refactor(langchain): refactor optional imports logic in langchain_v1

[cbornet]

• Use importlib to load dynamically the classes
• Removes missing package warnings

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Preview	Comments	Updated (UTC)
langchain	Ignored	Preview		Sep 11, 2025 2:30pm

[codspeed-hq]

CodSpeed WallTime Performance Report

Merging #32813 will not alter performance

_{Comparing cbornet:langchain_v1-refactor-imports (21a3b8f) with master (af17774)}

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

Summary

✅ 13 untouched

[codspeed-hq]

CodSpeed Instrumentation Performance Report

Merging #32813 will not alter performance

_{Comparing cbornet:langchain_v1-refactor-imports (21a3b8f) with master (af17774)}

Summary

✅ 14 untouched

[eyurtsev]

I think we don't want to load things on __init__ -- we want to have as fast of an import of langchain / langchain.chat_models as possible. Users would ideally only pay an import cost for the things they use

Refactor OK, but let's make it lazy and memoize appropriately?

[cbornet]

Ruff rule PLC0415 recommends to do the imports at the module loading.
I think the langchain_* modules are quite fast to import, aren't they ?

[eyurtsev]

it's been a significant problem which is why we've swapped to lazy imports in many places. I think we saw import time grow to ~0.5 seconds, which is way too slow

[eyurtsev]

So we want as little of the code loaded up front, so that end of the day only the parts of the code that the user needs are loaded

[cbornet]

Ok. I can switch to a cached lazy load then.
Note that the big load time may come from one package alone. It would be interesting to identify it.

[cbornet]

done.