feat(langchain): middleware support in create_agent
#32828
Merged
+2,655
−293
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
CodSpeed WallTime Performance ReportMerging #32828 will not alter performanceComparing
|
CodSpeed Instrumentation Performance ReportMerging #32828 will not alter performanceComparing Summary
|
sydney-runkle
added
the
codspeed-ignore
This was referenced Sep 5, 2025
sydney-runkle
force-pushed
the
sr/new-agent
branch
from
e04d329 to
26bef49
Compare
hwchase17
reviewed
Sep 6, 2025
| messages = state["messages"] | ||
| self._ensure_message_ids(messages) | ||
|
|
||
| total_tokens = self.token_counter(messages) |
There was a problem hiding this comment.
do we want to cache this or something? recomputing kind of annoying?
libs/langchain_v1/langchain/agents/middleware/human_in_the_loop.py
Outdated
Show resolved
Hide resolved
…32837) Main change here is the removal of `AgentJump`. Typing things as `AgentJump` doesn't really make sense, given that we often want to perform an agent jump **and** update the state of the graph, potentially including updates to state extensions from a given middleware. For example, previously we had: ```py def after_model(state: AgentState) -> AgentUpdate | AgentJump | None: ... ``` But what if my middleware adds a `model_calls` attribute? `AgentJump` has no way of inheriting that newly defined state, but I still might want to do something like: ```py def after_model(state: AgentState) -> AgentUpdate | AgentJump | None: return {'model_calls': X, 'jump_to': '__end__'} ``` Then the return type `AgentUpdate | AgentJump | None` doesn't represent the return value. There's no great way to type partial state updates yet, hence we use `dict[str, Any]`. This is a Python limitation. I've removed the conditional logic for signature parsing on `AgentJump` and we just always add conditional edges for middleware nodes now. Other things: * adding docstrings + other linting fixes (sort of a general cleanup of this module). * removing tool calling / model calling limits as those applications aren't fully fleshed out / we want to just publish these 3 middlewares first.
sydney-runkle
changed the title
create_agent
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Overview
Adding new
AgentMiddlewareprimitive that supportsbefore_model,after_model, andprepare_model_requesthooks.This is very exciting! It makes our
create_agentprebuilt much more extensible + capable. Still in alpha and subject to change.This is different than the initial implementation in that it:
TypedDictforAgentState-- dataclass subclassing is tricky w/ required values + required decoratorsmodel_settingstoModelRequestso that we can pass through things to bind (like cache kwargs for anthropic middleware)TODOs
top prio
other ones
secondary prio
AgentUpdateandAgentJump, at least in Python)docs
See langchain-ai/docs#390
open questions
Lots of open questions right now, many of them inlined as comments for the short term, will catalog some more significant ones here.