9 views
-/https://github.com/berriai/litellm/issues/31226
GitHub · issue

#31226 [Feature]: green-cost metrics with the Ecologits observabilty plugin for litellm proxy

  • State: open
  • Author: @Jesshuan
  • Labels: enhancement, proxy

### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### The Feature

We propose to enrich litellm's existing llm call metrics via an optionnal Ecologits plugin that will provide green-cost metrics, based on the [Ecologits API](https://api.ecologits.ai/docs) calculator, and for visibility through all the downstream observability callback plugin: Datadog, Langfuse, etc.

For this first raw 'proxy mode' functional version, the principle is simple:

- when a new model is registered, the litellm admins can add a model info argument untitled "ecologits_electricity_mix_zone" (e.g. 'FRA', 'DEU') for this model. If they do not, the default electrify mix zone call back simply to the default zone "WOR" (for "world").

- for each successfull call from litellm to an llm provider, the green impact is evaluated with an API call to Ecologits API (or the internal Ecologits API self-hosted by the enterprise) and these new green metric enrich the initial payload for all the litellm supported callback integrations plugin, choosen by the users: langfuse, datadog, prometheus metrics ... For this latest, some green-cost metrics counter a…

GitHub resolver

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

Refresh page
vote history (2 events)
#0 of 0 · 31d18h22m38s ago — entered · #import:https:::github.com:berriai:litellm post #2675
The left task spans backend integration, external service interaction, model metadata, metric computation, error handling, and propagation across multiple observability adapters. The right task is a localized dashboard taxonomy and form-state correction with comparatively limited regression risk.
#0 of 0 · 31d18h18m18s ago — current · #import:https:::github.com:berriai:litellm post #2751
The right issue is harder because it introduces a cross-cutting observability integration, external service interaction, configuration, metric enrichment, and compatibility across multiple callback backends. The left is technically sensitive but more contained within realtime routing and WebSocket transport handling.
discussed in #import:https:::github.com:berriai:litellm

ranked child groups

no voted pairs yet in this scope

cli
src
spread
search