test: migrate openai, openai_like and openrouter legacy tests to tests/unit

[devin-ai-integration]

TLDR

Problem this solves:

• 20 legacy test files sat in tests/test_litellm, outside the mutation-proven unit harness

How it solves it:

• Every test judged with a red then green source mutation, all 221 kept
• 20 files git mv'd to the mirrored path under tests/unit, 4 orphan __init__.py removed
• Empty __init__.py added to the 17 tests/unit directories this PR creates, so same-basename files cannot collide at collection

Counts: kept 221 / deleted 0 / rewritten 0 (239 node ids, 18 from parametrize)

moved 239, passed 239 in [REDACTED SECRET]/gh/BerriAI/litellm/2194947 (every node id from this PR is passed in that job's junit)

That job is red from 4 collection errors in files this PR does not touch: base_llm/batches/test_transformation.py, fastcrw/search/test_transformation.py, nvidia_riva/audio_transcription/test_handler.py and test_transformation.py collide by basename with anthropic/batches/ on main (no __init__.py), all landed on main between 07:50 and 08:09 UTC before this branch's first unit run. installing_litellm_on_python* and google_generate_content_endpoint_testing fail with the proxy master key boot check, which also fails on main (job 2194008)

Now uncovered: nothing, no test was deleted

Per-test ledger

llms/openai/embeddings/guardrail_translation/test_embeddings_guardrail_handler.py
test_embeddings_handler_string_input | keep unit | mutated: litellm/llms/openai/embeddings/guardrail_translation/handler.py:OpenAIEmbeddingsHandler._process_string_input: writes the whole guardrailed texts list into data["input"] instead of texts[0] | red then green
test_embeddings_handler_list_of_strings_input | keep unit | mutated: litellm/llms/openai/embeddings/guardrail_translation/handler.py:OpenAIEmbeddingsHandler._process_list_input: writes only guardrailed_texts[:1] into data["input"] | red then green
test_embeddings_guardrail_translation_mappings | keep unit | mutated: litellm/llms/openai/embeddings/guardrail_translation/__init__.py:guardrail_translation_mappings: drops the CallTypes.aembedding entry | red then green

llms/openai/image_generation/test_gpt_transformation.py
test_transform_image_generation_response_keeps_provider_echo | keep unit | mutated: litellm/llms/openai/image_generation/gpt_transformation.py:GPTImageGenerationConfig.transform_image_generation_response: overwrites the echoed quality with optional_params.get("quality", "high") instead of keeping the provider value | red then green

llms/openai/image_generation/test_image_generation_guardrail_handler.py
TestHandlerDiscovery::test_handler_discovered_for_image_generation | keep unit | mutated: litellm/llms/openai/image_generation/guardrail_translation/__init__.py:guardrail_translation_mappings: drops the CallTypes.image_generation entry | red then green
TestHandlerDiscovery::test_handler_discovered_for_aimage_generation | keep unit | mutated: litellm/llms/openai/image_generation/guardrail_translation/__init__.py:guardrail_translation_mappings: drops the CallTypes.aimage_generation entry | red then green
TestInputProcessing::test_process_simple_prompt | keep unit | mutated: litellm/llms/openai/image_generation/guardrail_translation/handler.py:OpenAIImageGenerationHandler.process_input_messages: writes the original prompt back instead of guardrailed_texts[0] | red then green
TestInputProcessing::test_process_no_prompt | keep unit | mutated: litellm/llms/openai/image_generation/guardrail_translation/handler.py:OpenAIImageGenerationHandler.process_input_messages: writes data["prompt"] = "" when no prompt is present | red then green
TestInputProcessing::test_process_empty_prompt | keep unit | mutated: litellm/llms/openai/image_generation/guardrail_translation/handler.py:OpenAIImageGenerationHandler.process_input_messages: writes data["prompt"] = "" when prompt is None | red then green
TestInputProcessing::test_process_complex_prompt | keep unit | mutated: litellm/llms/openai/image_generation/guardrail_translation/handler.py:OpenAIImageGenerationHandler.process_input_messages: writes the original prompt back instead of guardrailed_texts[0] | red then green
TestOutputProcessing::test_process_output_returns_unchanged | keep unit | mutated: litellm/llms/openai/image_generation/guardrail_translation/handler.py:OpenAIImageGenerationHandler.process_output_response: returns a fresh empty ImageResponse instead of the input | red then green
TestPIIMaskingScenario::test_pii_masking_in_prompt | keep unit | mutated: litellm/llms/openai/image_generation/guardrail_translation/handler.py:OpenAIImageGenerationHandler.process_input_messages: writes the original prompt back instead of guardrailed_texts[0] | red then green

llms/openai/image_generation/test_openai_image_generation_extra_headers.py
TestImageGenerationExtraHeaders::test_sync_image_generation_with_headers | keep unit | mutated: litellm/llms/openai/openai.py:OpenAIChatCompletion.image_generation: calls images.generate(**data) so caller headers never reach extra_headers | red then green
TestImageGenerationExtraHeaders::test_sync_image_generation_without_headers | keep unit | mutated: litellm/llms/openai/openai.py:OpenAIChatCompletion.image_generation: always passes extra_headers={} to images.generate | red then green
TestImageGenerationExtraHeaders::test_async_image_generation_with_headers | keep unit | mutated: litellm/llms/openai/openai.py:OpenAIChatCompletion.aimage_generation: calls images.generate(**data) so caller headers never reach extra_headers | red then green
TestImageGenerationExtraHeaders::test_async_image_generation_without_headers | keep unit | mutated: litellm/llms/openai/openai.py:OpenAIChatCompletion.aimage_generation: always passes extra_headers={} to images.generate | red then green
TestImageGenerationExtraHeaders::test_caller_headers_never_reach_the_logged_request_body | keep unit | mutated: litellm/llms/openai/openai.py:OpenAIChatCompletion.image_generation / aimage_generation: calls images.generate(**data) so caller headers never reach extra_headers (both parametrize cases) | red then green
TestImageGenerationExtraHeaders::test_sync_image_generation_forwards_headers_to_async | keep unit | mutated: litellm/llms/openai/openai.py:OpenAIChatCompletion.image_generation: forwards headers=None to aimage_generation when aimg_generation is True | red then green
TestImageGenerationEntryPointHeaders::test_extra_headers_reach_openai_provider | keep unit | mutated: litellm/images/main.py:image_generation: reads extra_headers as None so the caller's headers are dropped before reaching the provider | red then green

