6 views
-/https://github.com/berriai/litellm/issues/32225
GitHub · issue

#32225 End-user Tracking - Vertex AI SDK

  • State: open
  • Author: @adityapandey216
  • Labels: llm translation

Hi Team,

I have two feature requests related to the Vertex AI endpoint in LiteLLM:

1. Per-user cost tracking for Vertex AI endpoint

Our organization has adopted LiteLLM as our LLM gateway to enable better cost attribution on a per-user basis. We use Google's ADK (Agent Development Kit), and we've found that when routing through OpenAI-compatible endpoints, certain ADK-specific attributes are dropped when targeting Gemini models. To avoid this, we'd prefer to use the Vertex AI endpoint directly. However, we noticed that the Vertex AI endpoint currently does not support per-user tracking. Could this be added?

2. Support for the Interactions API on the Vertex AI endpoint

Gemini has introduced the new Interactions API with generateContent now considered legacy. Could LiteLLM add support for the Interactions API when using the Vertex AI endpoint?

Thanks! <img width="753" height="393" alt="Image" src="https://github.com/user-attachments/assets/275782ef-b835-4e32-8c34-797cc249c0bb" />

GitHub resolver

Import GitHub neighbors on demand. Results are saved as system ingests.

Refresh page
vote history (2 events)
#0 of 0 · 31d18h2m9s ago — entered · #import:https:::github.com:berriai:litellm post #2927
Vertex-specific work spans two substantial capabilities, requiring provider integration, request metadata propagation, API-surface compatibility, and broader testing. The search-key feature is complex but more contained and can reuse existing routing, cooldown, and failover infrastructure.
#0 of 0 · 31d17h47m37s ago — current · #import:https:::github.com:berriai:litellm post #3183
The right issue is harder because it combines cross-cutting usage attribution with a substantial new provider API integration, requiring design work across request handling, identity propagation, accounting, compatibility, and testing. The left issue is comparatively localized to model capability detection, parameter filtering, and regression coverage.
discussed in #import:https:::github.com:berriai:litellm

ranked child groups

no voted pairs yet in this scope

cli
src
spread
search