test(audio): cover Gemini transcription dispatch

This commit is contained in:
Marcin Juraszek 2026-09-22 20:46:38 +00:00
parent b646791578
commit 2ab4fb33b2
2 changed files with 86 additions and 0 deletions

View file

@ -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.

View file

@ -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"],
}