llms/openai/speech/test_text_to_speech_guardrail_handler.py
TestHandlerDiscovery::test_handler_discovered_for_speech | keep unit | mutated: litellm/llms/openai/speech/guardrail_translation/__init__.py:guardrail_translation_mappings: drops the CallTypes.speech entry | red then green
TestHandlerDiscovery::test_handler_discovered_for_aspeech | keep unit | mutated: litellm/llms/openai/speech/guardrail_translation/__init__.py:guardrail_translation_mappings: drops the CallTypes.aspeech entry | red then green
TestInputProcessing::test_process_simple_input_text | keep unit | mutated: litellm/llms/openai/speech/guardrail_translation/handler.py:OpenAITextToSpeechHandler.process_input_messages: writes the original input back instead of guardrailed_texts[0] | red then green
TestInputProcessing::test_process_no_input | keep unit | mutated: litellm/llms/openai/speech/guardrail_translation/handler.py:OpenAITextToSpeechHandler.process_input_messages: writes data["input"] = "" when no input is present | red then green
TestInputProcessing::test_process_empty_input | keep unit | mutated: litellm/llms/openai/speech/guardrail_translation/handler.py:OpenAITextToSpeechHandler.process_input_messages: writes data["input"] = "" when input is None | red then green
TestInputProcessing::test_process_long_input | keep unit | mutated: litellm/llms/openai/speech/guardrail_translation/handler.py:OpenAITextToSpeechHandler.process_input_messages: writes the original input back instead of guardrailed_texts[0] | red then green
TestInputProcessing::test_process_with_all_optional_params | keep unit | mutated: litellm/llms/openai/speech/guardrail_translation/handler.py:OpenAITextToSpeechHandler.process_input_messages: writes the original input back instead of guardrailed_texts[0] | red then green
TestOutputProcessing::test_output_processing_is_noop | keep unit | mutated: litellm/llms/openai/speech/guardrail_translation/handler.py:OpenAITextToSpeechHandler.process_output_response: returns None instead of the response | red then green
TestPIIMaskingScenario::test_pii_masking_in_tts_input | keep unit | mutated: litellm/llms/openai/speech/guardrail_translation/handler.py:OpenAITextToSpeechHandler.process_input_messages: writes the original input back instead of guardrailed_texts[0] | red then green
TestPIIMaskingScenario::test_phone_message_pii_redaction | keep unit | mutated: litellm/llms/openai/speech/guardrail_translation/handler.py:OpenAITextToSpeechHandler.process_input_messages: writes the original input back instead of guardrailed_texts[0] | red then green
TestContentModerationScenario::test_profanity_filtering_before_tts | keep unit | mutated: litellm/llms/openai/speech/guardrail_translation/handler.py:OpenAITextToSpeechHandler.process_input_messages: writes the original input back instead of guardrailed_texts[0] | red then green
TestMultilingualTTS::test_multilingual_input_preserved | keep unit | mutated: litellm/llms/openai/speech/guardrail_translation/handler.py:OpenAITextToSpeechHandler.process_input_messages: writes the original input back instead of guardrailed_texts[0] | red then green
TestMultilingualTTS::test_all_voice_options | keep unit | mutated: litellm/llms/openai/speech/guardrail_translation/handler.py:OpenAITextToSpeechHandler.process_input_messages: writes the original input back instead of guardrailed_texts[0] | red then green

llms/openai/evals/test_openai_evals_transformation.py
test_validate_environment_sets_headers | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.validate_environment: sets Authorization to the bare api key without the "Bearer " prefix | red then green
test_validate_environment_requires_api_key | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.validate_environment: returns headers instead of raising ValueError when no api key is found | red then green
test_get_complete_url_with_eval_id | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.get_complete_url: returns f"{api_base}/v1/{endpoint}" even when eval_id is given | red then green
test_get_complete_url_encodes_eval_id | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.get_complete_url: interpolates the raw eval_id instead of encode_url_path_segment(eval_id) | red then green
test_get_complete_url_without_eval_id | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.get_complete_url: returns f"{api_base}/{endpoint}" dropping the /v1 prefix | red then green
test_transform_create_eval_request | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.transform_create_eval_request: drops the testing_criteria key from the request body | red then green
test_transform_create_eval_response | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.transform_create_eval_response: validates {**response_json, "name": ""} so the eval name is blanked | red then green
test_transform_list_evals_request | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.transform_list_evals_request: never copies "after" into query_params | red then green
test_transform_list_evals_response | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.transform_list_evals_response: forces has_more=True on the parsed response | red then green
test_transform_update_eval_request | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.transform_update_eval_request: drops the metadata key from the request body | red then green
test_transform_delete_eval_request | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.transform_delete_eval_request: builds the URL without eval_id (collection URL) | red then green
test_transform_delete_eval_response | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.transform_delete_eval_response: forces deleted=False on the parsed response | red then green
test_transform_cancel_eval_request | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.transform_cancel_eval_request: drops the trailing /cancel from the URL | red then green
test_transform_cancel_eval_response | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.transform_cancel_eval_response: forces id="" on the parsed response | red then green
test_transform_run_requests_encode_eval_and_run_ids | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.transform_cancel_run_request: interpolates the raw run_id instead of encoded_run_id | red then green
test_transform_get_eval_response | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.transform_get_eval_response: validates {**response_json, "name": ""} so the eval name is blanked | red then green
test_transform_update_eval_response | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.transform_update_eval_response: validates {**response_json, "name": ""} so the eval name is blanked | red then green
test_transform_create_run_response | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.transform_create_run_response: forces status="failed" on the parsed run | red then green
test_transform_list_runs_request | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.transform_list_runs_request: hardcodes query_params["order"] = "desc" | red then green
test_transform_list_runs_response | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.transform_list_runs_response: forces has_more=True on the parsed response | red then green
test_transform_get_run_response | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.transform_get_run_response: forces status="queued" on the parsed run | red then green
test_transform_cancel_run_response | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.transform_cancel_run_response: forces status="queued" on the parsed run | red then green
test_transform_delete_run_request | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.transform_delete_run_request: returns request_body {"run_id": run_id} instead of {} | red then green
test_transform_delete_run_response | keep unit | mutated: litellm/llms/openai/evals/transformation.py:OpenAIEvalsConfig.transform_delete_run_response: forces deleted=False on the parsed response | red then green

llms/openai/transcriptions/test_transcription_duration_hidden.py
TestDiarizedJsonUsageParsing::test_fractional_duration_seconds_does_not_raise | keep unit | mutated: litellm/litellm_core_utils/llm_response_utils/convert_dict_to_response.py:convert_to_model_response_object: builds TranscriptionUsageDurationObject with seconds=int(usage["seconds"]) truncating 295.8 to 295 | red then green
TestDiarizedJsonUsageParsing::test_usage_duration_object_accepts_float_seconds | keep unit | mutated: litellm/types/utils.py:TranscriptionUsageDurationObject: seconds field typed int instead of float | red then green
TestTranscriptionDurationNotInResponseBody::test_convert_dict_stores_internal_duration_in_hidden_params | keep unit | mutated: litellm/litellm_core_utils/llm_response_utils/convert_dict_to_response.py:convert_to_model_response_object: sets _audio_transcription_duration as a response attribute instead of _hidden_params["audio_transcription_duration"] | red then green
TestTranscriptionDurationNotInResponseBody::test_convert_dict_preserves_provider_duration | keep unit | mutated: litellm/litellm_core_utils/llm_response_utils/convert_dict_to_response.py:convert_to_model_response_object: removes "duration" from the copied optional_keys | red then green
TestTranscriptionDurationNotInResponseBody::test_plain_json_response_has_no_duration | keep unit | mutated: litellm/litellm_core_utils/llm_response_utils/convert_dict_to_response.py:convert_to_model_response_object: sets every optional key with response_object.get(key, 0.0) so duration becomes 0.0 | red then green
TestCostCalculatorReadsDurationFromHiddenParams::test_completion_cost_uses_hidden_params_duration | keep unit | mutated: litellm/cost_calculator.py:completion_cost: reads duration only from getattr(completion_response, "duration", 0.0), ignoring _hidden_params | red then green
TestCostCalculatorReadsDurationFromHiddenParams::test_completion_cost_falls_back_to_response_duration | keep unit | mutated: litellm/cost_calculator.py:completion_cost: falls back to 0.0 instead of completion_response.duration | red then green
TestCostCalculatorReadsDurationFromHiddenParams::test_completion_cost_defaults_to_zero_duration | keep unit | mutated: litellm/cost_calculator.py:completion_cost: default duration 1.0 instead of 0.0 when neither source has one | red then green

