mirror of
https://github.com/BerriAI/litellm.git
synced 2026-09-17 23:51:30 +00:00
e0d6c85fb3
Backend resolver fix (logging_exporter_access.py, litellm_pre_call_utils.py):
auto_enable=true previously ignored access.teams/orgs entirely — any request
triggered the destination regardless of its access grants.
New scoped behavior:
auto_enable=true + access.global=true -> fires for all requests
auto_enable=true + access.teams=[A] -> fires only for teamA requests
auto_enable=true + access.orgs=[O] -> fires only for orgO requests
auto_enable=true + empty access -> proxy-wide (backward compat)
Introduced _has_explicit_access_grants() as the single check point in both
the request-time resolver and is_destination_visible(). 18 new unit tests;
all 62 pass. Live verified: 24/24 regression cases with real Anthropic spend.
UI fixes (LoggingCallbacksTable.tsx, types.ts, settings.tsx):
Mode column:
auto_enable=true -> Auto-enabled (orange badge with tooltip)
auto_enable=false -> Manual assignment
Scope column (explicit access grants only, no runtime behavior):
access.global=true -> Global access
access.teams=[...] -> team: ...
access.orgs=[...] -> org: ...
empty/absent -> - (regardless of auto_enable)
Tooltip on Auto-enabled explains proxy-wide backward compat when access
is empty and guides admins to add explicit grants.
Co-authored-by: Mateo Wang <mateo-berri@users.noreply.github.com>
|
||
|---|---|---|
| .. | ||
| litellm-dashboard | ||
| Dockerfile | ||
| nginx.conf | ||