builders · review · first seen Nov 4, 10:33 UTC

LangGraph SDK removes fairness_weight and fixes run stream rejoining

LangGraph SDK 0.4.12 shipped with a removed PrioritySettings field and a fix for rejoining a run stream that closes before the run ends. The decision for builders is whether either change touches code they run today.

L0 · primary1 source3 claims checked4 min read

What shipped

LangGraph SDK 0.4.12 was released. The supplied release excerpt lists two changes since sdk==0.4.11: removal of fairness_weight from PrioritySettings and a fix for rejoining a run stream that closes before the run ends. Both are shipped changes, not scheduled work. That makes this a version to evaluate against current SDK usage rather than a feature to plan around before release. The excerpt does not describe any other changes in 0.4.12, so this review does not treat it as a broader account of the release. [1][2][3]

The PrioritySettings change is a removal: fairness_weight is no longer part of that setting in 0.4.12. For a builder who supplies that field, the immediate review question is where it appears in their code and configuration, and what they intend to do without it. The release excerpt does not name a replacement field or describe how priorities behave after the removal. It also does not say what happens if existing code still supplies fairness_weight. Those points need checking before treating the version change as a routine dependency update. [1]

The stream change addresses a narrower failure case. Version 0.4.12 fixes rejoining a run stream that closes before the run ends. The important distinction is between the stream closing and the run ending: the cited fix concerns a stream that closes while its run has not ended. The release excerpt does not explain how the fix works or specify every circumstance in which a stream can close. Builders should therefore test the case they actually rely on, rather than infer a general guarantee about all stream interruptions. [1]

Why it matters if you build

If your workflow constructs PrioritySettings with fairness_weight, moving to 0.4.12 requires a decision about that use. Start by finding each place the field is supplied and identifying what that setting was meant to accomplish in your application. The confirmed change is removal of the field, not preservation of its effect under another name. Until you have checked the behavior you need, deleting the field merely to get past an update would leave that intent unexamined. If you do not use fairness_weight, this particular change may require no code edit, though that is a conclusion to verify in your own usage. [1]

If your workflow consumes a run stream and needs to rejoin after it closes before the run ends, the fix is directly relevant. A useful evaluation is to reproduce that sequence with the SDK version you use now, then repeat it with 0.4.12 and compare the result. That is a test suggestion, not a claim that the failure will occur in every application or that the fix covers every closure. The supplied excerpt gives no performance measure, operational cost, or limit for rejoining, so none can be put into an upgrade calculation from this source alone. [1][2]

These changes can pull an upgrade decision in different directions. The stream fix gives affected builders a reason to evaluate 0.4.12 now; the PrioritySettings removal gives builders who use fairness_weight a reason to inspect their dependency before adopting it. Neither point establishes the size of the migration or the value of the fix for an application that has not encountered the stated stream case. The practical cost is the time needed to check your own settings and stream behavior. The excerpt supplies no estimate for that work, so the decision rests on those checks rather than a published cost figure. [1][2][3]

How to try it today

1. Find uses of PrioritySettings and fairness_weight in the code and configuration you would update.
2. For each use, record what behavior you need; do not assume the excerpt names a replacement.
3. Evaluate LangGraph SDK 0.4.12 against that usage before adopting it.
4. If you consume run streams, test rejoining one that closes before its run ends.
5. Decide on the update only after checking both the setting you use and the stream case you depend on.

What we checked

Every claim below was walked down the source ladder. The rung is how close the source is to the thing itself; the tense is the source's own.

1. shipped L0 · primary

   LangGraph SDK 0.4.12 was released.

   per GitHub releases · langchain-ai/langgraph (L0)
2. shipped L0 · primary

   LangGraph SDK 0.4.12 removes fairness_weight from PrioritySettings.

   per GitHub releases · langchain-ai/langgraph (L0)
3. shipped L0 · primary

   LangGraph SDK 0.4.12 fixes rejoining a run stream that closes before the run ends.

   per GitHub releases · langchain-ai/langgraph (L0)

What is not confirmed

The supplied release excerpt does not establish a replacement for fairness_weight, the behavior of code that still supplies it, or the resulting priority behavior. It does not specify the stream fix's implementation, its limits, or whether it covers closure cases other than a stream closing before its run ends. It provides no performance results, cost figures, or migration instructions. Those are open questions for a builder's own evaluation, not details that can be inferred from the two release-note lines. [1][2]

Who should act, who should wait

Evaluate 0.4.12 now if you rely on rejoining a run stream after it closes before the run ends; the shipped fix addresses that stated case. Also review it now if your code supplies fairness_weight, because the shipped release removes that field from PrioritySettings. Wait before adopting it if you need an established replacement for fairness_weight or an assurance about stream cases beyond the one named: neither is provided here. If neither change touches your workflow, these claims alone do not establish an urgent reason to update today. [1][2][3]

Sources

• langgraph-sdk==0.4.12 — GitHub releases · langchain-ai/langgraph L0 · primary