Gemini returns finishReason="STOP" even when tool calls are present,
and sends tool_calls and finishReason in separate streaming chunks.
The ModelResponseIterator now tracks tool_calls across chunks and
correctly maps finish_reason to "tool_calls" per the OpenAI spec.
Fixes#21041
Remove logit_bias, modalities, prediction, audio, web_search_options
from supported params for all GPT-5 reasoning models (OpenAI rejects
them). Add logprobs, top_p, top_logprobs for gpt-5.1/5.2 which support
them when reasoning_effort="none".
Related to #21572
gpt-5-search-api models were routed through OpenAIGPT5Config which
listed params like n, temperature, tools, reasoning_effort as supported,
but OpenAI rejects all of these for search models.
Fixes#21572
- Add warning when credential name is not found in credential_list
- Remove litellm_credential_name from credentials dict after resolution
Addresses Greptile bot review comments on PR #21502
- passthrough/utils.py: change `request_query_params` from `Dict` to
`Mapping` so callers passing `dict[str, str]` satisfy the type checker
(dict is invariant; Mapping is covariant in its value type)
- pass_through_endpoints.py: cast `param_default_query_params` to
`Optional[dict]` consistent with the existing cast pattern for other
params extracted from the untyped `target_params` dict
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add test to verify that get_deployment_credentials_with_provider correctly
resolves litellm_credential_name to actual credential values and removes
the credential name from the returned dictionary.
These constants were added as hardcoded values in #21533 without environment
variable support, causing test_all_numeric_constants_can_be_overridden to fail.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
For multi-turn conversations, convert thinking_blocks on assistant
messages into content blocks prepended before the rest of the content,
so reasoning context is passed back to the hosted_vllm API.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Virtual keys only track created_at and updated_at, which don't indicate
when a key was last used. This adds a last_active field that gets updated
during the async batch spend update, giving admins visibility into which
keys are actively being used.
Changes:
- Add last_active DateTime? to VerificationToken and
DeletedVerificationToken in all 3 schema files and Python types
- Set last_active in the batch key spend update alongside spend increment
- Add Last Active column to virtual keys UI table with info popover
and hover tooltip showing full date/time with timezone
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