llms/openai/transcriptions/test_whisper_transformation.py
TestWhisperTransformRequestResponseFormat::test_defaults_to_verbose_json_when_unset | keep unit | mutated: litellm/llms/openai/transcriptions/whisper_transformation.py:OpenAIWhisperAudioTranscriptionConfig.transform_audio_transcription_request: defaults response_format to "json" instead of "verbose_json" | red then green
TestWhisperTransformRequestResponseFormat::test_respects_explicit_json | keep unit | mutated: litellm/llms/openai/transcriptions/whisper_transformation.py:OpenAIWhisperAudioTranscriptionConfig.transform_audio_transcription_request: unconditionally overwrites response_format with "verbose_json" | red then green
TestWhisperTransformRequestResponseFormat::test_respects_explicit_text | keep unit | mutated: litellm/llms/openai/transcriptions/whisper_transformation.py:OpenAIWhisperAudioTranscriptionConfig.transform_audio_transcription_request: unconditionally overwrites response_format with "verbose_json" | red then green
TestWhisperTransformRequestResponseFormat::test_preserves_verbose_json_when_set | keep unit | mutated: litellm/llms/openai/transcriptions/whisper_transformation.py:OpenAIWhisperAudioTranscriptionConfig.transform_audio_transcription_request: rewrites an explicit "verbose_json" to "json" | red then green
TestWhisperTransformResponse::test_parses_json_response | keep unit | mutated: litellm/llms/openai/transcriptions/whisper_transformation.py:OpenAIWhisperAudioTranscriptionConfig.transform_audio_transcription_response: returns TranscriptionResponse(text="") for JSON bodies | red then green
TestWhisperTransformResponse::test_parses_plain_text_response | keep unit | mutated: litellm/llms/openai/transcriptions/whisper_transformation.py:OpenAIWhisperAudioTranscriptionConfig.transform_audio_transcription_response: returns TranscriptionResponse(text="") for plain-text bodies | red then green
TestWhisperTransformResponse::test_malformed_json_body_with_json_content_type_raises | keep unit | mutated: litellm/llms/openai/transcriptions/whisper_transformation.py:OpenAIWhisperAudioTranscriptionConfig.transform_audio_transcription_response: swallows JSONDecodeError for application/json bodies and returns the text as a transcription | red then green
TestWhisperTransformResponse::test_json_content_type_match_is_case_insensitive | keep unit | mutated: litellm/llms/openai/transcriptions/whisper_transformation.py:OpenAIWhisperAudioTranscriptionConfig.transform_audio_transcription_response: compares content-type without .lower() | red then green

llms/openai/transcriptions/test_audio_transcription_guardrail_handler.py
TestHandlerDiscovery::test_handler_discovered_for_transcription | keep unit | mutated: litellm/llms/openai/transcriptions/guardrail_translation/__init__.py:guardrail_translation_mappings: dropped the CallTypes.transcription entry so lookup returns nothing | red then green
TestHandlerDiscovery::test_handler_discovered_for_atranscription | keep unit | mutated: litellm/llms/openai/transcriptions/guardrail_translation/__init__.py:guardrail_translation_mappings: dropped the CallTypes.atranscription entry so lookup returns nothing | red then green
TestInputProcessing::test_input_processing_is_noop | keep unit | mutated: litellm/llms/openai/transcriptions/guardrail_translation/handler.py:OpenAIAudioTranscriptionHandler.process_input_messages: returns data with the "file" key stripped instead of the untouched dict | red then green
TestOutputProcessing::test_process_simple_transcription | keep unit | mutated: litellm/llms/openai/transcriptions/guardrail_translation/handler.py:OpenAIAudioTranscriptionHandler.process_output_response: writes original_text back instead of guardrailed_texts[0] | red then green
TestOutputProcessing::test_process_transcription_with_metadata | keep unit | mutated: litellm/llms/openai/transcriptions/guardrail_translation/handler.py:OpenAIAudioTranscriptionHandler.process_output_response: sets response.duration = None and response.language = None after applying the guardrailed text | red then green
TestOutputProcessing::test_process_empty_text | keep unit | mutated: litellm/llms/openai/transcriptions/guardrail_translation/handler.py:OpenAIAudioTranscriptionHandler.process_output_response: sets response.text = "" instead of returning the response untouched when text is None | red then green
TestOutputProcessing::test_process_long_transcription | keep unit | mutated: litellm/llms/openai/transcriptions/guardrail_translation/handler.py:OpenAIAudioTranscriptionHandler.process_output_response: writes original_text back instead of guardrailed_texts[0] | red then green
TestPIIMaskingScenario::test_pii_masking_in_transcription | keep unit | mutated: litellm/llms/openai/transcriptions/guardrail_translation/handler.py:OpenAIAudioTranscriptionHandler.process_output_response: writes original_text back instead of guardrailed_texts[0] | red then green
TestPIIMaskingScenario::test_meeting_transcription_pii_redaction | keep unit | mutated: litellm/llms/openai/transcriptions/guardrail_translation/handler.py:OpenAIAudioTranscriptionHandler.process_output_response: writes original_text back instead of guardrailed_texts[0] | red then green
TestContentModerationScenario::test_profanity_filtering | keep unit | mutated: litellm/llms/openai/transcriptions/guardrail_translation/handler.py:OpenAIAudioTranscriptionHandler.process_output_response: writes original_text back instead of guardrailed_texts[0] | red then green
TestMultilingualTranscription::test_transcription_preserves_language_info | keep unit | mutated: litellm/llms/openai/transcriptions/guardrail_translation/handler.py:OpenAIAudioTranscriptionHandler.process_output_response: sets response.language = None after applying the guardrailed text | red then green

