Horrible memory footprint...

[false-vacuum]

Why does just importing this use 500+ mib of memory? That's insane.

Steps to reproduce

• check memory usage.
• Import vapi

from vapi import Vapi

• Check memory usage.
• You just leaked half a gig of memory.

[leandermaerkisch]

I faced the same issue. Deployment crashed because OOM error. Took a while to figure this out. Would be great if this is fixed soon

[false-vacuum]

@leandermaerkisch The latest pydantic causes an unbounded memory leak. If you pin pydantic==2.10.6, it should fix this.... This auto-generated SDK really needs some work.

[leandermaerkisch]

That's great news, thanks a bunch @false-vacuum!

[dsinghvi]

@leandermaerkisch @false-vacuum we just reduced some of the sizes of __init__.py to reduce the footprint by ~3x. We'll be making an update later to use typedicts instead of pydantic models to further reduce the memory pressure.

Sorry about the inconvenience so far!

[leandermaerkisch]

Thanks for fixing this rapidly!

…

On Fri, 4 Apr 2025 at 01:53, Deep Singhvi ***@***.***> wrote: @leandermaerkisch <https://github.com/leandermaerkisch> @false-vacuum <https://github.com/false-vacuum> we just reduced some of the sizes of __init__.py to reduce the footprint by ~3x. We'll be making an update later to use typedicts instead of pydantic models to further reduce the memory pressure. Sorry about the inconvenience so far! — Reply to this email directly, view it on GitHub <#11 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AIZFPQ45KI6INDROBV7BB5L2XXJYTAVCNFSM6AAAAABYTJJDIKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDONZXGMYDENZXGA> . You are receiving this because you were mentioned.Message ID: ***@***.***> [image: dsinghvi]*dsinghvi* left a comment (VapiAI/server-sdk-python#11) <#11 (comment)> @leandermaerkisch <https://github.com/leandermaerkisch> @false-vacuum <https://github.com/false-vacuum> we just reduced some of the sizes of __init__.py to reduce the footprint by ~3x. We'll be making an update later to use typedicts instead of pydantic models to further reduce the memory pressure. Sorry about the inconvenience so far! — Reply to this email directly, view it on GitHub <#11 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AIZFPQ45KI6INDROBV7BB5L2XXJYTAVCNFSM6AAAAABYTJJDIKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDONZXGMYDENZXGA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

-- Best regards / Mit freundlichen Grüßen, Leander Märkisch