond in {language}"
    return [{"role": "system", "content": system_msg}] + state["messages"]


graph = create_react_agent(
    model,
    tools,
    prompt=prompt,
    pre_model_hook=summarization_node,
    state_schema=State,
    checkpointer=checkpointer
)
```
 2025-04-02 15:41:31 +00:00
Nuno CamposandNuno Campos 6fe319ed1b WIP: Add migrate_checkpoint 2025-04-02 07:24:55 -07:00
Vadym BardaandGitHub 55f922cf2f langgraph: release 0.3.23 (#4141) 2025-04-02 09:50:18 -04:00
Nuno Campos 6efeefe424 Maintain checkpoint LATEST_VERSION constant in langgraph lib 
- This should be controlled by the langgraph version, not the version of langgraph-checkpoint installed
 2025-04-01 21:46:38 -07:00
William FHandGitHub da8b8c606a Release CLI (#4124) 
Signed-off-by: William Fu-Hinthorn <13333726+hinthornw@users.noreply.github.com>
 2025-04-01 21:28:58 -07:00
William FHandGitHub dcda8c24d6 Add checkpointer configuration support in langgraph.json (#4122) 
Signed-off-by: William Fu-Hinthorn <13333726+hinthornw@users.noreply.github.com>
 2025-04-01 21:16:12 -07:00
Vadym BardaandGitHub e3d697620b langgraph: support removing all messages with RemoveMessage (#4117) 2025-04-01 17:51:54 -04:00
Nuno Campos 30883729f0 0.3.22 2025-04-01 07:54:24 -07:00
William Fu-Hinthorn 9bd78ed483 Allow blocking in dev 
Signed-off-by: William Fu-Hinthorn <13333726+hinthornw@users.noreply.github.com>
 2025-04-01 05:45:47 -07:00
Nuno Campos e85b7e6cd9 Lint 2025-03-31 18:46:51 -07:00
Nuno Campos 94fa46f9fa Avoid creating checkpoint unless we're saving it 
- When checkpointing is disabled don't call create_checkpoint in PregelLoop
- In local_read apply writes directly to copies of updated channels
- Add BaseChannel.copy() method to create channel copies with less overhead
 2025-03-31 18:37:13 -07:00
Nuno Campos fd64ada9de Avoid raise-catch strategy in BaseChannel.checkpoint() 
- This mirrors the work done earlier on BaseChannel.get()
- Comparing to a sentinel value is significantly faster than raising and catching an exception
 2025-03-31 17:48:19 -07:00
Nuno Campos 5e9e7b79fe Lint 2025-03-31 17:36:00 -07:00
Nuno Campos c49a077789 Lint 2025-03-31 17:24:26 -07:00
Nuno Campos 881b07cf7f Lint 2025-03-31 16:36:16 -07:00
Nuno Campos 0425d4e65d Update 2025-03-31 16:29:07 -07:00
Nuno Campos 118016a21c Add fast path to serialize None values 2025-03-31 16:15:56 -07:00
Nuno Campos 067b99c789 Lazily create atomic counters in pregel scratchpad 
- many times these aren't actually used, so makes sense to delay creation until needed
 2025-03-31 15:19:53 -07:00
Tat Dat Duong aef463c633 fix(sdk-js): send accepts any input 2025-04-01 00:15:32 +02:00
Nuno Campos e9e9a96a0d Reduce perf impact of pregel scratchpad creation 
- make scratchpad class frozen now that its members are never reassigned
- replace next(gen expr) with for-loop to avoid allocating generator objects
 2025-03-31 14:50:10 -07:00
Tat Dat Duong f629f68ec3 feat(cli): add support for gen ui config 2025-03-28 23:05:04 +01:00
Tat Dat Duong 2808a7859a Bump to 0.0.62 2025-03-28 21:39:44 +01:00
Tat Dat Duong fb3c61ea4f feat(sdk-js): run optimistic values mutator before any network request 2025-03-28 21:39:35 +01:00
Tat Dat Duong 9abc1c8174 release(sdk-js): 0.0.61 2025-03-28 15:23:25 +01:00
Tat Dat Duong b8b973fc0c fix(sdk-js): do not await for client.runs.stream, as it is already async generator 2025-03-28 15:15:54 +01:00
David DuongandGitHub 5401d2ea81 feat(sdk-js): add option to manually provide implementation for shared modules (#4042) 2025-03-28 14:04:09 +01:00
Nuno Campos 49bb08a3f9 Update prebuilt test 2025-03-27 17:58:29 -07:00
Nuno Campos 4f2e9b838f Reduce the number of channels created for each node by 50% 
- Used to be 2 channels per node, it is now one per node, which is the minimum
- Now both hard edges, conditional edges, entrypoint and conditional entrypoint all use the same channel to trigger a node
 2025-03-27 17:47:34 -07:00
William FHandGitHub 7f079adfee Update auth user type (#4062) 2025-03-27 16:09:54 -07:00
William Fu-Hinthorn 4c74af606f Update auth user type 2025-03-27 16:03:11 -07:00
Nuno CamposandGitHub 522caa643f In Python 3.12 or above, use asyncio eager task factory (#4055) 
- This is a performance improvement when calling async functions that do
not use await, as they are run immediately and never scheduled in the
loop
 2025-03-27 13:36:04 -07:00
Nuno Campos ef71656f05 Fix 2025-03-27 13:14:39 -07:00
Tat Dat Duong 3ba7c7fbed Do not throw error if window is undefined due to Next 2025-03-27 21:03:54 +01:00
Tat Dat Duong 5cdab86d48 Fix object assignment 2025-03-27 20:58:39 +01:00
Nuno CamposandGitHub f5fe7e5195 Remove internal frames from stack traces (#4054) 
- For exceptions raised in user code (ie. nodes or edges) remove
internal frames from the stack trace
 2025-03-27 12:41:54 -07:00
Nuno Campos 1c5a354a7d In Python 3.12 or above, use asyncio eager task factory 
- This is a performance improvement when calling async functions that do not use await, as they are run immediately and never scheduled in the loop
 2025-03-27 11:00:37 -07:00
Nuno Campos bfd271e00e Remove internal frames from stack traces 
- For exceptions raised in user code (ie. nodes or edges) remove internal frames from the stack trace
 2025-03-27 10:58:13 -07:00
Vadym BardaandGitHub 9647b1e55f langgraph: use correct type for node destination annotations (#4053) 
Fixes https://github.com/langchain-ai/langgraph/issues/4051
 2025-03-27 16:57:39 +00:00