litellm/tests/unit/proxy/telemetry/__init__.py
devin-ai-integration[bot] bfcb0ff474
feat(telemetry): record proxy requests and provider attempts into the telemetry sink (#45489)
* feat(telemetry): record proxy requests and provider attempts into the telemetry sink

* fix(telemetry): validate env settings, fix the window and only record the request's own provider calls

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

* fix(telemetry): bind the call id from a typed local

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

* fix(telemetry): record a finished request into the sink current when it finishes

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

* fix(telemetry): record every failed provider attempt, including retries and fallbacks on a shared logging object

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

* fix(telemetry): report timeouts and connection failures as their own provider status classes

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

* fix(telemetry): record failed attempts from the per-deployment failure hook instead of a new core logging hook

Reverts the async_log_attempt_failure_event hook added to CustomLogger and
litellm_logging in 3421b38685. The router already calls
async_post_call_failure_deployment_hook once per failed attempt, retries and
fallbacks included, so every failed fallback still gets its own row without
touching the shared logging object.

* fix(telemetry): register the attempt logger only once a telemetry group is turned on

A proxy that never opts in now carries no telemetry callback. Once registered
the logger stays, returning early while every group is off.

---------

Co-authored-by: nate <nate@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-10-13 18:46:06 -07:00

0 lines
Python