mirror of
https://github.com/BerriAI/litellm.git
synced 2026-09-24 00:52:24 +00:00
1260f09679
resolve_embedding_router gated on exact model-name membership, so wildcard/pattern (bedrock/*) and visible model_group_alias embedding models fell back to a direct litellm.embedding() call and lost deployment-level auth (e.g. Bedrock aws_role_name), the same class of failure as #28244. Resolve via Router.get_model_list(model_name=...), which unifies exact + visible alias + provider-prefixed wildcard/pattern, and drop the now-redundant llm_model_list arg. get_model_list is a listing resolver, not a faithful 'would the Router route this' predicate: hidden model_group_alias entries, a bare model name matched only against a provider-prefixed wildcard, a raw litellm_params.model / model_info.id deployment address, and team-public names without a team_id still resolve empty and keep falling back to a direct embedding call. These are niche embedding-model configs and degrade to current behavior; documented as known gaps. |
||
|---|---|---|
| .. | ||
| test_azure_blob_cache.py | ||
| test_caching.py | ||
| test_caching_handler.py | ||
| test_check_and_fix_namespace_none_guard.py | ||
| test_dual_cache.py | ||
| test_embedding_router.py | ||
| test_gcs_cache.py | ||
| test_in_memory_cache.py | ||
| test_llm_caching_handler.py | ||
| test_llm_client_cache_e2e.py | ||
| test_qdrant_semantic_cache.py | ||
| test_redis_cache.py | ||
| test_redis_cluster_cache.py | ||
| test_redis_connection_pool.py | ||
| test_redis_semantic_cache.py | ||
| test_s3_cache.py | ||
| test_valkey_semantic_cache.py | ||