From 2ab4fb33b2ac970615cb8844c2b70644e2f6b981 Mon Sep 17 00:00:00 2001 From: Marcin Juraszek <10106352+korasino@users.noreply.github.com> Date: Tue, 22 Sep 2026 20:46:38 +0000 Subject: [PATCH] test(audio): cover Gemini transcription dispatch --- .../test_get_supported_openai_params.py | 29 ++++++++++ .../test_gemini_realtime_transcription.py | 57 +++++++++++++++++++ 2 files changed, 86 insertions(+) create mode 100644 tests/test_litellm/llms/gemini/realtime/test_gemini_realtime_transcription.py diff --git a/tests/test_litellm/litellm_core_utils/test_get_supported_openai_params.py b/tests/test_litellm/litellm_core_utils/test_get_supported_openai_params.py index f9e285cf9fb..d40cf46420c 100644 --- a/tests/test_litellm/litellm_core_utils/test_get_supported_openai_params.py +++ b/tests/test_litellm/litellm_core_utils/test_get_supported_openai_params.py @@ -1,5 +1,6 @@ import pytest +import litellm from litellm.litellm_core_utils.get_supported_openai_params import ( get_supported_openai_params, ) @@ -8,6 +9,34 @@ BEDROCK_REAL_MODEL = "eu.anthropic.claude-haiku-4-5-20251001-v1:0" BEDROCK_LABEL = "claude-haiku-4-5" +def test_gemini_transcription_dispatch_returns_batch_params(monkeypatch): + monkeypatch.setitem( + litellm.model_cost, + "gemini/gemini-3.5-transcribe", + {"audio_transcription_config": "gemini_transcribe"}, + ) + + params = get_supported_openai_params( + model="gemini-3.5-transcribe", + custom_llm_provider="gemini", + request_type="transcription", + ) + + assert params is not None + assert "keywords" in params + assert "timestamp_granularities" in params + + +def test_gemini_live_transcription_dispatch_does_not_use_batch_params(): + params = get_supported_openai_params( + model="gemini-3.5-transcribe-live", + custom_llm_provider="gemini", + request_type="transcription", + ) + + assert params is None + + def test_base_model_label_does_not_strip_bedrock_tools(): """Regression for #29618. diff --git a/tests/test_litellm/llms/gemini/realtime/test_gemini_realtime_transcription.py b/tests/test_litellm/llms/gemini/realtime/test_gemini_realtime_transcription.py new file mode 100644 index 00000000000..e20945072d0 --- /dev/null +++ b/tests/test_litellm/llms/gemini/realtime/test_gemini_realtime_transcription.py @@ -0,0 +1,57 @@ +import json + +import pytest + +from litellm.llms.gemini.realtime.transformation import GeminiRealtimeConfig + + +@pytest.mark.parametrize( + ("transcription", "expected"), + [ + ("invalid", {}), + ({}, {}), + ({"language": "", "keywords": ["", 42]}, {}), + ( + {"language": "pl", "keywords": ["", 42, "term"]}, + {"languageCodes": ["pl-PL"], "customVocabulary": ["term"]}, + ), + ], +) +def test_session_update_maps_transcription_values(transcription, expected): + messages = GeminiRealtimeConfig().transform_realtime_request( + json.dumps( + { + "type": "session.update", + "session": {"input_audio_transcription": transcription}, + } + ), + "gemini-3.5-transcribe-live", + session_configuration_request=None, + ) + + setup = json.loads(messages[0]) + assert setup["setup"]["inputAudioTranscription"] == expected + + +def test_session_update_maps_language_and_keywords(): + messages = GeminiRealtimeConfig().transform_realtime_request( + json.dumps( + { + "type": "session.update", + "session": { + "input_audio_transcription": { + "language": "pl", + "keywords": ["alpha", "beta"], + } + }, + } + ), + "gemini-3.5-transcribe-live", + session_configuration_request=None, + ) + + setup = json.loads(messages[0]) + assert setup["setup"]["inputAudioTranscription"] == { + "languageCodes": ["pl-PL"], + "customVocabulary": ["alpha", "beta"], + }