18 views
-/https://github.com/berriai/litellm/issues/24680
GitHub · issue

#24680 [Bug]: Valid API key not accepted: "Invalid proxy server token passed"

  • State: open
  • Author: @yuri-alias
  • 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?

Since updating to 1.82.3, some of the API keys that were created in liteLLM can no longer access any of the models.

Example with a valid key using curl (Same error appears when using the playground in the web UI)

``` yuri🐝bee:~> curl -s https://<liteLLM_url>/key/info -H "Authorization: Bearer <api_key>" | jq { "error": { "message": "Authentication Error, Invalid proxy server token passed. Received API Key = sk-...m31Q, Key Hash (Token) =<hash>. Unable to find token in cache or `LiteLLM_VerificationTokenTable`", "type": "token_not_found_in_db", "param": "key", "code": "401" } }

```

This key created a long time ago and has worked perfectly well. But since the update, the key can't make any sort of request, neither administrative (/key/info) nor model related (completions). The key is not disabled, or expired, or canceled.

We've been troubleshooting this for a few hours, and we can't determine why this happens.

I've found a few other issues that may be releated:

#13887 - Not applicable, since we have not blocked this …

GitHub resolver

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

Refresh page
vote history (3 events)
#0 of 0 · 31d17h56m3s ago — entered · #import:https:::github.com:berriai:litellm post #3145
The left issue is harder because it likely requires tracing a release regression across authentication, token caching, database lookup, and backward compatibility, while the right is a relatively localized logging and exception-severity adjustment.
The authentication regression is harder because it likely requires tracing proxy token lookup, caching, database compatibility, upgrade behavior, and multiple request paths, with higher regression risk. The media feature is comparatively bounded to provider-specific content normalization, validation, and targeted multimodal tests.
#0 of 0 · 31d17h38m24s ago — current · #import:https:::github.com:berriai:litellm post #3440
29452 requires broader cross-cutting architecture, provider/auth-mode integration, secret lifecycle design, and compatibility work, while 24680 is primarily a focused proxy authentication persistence/debugging fix.
discussed in #import:https:::github.com:berriai:litellm

ranked child groups

no voted pairs yet in this scope

cli
src
spread
search