llms/openai/vector_store_files/test_openai_vector_store_files_transformation.py
test_validate_environment_sets_headers | keep unit | mutated: litellm/llms/openai/vector_store_files/transformation.py:OpenAIVectorStoreFilesConfig.validate_environment: no longer adds the OpenAI-Beta assistants header | red then green
test_get_complete_url | keep unit | mutated: litellm/llms/openai/vector_store_files/transformation.py:OpenAIVectorStoreFilesConfig.get_complete_url: returns .../vector_stores/{id} without the trailing /files | red then green
test_get_complete_url_encodes_vector_store_id | keep unit | mutated: litellm/llms/openai/vector_store_files/transformation.py:OpenAIVectorStoreFilesConfig.get_complete_url: interpolates the raw vector_store_id instead of encode_url_path_segment(...) | red then green
test_transform_create_request | keep unit | mutated: litellm/llms/openai/vector_store_files/transformation.py:OpenAIVectorStoreFilesConfig.transform_create_vector_store_file_request: filtered_attributes = None so "attributes" is popped from the payload | red then green
test_transform_list_request | keep unit | mutated: litellm/llms/openai/vector_store_files/transformation.py:OpenAIVectorStoreFilesConfig.transform_list_vector_store_files_request: drops the "order" key from the returned query params | red then green
test_transform_file_request_encodes_file_id | keep unit | mutated: litellm/llms/openai/vector_store_files/transformation.py:OpenAIVectorStoreFilesConfig.transform_retrieve_vector_store_file_content_request: interpolates the raw file_id instead of encode_url_path_segment(...) | red then green
test_transform_create_response | keep unit | mutated: litellm/llms/openai/vector_store_files/transformation.py:OpenAIVectorStoreFilesConfig.transform_create_vector_store_file_response: returns {**response.json(), "status": "in_progress"} | red then green
test_transform_list_response | keep unit | mutated: litellm/llms/openai/vector_store_files/transformation.py:OpenAIVectorStoreFilesConfig.transform_list_vector_store_files_response: returns {**response.json(), "has_more": True} | red then green
test_transform_delete_response | keep unit | mutated: litellm/llms/openai/vector_store_files/transformation.py:OpenAIVectorStoreFilesConfig.transform_delete_vector_store_file_response: returns {**response.json(), "deleted": False} | red then green

llms/openai/vector_stores/test_openai_vector_stores_transformation.py
TestOpenAIVectorStoreAPIConfig::test_transform_create_vector_store_request_with_metadata_empty_or_none | keep unit | mutated: litellm/llms/openai/vector_stores/transformation.py:OpenAIVectorStoreConfig.transform_create_vector_store_request: metadata_payload = add_openai_metadata(metadata) or {"litellm": "true"} | red then green (both params)
TestOpenAIVectorStoreAPIConfig::test_transform_create_vector_store_request_with_large_metadata | keep unit | mutated: litellm/utils.py:add_openai_metadata: trims to 15 keys (idx < 15) instead of 16 | red then green
TestOpenAIVectorStoreAPIConfig::test_transform_search_vector_store_request_encodes_vector_store_id | keep unit | mutated: litellm/llms/openai/vector_stores/transformation.py:OpenAIVectorStoreConfig.transform_search_vector_store_request: interpolates the raw vector_store_id instead of encode_url_path_segment(...) | red then green
TestOpenAIVectorStoreAPIConfig::test_transform_search_vector_store_request_preserves_query_string | keep unit | mutated: litellm/llms/openai/vector_stores/transformation.py:OpenAIVectorStoreConfig.transform_search_vector_store_request: builds f"{base_url}/{id}/search" dropping the api_base query string | red then green

llms/openai/videos/test_openai_video_transformation.py
test_video_content_request_encodes_video_id_path_segment | keep unit | mutated: litellm/llms/openai/videos/transformation.py:OpenAIVideoConfig.transform_video_content_request: interpolates original_video_id unencoded into the /content URL | red then green
test_video_content_request_encodes_variant_query_param | keep unit | mutated: litellm/llms/openai/videos/transformation.py:OpenAIVideoConfig.transform_video_content_request: appends ?variant={variant} without quote(...) | red then green
test_wrapped_character_id_is_decoded_then_encoded_as_path_segment | keep unit | mutated: litellm/llms/openai/videos/transformation.py:OpenAIVideoConfig.transform_video_get_character_request: interpolates original_character_id unencoded into /characters/{id} | red then green
test_video_edit_request_forwards_uploaded_file_as_multipart | keep unit | mutated: litellm/llms/openai/videos/transformation.py:OpenAIVideoConfig.transform_video_edit_request: returns (url, form_data, None) dropping the multipart files tuple | red then green
test_video_edit_request_without_file_sends_json_id_reference | keep unit | mutated: litellm/llms/openai/videos/transformation.py:OpenAIVideoConfig.transform_video_edit_request: sends {"video_id": id} instead of {"video": {"id": id}} | red then green

llms/openai_like/chat/test_openai_like_chat_transformation.py
test_sanitize_usage_obj_handles_null_tokens | keep unit | mutated: litellm/llms/openai_like/chat/transformation.py:OpenAILikeChatConfig._sanitize_usage_obj: writes `usage[key] = 1` instead of `usage[key] = 0` for None *_tokens values | red then green
test_sanitize_usage_obj_no_usage | keep unit | mutated: litellm/llms/openai_like/chat/transformation.py:OpenAILikeChatConfig._sanitize_usage_obj: replaces the `if "usage" in response_json and isinstance(...)` guard with `if True:` so the sanitizer touches responses with no usage key (KeyError) | red then green
test_sanitize_usage_obj_valid_usage | keep unit | mutated: litellm/llms/openai_like/chat/transformation.py:OpenAILikeChatConfig._sanitize_usage_obj: adds `response_json["usage"] = {}` before return so the usage object is replaced with an empty dict (note: the test's shallow copy hides in-place mutations; rebinding the top-level key is what it detects) | red then green

llms/openai_like/embedding/test_openai_like_embedding.py
TestOpenAILikeEmbeddingHandler::test_encoding_format_none_filtered_out | keep unit | mutated: litellm/llms/openai_like/embedding/handler.py:OpenAILikeEmbeddingHandler.embedding: replaces the `v not in (None, "")` filter with `dict(optional_params)` so None values are sent in the request body | red then green
TestOpenAILikeEmbeddingHandler::test_encoding_format_empty_string_filtered_out | keep unit | mutated: litellm/llms/openai_like/embedding/handler.py:OpenAILikeEmbeddingHandler.embedding: same `dict(optional_params)` no-filter mutation sends "" through | red then green
TestOpenAILikeEmbeddingHandler::test_encoding_format_float_preserved | keep unit | mutated: litellm/llms/openai_like/embedding/handler.py:OpenAILikeEmbeddingHandler.embedding: appends `"encoding_format": "mutated"` into data after the filtered params spread, overwriting the caller value | red then green
TestOpenAILikeEmbeddingHandler::test_encoding_format_base64_preserved | keep unit | mutated: litellm/llms/openai_like/embedding/handler.py:OpenAILikeEmbeddingHandler.embedding: same overwrite mutation | red then green
TestOpenAILikeEmbeddingHandler::test_other_optional_params_preserved | keep unit | mutated: litellm/llms/openai_like/embedding/handler.py:OpenAILikeEmbeddingHandler.embedding: `dict(optional_params)` no-filter mutation leaks encoding_format=None into the body | red then green
TestOpenAILikeEmbeddingHandler::test_no_optional_params | keep unit | mutated: litellm/llms/openai_like/embedding/handler.py:OpenAILikeEmbeddingHandler.embedding: injected default `"encoding_format": "mutated"` appears in the body | red then green

