[Feature]: Add missing post_call hook to Lakera guardrails

[eurogig]

The Feature

Current State

async_pre_call_hook implemented (pre_call)
async_moderation_hook implemented (during_call)
async_post_call_success_hook is missing
...therefore the post_call mode doesn't work

Target State

All three hooks implemented
post_call mode fully functional
Monitor mode works with post_call hook

Motivation, pitch

Lakera is excellent as detecting PII exflitration and harmful generated content inoutgoing content from an LLM. For some unknown reason the initial implementation of Lakera support only added pre_call and during_call support. Let's fix that so we can enjoy the full benefits of Lakera's security.

LiteLLM is hiring a founding backend engineer, are you interested in joining us and shipping to all our users?

No

Twitter / LinkedIn details

@SteveGiguere / https://www.linkedin.com/in/stevegiguere/

[hamzaq453]

Hi @eurogig ! I'm interested in implementing this feature. I'll add the async_post_call_success_hook to complete the Lakera guardrails integration. Let me know if there are any specific implementation considerations I should be aware of. Otherwise, I'll start working on it and submit a PR.