6 views
-/https://github.com/berriai/litellm/issues/27300
GitHub · issue

#27300 [Bug]: max_budget is ignored after reset

  • State: open
  • Author: @DAGRSAG
  • Labels: bug, proxy

### Check for existing issues

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

### What happened?

After the monthly reset, the spend for each key is set to 0 again. And also in the last month it worked, that if a key reaches the max_budget, new requests are rejected due to the limit

But now after the reset, it seems to be ignored. We have already some key which run over the limits and can still send requests.

If a change the max_budget for a key (spend 193, old budget 200 to now 180) requests immediately fail due to budget limit.

Some logs I can found are posted below

### Steps to Reproduce

1. Define a budget 2. Monthly reset happen 3. Budget is ignored

### Relevant log output

```shell 15:21:57 - LiteLLM Proxy:ERROR: reset_budget_job.py:660 - Failed to reset budget windows for keys: Unable to match input value to any allowed input type for the field. Parse errors: [`where.budget_limits.not`: A value is required but not set, `where.budget_limits.not`: A value is required but not set, `where.budget_limits.not`: A value is required but not set] File "/app/.venv/lib/python3.13/site-packages/litellm/proxy/common_utils/reset_budget_job.…

GitHub resolver

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

Refresh page
vote history (2 events)
#0 of 0 · 31d18h19m39s ago — entered · #import:https:::github.com:berriai:litellm post #2625
The first issue is harder because it involves scheduled state transitions, persistence-layer validation, budget-accounting correctness, and regression coverage across reset and enforcement paths. The second is comparatively localized to request-hook response handling and HTTP behavior, with a narrower implementation and test surface.
#0 of 0 · 31d17h40m37s ago — current · #import:https:::github.com:berriai:litellm post #3302
The managed-settings feature spans new proxy configuration architecture, tool-specific semantics, persistence, authorization, rollout behavior, and broad integration testing. The budget issue is comparatively contained to diagnosing the reset path and correcting validation or update logic, with narrower regression testing.
discussed in #import:https:::github.com:berriai:litellm

ranked child groups

no voted pairs yet in this scope

cli
src
spread
search