llms/openai_like/messages/test_openai_like_anthropic_messages_transformation.py
test_get_complete_url_handles_api_base_variants | keep unit | mutated: litellm/llms/openai_like/messages/transformation.py:OpenAILikeAnthropicMessagesConfig.get_complete_url: drops `base.removesuffix("/v1")` so api_base ending in /v1 produces /v1/v1/messages | red then green
test_get_complete_url_requires_api_base | keep unit | mutated: litellm/llms/openai_like/messages/transformation.py:OpenAILikeAnthropicMessagesConfig.get_complete_url: returns "" instead of raising ValueError when api_base is missing | red then green
test_request_stays_in_anthropic_shape | keep unit | mutated: litellm/llms/openai_like/messages/transformation.py:OpenAILikeAnthropicMessagesConfig.transform_anthropic_messages_request: passes `model=f"{model}-mutated"` to the super() transform so payload["model"] changes | red then green
test_request_requires_max_tokens | keep unit | mutated: litellm/llms/anthropic/experimental_pass_through/messages/transformation.py:AnthropicMessagesConfig.transform_anthropic_messages_request: defaults `max_tokens = 1024` instead of raising AnthropicError when max_tokens is missing | red then green
test_validate_environment_sets_bearer_and_anthropic_defaults | keep unit | mutated: litellm/llms/openai_like/messages/transformation.py: DEFAULT_ANTHROPIC_API_VERSION changed "2023-06-01" -> "1999-01-01" | red then green
test_validate_environment_does_not_overwrite_caller_headers | keep unit | mutated: litellm/llms/openai_like/messages/transformation.py:OpenAILikeAnthropicMessagesConfig.validate_anthropic_messages_environment: `needs_auth` becomes `bool(api_key)` so a caller authorization header is overwritten | red then green
test_validate_environment_preserves_standard_cased_caller_headers | keep unit | mutated: litellm/llms/openai_like/messages/transformation.py:OpenAILikeAnthropicMessagesConfig.validate_anthropic_messages_environment: `present` computed as `set(headers)` instead of lowercased keys, so "Authorization" is not detected and a duplicate lowercase header is added | red then green
test_validate_environment_honors_x_api_key_when_present | keep unit | mutated: litellm/llms/openai_like/messages/transformation.py:OpenAILikeAnthropicMessagesConfig.validate_anthropic_messages_environment: `needs_auth = bool(api_key)` ignores x-api-key and still injects authorization | red then green
test_validate_environment_injects_anthropic_beta_for_context_management | keep unit | mutated: litellm/llms/anthropic/experimental_pass_through/messages/transformation.py:AnthropicMessagesConfig._update_headers_with_anthropic_beta: `if has_other:` -> `if False:` so the context-management-2025-06-27 beta is never added | red then green
test_validate_environment_injects_anthropic_beta_for_fast_mode | keep unit | mutated: litellm/llms/anthropic/experimental_pass_through/messages/transformation.py:AnthropicMessagesConfig._update_headers_with_anthropic_beta: speed check changed `== "fast"` to `== "slow"` so fast-mode-2026-02-01 is not added | red then green
test_validate_environment_merges_existing_anthropic_beta | keep unit | mutated: litellm/llms/anthropic/experimental_pass_through/messages/transformation.py:AnthropicMessagesConfig._update_headers_with_anthropic_beta: `beta_values.update(existing_beta)` -> `beta_values.update(set())` so caller anthropic-beta values are dropped | red then green
test_request_strips_advisor_blocks_when_advisor_tool_absent | keep unit | mutated: litellm/llms/anthropic/experimental_pass_through/messages/transformation.py:AnthropicMessagesConfig.transform_anthropic_messages_request: `if not _has_advisor:` -> `if False:` so advisor blocks are never stripped | red then green
test_request_maps_reasoning_effort_to_thinking | keep unit | mutated: litellm/llms/anthropic/experimental_pass_through/messages/transformation.py:AnthropicMessagesConfig._translate_reasoning_effort_to_anthropic: `optional_params.pop("reasoning_effort", None)` -> `.get(...)` so reasoning_effort stays in the payload | red then green
test_passthrough_disables_anthropic_beta_filtering | keep unit | mutated: litellm/llms/openai_like/messages/transformation.py:OpenAILikeAnthropicMessagesConfig.should_filter_anthropic_beta_headers: returns True instead of False | red then green
test_anthropic_beta_survives_provider_filter_on_passthrough_path | keep unit | mutated: litellm/llms/openai_like/messages/transformation.py:OpenAILikeAnthropicMessagesConfig.should_filter_anthropic_beta_headers: returns True instead of False (the filtered beta values no longer survive) | red then green
test_json_provider_messages_config_probes_capabilities_under_provider_slug | keep unit | mutated: litellm/llms/openai_like/messages/transformation.py:JSONProviderAnthropicMessagesConfig.custom_llm_provider: returns "anthropic" instead of the provider slug | red then green
test_request_strips_cache_control_ttl_everywhere | keep unit | mutated: litellm/llms/openai_like/messages/transformation.py:OpenAILikeAnthropicMessagesConfig.supports_cache_control_ttl: returns True so normalize_cache_control_in_anthropic_payload is skipped and ttl survives | red then green
test_request_defaults_missing_cache_control_type_and_drops_non_dict | keep unit | mutated: litellm/llms/openai_like/messages/transformation.py:OpenAILikeAnthropicMessagesConfig.supports_cache_control_ttl: returns True so normalization (type defaulting / non-dict drop) is skipped | red then green
test_native_anthropic_config_keeps_cache_control_ttl | keep unit | mutated: litellm/llms/anthropic/experimental_pass_through/messages/transformation.py:AnthropicMessagesConfig.transform_anthropic_messages_request: applies normalize_cache_control_in_anthropic_payload on the native path so ttl is stripped | red then green
test_deployment_opt_in_keeps_cache_control_ttl | keep unit | mutated: litellm/llms/openai_like/messages/transformation.py:OpenAILikeAnthropicMessagesConfig.__init__: `self._cache_control_ttl = cache_control_ttl` -> `= False` so the opt-in is ignored | red then green
test_json_provider_constraint_opts_into_cache_control_ttl | keep unit | mutated: litellm/llms/openai_like/messages/transformation.py:JSONProviderAnthropicMessagesConfig.__init__: `cache_control_ttl=bool(provider.constraints.get("cache_control_ttl"))` -> `cache_control_ttl=False` | red then green
test_request_strips_ttl_only_where_the_messages_api_defines_cache_control | keep unit | mutated: litellm/llms/anthropic/common_utils.py:_with_portable_cache_control: blocks without their own cache_control now recurse into nested Mapping values so tool_use.input/input_schema cache_control data is rewritten | red then green

