Underground-Digital/Workflow-Engine

1import os2from collections.abc import Generator3 4import pytest5 6from core.model_runtime.entities.llm_entities import LLMResult, LLMResultChunk, LLMResultChunkDelta7from core.model_runtime.entities.message_entities import AssistantPromptMessage, SystemPromptMessage, UserPromptMessage8from core.model_runtime.errors.validate import CredentialsValidateFailedError9from core.model_runtime.model_providers.replicate.llm.llm import ReplicateLargeLanguageModel10 11 12def test_validate_credentials():13    model = ReplicateLargeLanguageModel()14 15    with pytest.raises(CredentialsValidateFailedError):16        model.validate_credentials(17            model="meta/llama-2-13b-chat",18            credentials={19                "replicate_api_token": "invalid_key",20                "model_version": "f4e2de70d66816a838a89eeeb621910adffb0dd0baba3976c96980970978018d",21            },22        )23 24    model.validate_credentials(25        model="meta/llama-2-13b-chat",26        credentials={27            "replicate_api_token": os.environ.get("REPLICATE_API_KEY"),28            "model_version": "f4e2de70d66816a838a89eeeb621910adffb0dd0baba3976c96980970978018d",29        },30    )31 32 33def test_invoke_model():34    model = ReplicateLargeLanguageModel()35 36    response = model.invoke(37        model="meta/llama-2-13b-chat",38        credentials={39            "replicate_api_token": os.environ.get("REPLICATE_API_KEY"),40            "model_version": "f4e2de70d66816a838a89eeeb621910adffb0dd0baba3976c96980970978018d",41        },42        prompt_messages=[43            SystemPromptMessage(44                content="You are a helpful AI assistant.",45            ),46            UserPromptMessage(content="Who are you?"),47        ],48        model_parameters={49            "temperature": 1.0,50            "top_k": 2,51            "top_p": 0.5,52        },53        stop=["How"],54        stream=False,55        user="abc-123",56    )57 58    assert isinstance(response, LLMResult)59    assert len(response.message.content) > 060 61 62def test_invoke_stream_model():63    model = ReplicateLargeLanguageModel()64 65    response = model.invoke(66        model="mistralai/mixtral-8x7b-instruct-v0.1",67        credentials={68            "replicate_api_token": os.environ.get("REPLICATE_API_KEY"),69            "model_version": "2b56576fcfbe32fa0526897d8385dd3fb3d36ba6fd0dbe033c72886b81ade93e",70        },71        prompt_messages=[72            SystemPromptMessage(73                content="You are a helpful AI assistant.",74            ),75            UserPromptMessage(content="Who are you?"),76        ],77        model_parameters={78            "temperature": 1.0,79            "top_k": 2,80            "top_p": 0.5,81        },82        stop=["How"],83        stream=True,84        user="abc-123",85    )86 87    assert isinstance(response, Generator)88 89    for chunk in response:90        assert isinstance(chunk, LLMResultChunk)91        assert isinstance(chunk.delta, LLMResultChunkDelta)92        assert isinstance(chunk.delta.message, AssistantPromptMessage)93 94 95def test_get_num_tokens():96    model = ReplicateLargeLanguageModel()97 98    num_tokens = model.get_num_tokens(99        model="",100        credentials={101            "replicate_api_token": os.environ.get("REPLICATE_API_KEY"),102            "model_version": "2b56576fcfbe32fa0526897d8385dd3fb3d36ba6fd0dbe033c72886b81ade93e",103        },104        prompt_messages=[105            SystemPromptMessage(106                content="You are a helpful AI assistant.",107            ),108            UserPromptMessage(content="Hello World!"),109        ],110    )111 112    assert num_tokens == 14113