mirror of
https://github.com/BerriAI/litellm.git
synced 2026-11-15 09:12:18 +00:00
- deprecated registry members called with a None name read the registry source directly again,
as they did before registry() existed, so both cores return the old empty result instead of
registry()'s TypeError (get_model_group, resolve_served_model_name, get_model_from_alias,
get_routable_upstream_model, get_candidate_model_ids_for_route, get_remaining_model_group_usage)
- rust core keeps retry_after, timeout, num_retries, allowed_fails and max_fallbacks as given, as the
python core stores them; update_settings casts to int as python does; the crate reads them as numbers
- rust core adds the {"*": ...} fallback entry whenever default_fallbacks is given, even empty
- rust core passes the configured model_name to deployment ids and deployments as given and logs the
group names after set_model_list, so a non-string model_name and an alias without hidden fail as
they do on the python core
|
||
|---|---|---|
| .. | ||
| core | ||
| __init__.py | ||
| test_enforce_model_rate_limits.py | ||
| test_io_token_rate_limits.py | ||
| test_router.py | ||
| test_router_callback_hook_sequence.py | ||
| test_router_endpoints.py | ||
| test_router_entry_points.py | ||
| test_router_facade.py | ||
| test_router_fallbacks.py | ||
| test_router_get_deployments.py | ||
| test_router_helper_utils.py | ||
| test_router_host.py | ||
| test_router_max_parallel_requests.py | ||
| test_router_provider_endpoints.py | ||
| test_router_retries.py | ||
| test_router_surface_check.py | ||
| test_router_timeout.py | ||
| test_router_vector_store_endpoints.py | ||