llms/openrouter/test_openrouter_embedding_transformation.py
test_openrouter_embedding_supported_params | keep unit | mutated: litellm/llms/openrouter/embedding/transformation.py:OpenrouterEmbeddingConfig.get_supported_openai_params: drops "dimensions" from the returned list | red then green
test_openrouter_embedding_transform_request | keep unit | mutated: litellm/llms/openrouter/embedding/transformation.py:OpenrouterEmbeddingConfig.transform_embedding_request: keeps `model = model` instead of stripping the "openrouter/" prefix | red then green
test_openrouter_embedding_validate_environment | keep unit | mutated: litellm/llms/openrouter/embedding/transformation.py:OpenrouterEmbeddingConfig.validate_environment: Content-Type header "application/json" -> "text/plain" | red then green
test_openrouter_embedding_get_complete_url | keep unit | mutated: litellm/llms/openrouter/embedding/transformation.py:OpenrouterEmbeddingConfig.get_complete_url: returns f"{api_base}/embeddingz" instead of f"{api_base}/embeddings" | red then green
test_openrouter_embedding_map_params | keep unit | mutated: litellm/llms/openrouter/embedding/transformation.py:OpenrouterEmbeddingConfig.map_openai_params: `param in supported` -> `param not in supported` so supported params are dropped and unsupported kept | red then green

llms/openrouter/test_openrouter_provider_routing.py
TestOpenRouterNativeModelRouting::test_double_prefixed_strips_once | keep unit | mutated: litellm/litellm_core_utils/get_llm_provider_logic.py:_get_openai_compatible_provider_info: `model.split("/", 1)[1]` -> `model.split("/")[-1]` so double-prefixed native ids are stripped twice | red then green
TestOpenRouterNativeModelRouting::test_bridge_double_call_preserves_native_model | keep unit | mutated: litellm/litellm_core_utils/get_llm_provider_logic.py:get_llm_provider openrouter block: swaps the two returns so a remainder without "/" returns `remainder` (strips "openrouter/" from the native id) | red then green
TestOpenRouterNativeModelRouting::test_regular_models_still_strip_normally | keep unit | mutated: litellm/litellm_core_utils/get_llm_provider_logic.py:_get_openai_compatible_provider_info: `model.split("/", 1)[1]` -> `model.split("/")[-1]` drops the vendor namespace | red then green
TestOpenRouterNativeModelRouting::test_wildcard_deployment_strips_routing_prefix | keep unit | mutated: litellm/litellm_core_utils/get_llm_provider_logic.py:get_llm_provider openrouter block: returns `model` instead of `remainder` so the routing prefix is kept | red then green

llms/openrouter/chat/test_openrouter_chat_transformation.py
TestOpenRouterChatCompletionStreamingHandler::test_chunk_parser_successful | keep unit | mutated: litellm/llms/openrouter/chat/transformation.py:OpenRouterChatCompletionStreamingHandler.chunk_parser: sets `choice["delta"]["reasoning"] = "mutated"` so reasoning_content no longer matches the chunk's reasoning (setting reasoning_content=None stays green because the Delta model re-derives it from "reasoning") | red then green
TestOpenRouterChatCompletionStreamingHandler::test_chunk_parser_error_response | keep unit | mutated: litellm/llms/openrouter/chat/transformation.py:OpenRouterChatCompletionStreamingHandler.chunk_parser: `status_code=error_message["code"]` -> `status_code=500` | red then green
TestOpenRouterChatCompletionStreamingHandler::test_chunk_parser_key_error | keep unit | mutated: litellm/llms/openrouter/chat/transformation.py:OpenRouterChatCompletionStreamingHandler.chunk_parser: `except KeyError` -> `except TypeError` so KeyError propagates raw instead of OpenRouterException | red then green
test_openrouter_extra_body_transformation | keep unit | mutated: litellm/llms/openrouter/chat/transformation.py:OpenrouterConfig.transform_request: `response.update(extra_body)` -> `response.update({})` so provider routing keys are dropped | red then green
test_openrouter_cache_control_flag_removal | keep unit | mutated: litellm/llms/openrouter/chat/transformation.py:OpenrouterConfig.remove_cache_control_flag_from_messages_and_tools: non-supporting models return `messages, tools` unchanged instead of calling super() (the always-True _supports_cache_control_in_content mutation stays green because the flag is then moved into content) | red then green
test_openrouter_transform_request_with_cache_control | keep unit | mutated: litellm/llms/openrouter/chat/transformation.py:OpenrouterConfig._move_cache_control_to_content: string-content branch writes `"cache_control": None` instead of the flag | red then green
test_openrouter_transform_request_with_cache_control_list_content | keep unit | mutated: litellm/llms/openrouter/chat/transformation.py:OpenrouterConfig._move_cache_control_to_content: `if i == len(content) - 1` -> `if i == 0` so cache_control lands on the first block | red then green
test_openrouter_transform_request_with_cache_control_gemini | keep unit | mutated: litellm/llms/openrouter/chat/transformation.py:OpenrouterConfig._move_cache_control_to_content: string-content branch writes `"cache_control": None` | red then green
test_openrouter_transform_request_multiple_cache_controls | keep unit | mutated: litellm/llms/openrouter/chat/transformation.py:OpenrouterConfig._move_cache_control_to_content: first-block mutation (`i == 0`) | red then green
test_openrouter_cost_tracking_non_streaming | keep unit | mutated: litellm/llms/openrouter/chat/transformation.py:OpenrouterConfig.transform_request: `response["usage"] = {"include": True}` -> `{"include": False}` | red then green
test_openrouter_cost_tracking_streaming | keep unit | mutated: litellm/llms/openrouter/chat/transformation.py:OpenrouterConfig.transform_request: same usage include False mutation | red then green
test_openrouter_reasoning_models_allow_reasoning_effort_param | keep unit | mutated: litellm/llms/openrouter/chat/transformation.py:OpenrouterConfig.get_supported_openai_params: supports_reasoning check -> `if False:` so reasoning_effort is never appended | red then green
test_openrouter_non_reasoning_models_do_not_add_reasoning_effort | keep unit | mutated: litellm/llms/openrouter/chat/transformation.py:OpenrouterConfig.get_supported_openai_params: supports_reasoning check -> `if True:` so reasoning_effort is always appended | red then green
test_openrouter_reasoning_effort_max_maps_to_xhigh | keep unit | mutated: litellm/llms/openrouter/chat/transformation.py:OpenrouterConfig.map_openai_params: maps "max" to "high" instead of "xhigh" | red then green
test_openrouter_reasoning_effort_max_does_not_mutate_caller_dict | keep unit | mutated: litellm/llms/openrouter/chat/transformation.py:OpenrouterConfig.map_openai_params: `non_default_params["reasoning_effort"] = "xhigh"` in place instead of building a copied dict | red then green
test_openrouter_reasoning_effort_xhigh_passes_through | keep unit | mutated: litellm/llms/openrouter/chat/transformation.py:OpenrouterConfig.map_openai_params: adds `mapped_openai_params.pop("reasoning_effort", None)` so the param is dropped from the result | red then green
test_openrouter_reasoning_effort_high_passes_through | keep unit | mutated: litellm/llms/openrouter/chat/transformation.py:OpenrouterConfig.map_openai_params: same pop mutation | red then green

