chore(llma): update SDKs #367

carlos-marchal-ph · 2025-11-06T14:45:31Z

Update LLM provider SDKs to latest major versions

openai: 1.102.0 → 2.7.1
anthropic: 0.64.0 → 0.72.0
google-genai: 1.32.0 → 1.49.0
langchain-core: 0.3.75 → 1.0.3
langchain-openai: 0.3.32 → 1.0.2
langchain-anthropic: 0.3.19 → 1.0.1
langchain-community: 0.3.29 → 0.4.1
langgraph: 0.6.6 → 1.0.2

Tested locally with llm-analytics-tools to verify the major updates didn't break anything.

greptile-apps

_{6 files reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}

andrewm4894 · 2025-11-07T14:03:18Z

posthog/ai/langchain/callbacks.py

@@ -1,8 +1,8 @@
 try:
-    import langchain  # noqa: F401
+    import langchain_core  # noqa: F401


@carlos-marchal-ph does this mean we would not support LC <1.* on latest posthog sdk?

That's correct. That's why it's also a major release. It's not ideal but at least it gives user a choice, they can still stay on 6.9. I guess we could backport it and export it as langchain_v0.

hmm just trying to think through users who say have some agents working away using pre 1.0 LC but are doing all new agent dev in LC 1+

what would they do?

i think maybe would be problems for them?

not against any of this but just thinking through what if or worse case scenarios if lots of users start complaining

Nono, makes a lot of sense. I will implement an exported v0 wrapper. A shame that we are shipping all that code though, we really need to rethink how we approach the wrappers at some point.

yeah 100% going to only get more complicated

hmm just trying to think through users who say have some agents working away using pre 1.0 LC but are doing all new agent dev in LC 1+

actually not even sure how possible that is actually - it might sort of be.

@carlos-marchal-ph one option is to just go as-is but we just be ready with a draft pr for the v0 wrapper as a backup in case needed maybe?

eg maybe we just accept that we wont know (since we dont actually have version tracking in the sdk) and just be ready for if it does mess up a lot of folks

That makes sense too. In any case won't merge until Monday just to be safe.

andrewm4894

lgtm - just want to be ready if we do need a follow on PR if we do end up seeing lots of users have issues and needing backwards compatibility with langchain for a bit longer. if so i think we can handle it in a follow on pr perhaps.

we are a bit blind as we dont have version from the sdk so not really any good way for us to know.

so its better on balance i think to try be a bit cleaner like we are in the pr but just be ready for if it does cause issues for some folks.

pauldambra · 2025-11-10T15:32:36Z

CHANGELOG.md

i've not done a release of the python SDK for ages.

i'd be tempted to make it clear in this changelog note that this removes python 3.9 support.

i'd guess there's no user action on how they use the SDK, just whether we explicitly support the version, so no need for a "how to upgrade from 6 to 7" note

if we wanted to be very neat then we'd have a PR that removes 3.9 support, and then a PR that does the llma upgrades and we release those two changes as v7 but imho just making it clear in the changelog is enough

CHANGELOG.md

Co-authored-by: Paul D'Ambra <paul.dambra@gmail.com>

carlos-marchal-ph · 2025-11-10T17:37:54Z

Added support for Python 3.14. The only issue is that it's not compatible with Pydantic 1.X.

Pydantic is not a direct dependency of our SDK but we do accept Pydantic objects as input from users in case they are using it. I added a warning if we get passed a Pydantic V1 object on Python 3.14, which shouldn't happen with proper dependency management, but just in case.

pauldambra · 2025-11-10T19:02:09Z

pyproject.toml

    "types-six",
    "pre-commit",
-    "pydantic",
+    "pydantic>=2.12",


so... in order to support 3.14
we're dropping support for pydantic 1.0

if you want to use python <14 you have to upgrade pydantic anyway

let's make that clear in the changelog too... i can imagine that being a pain for people but i also don't know if there's a better way to approach the versioning/support

it would be fine to separate support 3.14 from this PR so you can ship the previous thing

I'm going to separate the PR to make sure to get this right. I'm not sure we really need to drop support for V1 for other versions of Python, as long as we make sure that we are testing against objects from both versions except for 3.14, which as you already said is intrinsically incompatible with V1.

Added support for 3.14 without dropping Pydantic V1 for the rest of versions in #373.

pauldambra · 2025-11-10T19:02:35Z

CHANGELOG.md

@@ -1,3 +1,17 @@
+# 7.0.0 - 2025-11-10
+
+NB Python 3.9 is no longer supported


let's also make clear here that we're dropping pydantic 1.0 support too

chore: update LLM SDKs

5249d02

carlos-marchal-ph requested a review from a team November 6, 2025 14:45

carlos-marchal-ph self-assigned this Nov 6, 2025

carlos-marchal-ph added the team/llm-analytics label Nov 6, 2025

carlos-marchal-ph changed the title ~~chore: update LLM SDKs~~ chore(llma): update SDKs Nov 6, 2025

greptile-apps bot reviewed Nov 6, 2025

View reviewed changes

fix: fully deprecate Python 3.9

3564d58

andrewm4894 reviewed Nov 7, 2025

View reviewed changes

Merge branch 'master' into chore(llma)/update-sdks

498d1cc

andrewm4894 approved these changes Nov 7, 2025

View reviewed changes

Merge branch 'master' into chore(llma)/update-sdks

ab1d79b

carlos-marchal-ph enabled auto-merge (squash) November 10, 2025 10:49

carlos-marchal-ph disabled auto-merge November 10, 2025 15:14

empty commit to unstuck CI

bdb374e

pauldambra reviewed Nov 10, 2025

View reviewed changes

CHANGELOG.md Outdated Show resolved Hide resolved

carlos-marchal-ph and others added 4 commits November 10, 2025 16:35

Update CHANGELOG.md

fb6ce16

Co-authored-by: Paul D'Ambra <paul.dambra@gmail.com>

feat: add python 3.14 compatibility

46f55e0

fix: notify users if using incompatible pydantic & python combo

157d53c

fix: formatting

34cdb98

carlos-marchal-ph requested a review from pauldambra November 10, 2025 17:38

pauldambra reviewed Nov 10, 2025

View reviewed changes

carlos-marchal-ph added 3 commits November 11, 2025 11:00

revert: python 3.14 support

fb7d0f6

Merge branch 'master' into chore(llma)/update-sdks

731276e

Merge branch 'master' into chore(llma)/update-sdks

826fe9d

carlos-marchal-ph enabled auto-merge (squash) November 11, 2025 18:10

carlos-marchal-ph merged commit 4820336 into master Nov 11, 2025
10 checks passed

carlos-marchal-ph deleted the chore(llma)/update-sdks branch November 11, 2025 18:12

		@@ -1,3 +1,17 @@
		# 7.0.0 - 2025-11-10

		NB Python 3.9 is no longer supported

chore(llma): update SDKs #367

chore(llma): update SDKs #367

Conversation

carlos-marchal-ph commented Nov 6, 2025

Uh oh!

greptile-apps bot left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

andrewm4894 left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

carlos-marchal-ph commented Nov 10, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

andrewm4894 left a comment •

edited

Loading