llms/openrouter/image_edit/test_openrouter_image_edit_transformation.py
TestOpenRouterImageEditTransformation::test_get_supported_openai_params | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:get_supported_openai_params: returns ["size", "quality"] dropping "n" | red then green
TestOpenRouterImageEditTransformation::test_use_multipart_form_data_returns_false | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:use_multipart_form_data: returns True instead of False | red then green
TestOpenRouterImageEditTransformation::test_map_openai_params_size | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:map_openai_params: aspect_ratio hardcoded to "9:16" instead of _map_size_to_aspect_ratio(value) | red then green
TestOpenRouterImageEditTransformation::test_map_openai_params_quality | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:_map_quality_to_image_size: "high" maps to "2K" instead of "4K" | red then green
TestOpenRouterImageEditTransformation::test_map_openai_params_size_and_quality | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:map_openai_params: aspect_ratio hardcoded to "9:16" | red then green
TestOpenRouterImageEditTransformation::test_map_openai_params_n_passthrough | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:map_openai_params: else-branch `mapped_params[key] = value` -> `pass` so n is dropped | red then green
TestOpenRouterImageEditTransformation::test_map_openai_params_unknown_quality_ignored | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:_map_quality_to_image_size: `.get(quality)` -> `.get(quality, "1K")` so unknown quality produces an image_config | red then green
TestOpenRouterImageEditTransformation::test_map_size_to_aspect_ratio_square | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:_map_size_to_aspect_ratio: "512x512" -> "2:3" | red then green
TestOpenRouterImageEditTransformation::test_map_size_to_aspect_ratio_landscape | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:_map_size_to_aspect_ratio: "1792x1024" -> "4:3" | red then green
TestOpenRouterImageEditTransformation::test_map_size_to_aspect_ratio_portrait | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:_map_size_to_aspect_ratio: "1024x1536" -> "3:2" | red then green
TestOpenRouterImageEditTransformation::test_map_size_to_aspect_ratio_unknown_defaults_to_1_1 | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:_map_size_to_aspect_ratio: `.get(size, "1:1")` -> `.get(size, "9:16")` | red then green
TestOpenRouterImageEditTransformation::test_map_quality_to_image_size | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:_map_quality_to_image_size: "high" -> "2K" | red then green
TestOpenRouterImageEditTransformation::test_map_quality_to_image_size_unknown_returns_none | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:_map_quality_to_image_size: `.get(quality, "1K")` default | red then green
TestOpenRouterImageEditTransformation::test_get_complete_url_default | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:get_complete_url: appends "/images/generations" instead of "/chat/completions" | red then green
TestOpenRouterImageEditTransformation::test_get_complete_url_with_custom_base | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:get_complete_url: same suffix mutation | red then green
TestOpenRouterImageEditTransformation::test_get_complete_url_with_complete_base | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:get_complete_url: `endswith("/chat/completions")` -> `endswith("/v1")` so the suffix is appended again | red then green
TestOpenRouterImageEditTransformation::test_validate_environment_with_api_key | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:validate_environment: `f"Bearer {api_key}"` -> `f"Token {api_key}"` | red then green
TestOpenRouterImageEditTransformation::test_validate_environment_with_secret_key | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:validate_environment: same Token prefix mutation | red then green
TestOpenRouterImageEditTransformation::test_validate_environment_missing_api_key_raises | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:validate_environment: `api_key = "fallback"` instead of raising ValueError | red then green
TestOpenRouterImageEditTransformation::test_transform_image_edit_request_basic | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:transform_image_edit_request: `"modalities": ["image", "text"]` -> `["text"]` | red then green
TestOpenRouterImageEditTransformation::test_transform_image_edit_request_with_bytesio | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:transform_image_edit_request: data URL `data:{mime};base64,{b64}` -> `data:{mime},{b64}` | red then green
TestOpenRouterImageEditTransformation::test_transform_image_edit_request_with_multiple_images | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:transform_image_edit_request: `image if isinstance(image, list) else [image]` -> `[image]` so a list is treated as one image and _read_image_bytes raises | red then green
TestOpenRouterImageEditTransformation::test_transform_image_edit_request_with_optional_params | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:transform_image_edit_request: `if key not in ("model","messages","modalities")` -> `if key in ()` so optional params are never copied | red then green
TestOpenRouterImageEditTransformation::test_transform_image_edit_request_base64_encoding | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:transform_image_edit_request: `base64.b64encode(image_bytes).decode("utf-8")` reversed (`[::-1]`) | red then green
TestOpenRouterImageEditTransformation::test_transform_image_edit_request_no_prompt | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:transform_image_edit_request: `if prompt:` -> `if True:` so a text part with None is added | red then green
TestOpenRouterImageEditTransformation::test_transform_image_edit_response_with_base64 | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:transform_image_edit_response: `image_url.startswith("data:")` -> `startswith("https:")` so data URLs go down the url path | red then green
TestOpenRouterImageEditTransformation::test_transform_image_edit_response_with_url | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:transform_image_edit_response: `url=image_url` -> `url=None` in the non-data branch | red then green
TestOpenRouterImageEditTransformation::test_transform_image_edit_response_usage_and_cost | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:_set_usage_and_cost: `output_tokens=image_tokens` -> `output_tokens=prompt_tokens` | red then green
TestOpenRouterImageEditTransformation::test_transform_image_edit_response_multiple_images | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:transform_image_edit_response: `startswith("https:")` flip | red then green
TestOpenRouterImageEditTransformation::test_transform_image_edit_response_json_error | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:transform_image_edit_response: `except Exception` -> `except KeyError` so JSONDecodeError propagates raw | red then green
TestOpenRouterImageEditTransformation::test_get_error_class | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:get_error_class: `status_code=status_code` -> `status_code=500` | red then green
TestOpenRouterImageEditTransformation::test_read_image_bytes_from_bytes | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:_read_image_bytes: bytes branch returns `image + b"pad"` | red then green
TestOpenRouterImageEditTransformation::test_read_image_bytes_from_bytesio | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:_read_image_bytes: BytesIO branch drops the `image.seek(current_pos)` restore | red then green
TestOpenRouterImageEditTransformation::test_read_image_bytes_unsupported_type | keep unit | mutated: litellm/llms/openrouter/image_edit/transformation.py:_read_image_bytes: returns b"" instead of raising ValueError | red then green

llms/openrouter/image_generation/test_openrouter_image_gen_transformation.py
TestOpenRouterImageGenerationTransformation::test_get_supported_openai_params | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:get_supported_openai_params: drops "n" from the list | red then green
TestOpenRouterImageGenerationTransformation::test_map_size_to_aspect_ratio_square | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:_map_size_to_aspect_ratio: "512x512" -> "2:3" | red then green
TestOpenRouterImageGenerationTransformation::test_map_size_to_aspect_ratio_landscape | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:_map_size_to_aspect_ratio: "1792x1024" -> "4:3" | red then green
TestOpenRouterImageGenerationTransformation::test_map_size_to_aspect_ratio_portrait | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:_map_size_to_aspect_ratio: "1024x1536" -> "3:2" | red then green
TestOpenRouterImageGenerationTransformation::test_map_size_to_aspect_ratio_auto | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:_map_size_to_aspect_ratio: "auto" -> "16:9" | red then green
TestOpenRouterImageGenerationTransformation::test_map_size_to_aspect_ratio_unknown | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:_map_size_to_aspect_ratio: `.get(size, "1:1")` -> `.get(size, "9:16")` | red then green
TestOpenRouterImageGenerationTransformation::test_map_quality_to_image_size_low | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:_map_quality_to_image_size: "low" -> "2K" | red then green
TestOpenRouterImageGenerationTransformation::test_map_quality_to_image_size_medium | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:_map_quality_to_image_size: "medium" -> "1K" | red then green
TestOpenRouterImageGenerationTransformation::test_map_quality_to_image_size_high | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:_map_quality_to_image_size: "high" -> "1K" | red then green
TestOpenRouterImageGenerationTransformation::test_map_quality_to_image_size_unknown | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:_map_quality_to_image_size: `.get(quality, "1K")` default | red then green
TestOpenRouterImageGenerationTransformation::test_map_openai_params_size_only | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:map_openai_params: `if key in supported_params` -> `not in` so size goes through the passthrough branch without image_config | red then green
TestOpenRouterImageGenerationTransformation::test_map_openai_params_quality_only | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:map_openai_params: same inverted-supported mutation | red then green
TestOpenRouterImageGenerationTransformation::test_map_openai_params_size_and_quality | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:map_openai_params: same inverted-supported mutation | red then green
TestOpenRouterImageGenerationTransformation::test_map_openai_params_with_n_parameter | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:map_openai_params: same inverted-supported mutation | red then green
TestOpenRouterImageGenerationTransformation::test_map_openai_params_unsupported_param_drop_false | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:map_openai_params: same inverted-supported mutation | red then green
TestOpenRouterImageGenerationTransformation::test_map_openai_params_unsupported_param_drop_true | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:map_openai_params: same inverted-supported mutation | red then green
TestOpenRouterImageGenerationTransformation::test_get_complete_url_default | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:get_complete_url: default literal -> "https://openrouter.ai/api/v1/images/generations" | red then green
TestOpenRouterImageGenerationTransformation::test_get_complete_url_with_custom_base | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:get_complete_url: appends "/images/generations" instead of "/chat/completions" | red then green
TestOpenRouterImageGenerationTransformation::test_get_complete_url_with_base_already_complete | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:get_complete_url: `endswith("/chat/completions")` -> `endswith("/v1")` | red then green
TestOpenRouterImageGenerationTransformation::test_validate_environment_with_api_key | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:validate_environment: `f"Bearer {api_key}"` -> `f"Token {api_key}"` | red then green
TestOpenRouterImageGenerationTransformation::test_validate_environment_with_secret_key | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:validate_environment: same Token prefix mutation | red then green
TestOpenRouterImageGenerationTransformation::test_transform_image_generation_request_basic | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:transform_image_generation_request: message role "user" -> "system" | red then green
TestOpenRouterImageGenerationTransformation::test_transform_image_generation_request_with_image_config | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:transform_image_generation_request: `if key not in [...]` -> `if key in []` so image_config/n are dropped | red then green
TestOpenRouterImageGenerationTransformation::test_transform_image_generation_response_with_base64_images | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:transform_image_generation_response: `image_url.startswith("data:")` -> `startswith("https:")` | red then green
TestOpenRouterImageGenerationTransformation::test_transform_image_generation_response_with_url_images | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:transform_image_generation_response: `url=image_url` -> `url=None` | red then green
TestOpenRouterImageGenerationTransformation::test_transform_image_generation_response_with_usage_and_cost | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:_set_usage_and_cost: `output_tokens=image_tokens` -> `output_tokens=total_tokens` | red then green
TestOpenRouterImageGenerationTransformation::test_transform_image_generation_response_multiple_images | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:transform_image_generation_response: `startswith("https:")` flip | red then green
TestOpenRouterImageGenerationTransformation::test_transform_image_generation_response_json_error | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:transform_image_generation_response: `except Exception` -> `except KeyError` | red then green
TestOpenRouterImageGenerationTransformation::test_transform_image_generation_response_transformation_error | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:transform_image_generation_response: `images = message.get("images", [])` -> `images = []` so no error is raised | red then green
TestOpenRouterImageGenerationTransformation::test_get_error_class | keep unit | mutated: litellm/llms/openrouter/image_generation/transformation.py:get_error_class: `status_code=status_code` -> `status_code=500` | red then green

User Flow

No user-facing behavior changes, test-only move

Relevant issues

Wave 2 phase 11 of the legacy test migration, follows #42103

Affected release

Linear ticket

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• I have added meaningful tests
• The handful of test files covering my change pass locally, e.g. uv run pytest tests/test_litellm/<your_test_file>.py -v. Leave the suites (make test-unit-*, make test-unit) to CI: it finishes in ~15 minutes where a laptop takes an hour or more
• My PR passes all required CI/CD checks (e.g., lint, schema.d.ts sync check, etc.)
• My PR's scope is as isolated as possible; it only solves 1 specific problem
• I have received a Greptile Confidence Score of at least 4/5 before requesting a maintainer review (Greptile reviews automatically once the PR is opened; only comment @greptileai to re-request a review after pushing changes)

Screenshots / Proof of Fix

Test-only move, proof is the CircleCI unit job junit linked above: 239 of 239 moved node ids passed

Link to Devin session: https://app.devin.ai/sessions/791d8c04e1e94bfe847e05193a265e16
Open in Devin Desktop: https://app.devin.ai/desktop/session/791d8c04e1e94bfe847e05193a265e16?variant=devin

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge because it only relocates unchanged tests into a recursively collected unit-test directory

Summary

This PR relocates 20 unchanged OpenAI, OpenAI-like, and OpenRouter test files into the mutation-tested unit suite and removes four orphaned package initializers

• All relocated files are exact renames with no test-content changes
• The CircleCI unit job recursively discovers their new paths under tests/unit
• No production behavior or security boundary changes

_{Reviews (1) · Last reviewed commit: "test: migrate openai, openai_like and op..."}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!