|
| 1 | +#!/usr/bin/env python |
| 2 | +###################################################################################################################### |
| 3 | +# Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved. # |
| 4 | +# # |
| 5 | +# Licensed under the Apache License, Version 2.0 (the "License"). You may not use this file except in compliance # |
| 6 | +# with the License. A copy of the License is located at # |
| 7 | +# # |
| 8 | +# http://www.apache.org/licenses/LICENSE-2.0 # |
| 9 | +# # |
| 10 | +# or in the 'license' file accompanying this file. This file is distributed on an 'AS IS' BASIS, WITHOUT WARRANTIES # |
| 11 | +# OR CONDITIONS OF ANY KIND, express or implied. See the License for the specific language governing permissions # |
| 12 | +# and limitations under the License. # |
| 13 | +###################################################################################################################### |
| 14 | + |
| 15 | +import json |
| 16 | +import os |
| 17 | +from copy import deepcopy |
| 18 | +from unittest.mock import patch |
| 19 | + |
| 20 | +import pytest |
| 21 | +from clients.builders.anthropic_builder import AnthropicBuilder |
| 22 | +from langchain.callbacks.streaming_aiter import AsyncIteratorCallbackHandler |
| 23 | +from llm_models.anthropic import AnthropicLLM |
| 24 | +from llm_models.rag.anthropic_retrieval import AnthropicRetrievalLLM |
| 25 | +from shared.memory.ddb_chat_memory import DynamoDBChatMemory |
| 26 | +from utils.constants import ( |
| 27 | + CONVERSATION_ID_EVENT_KEY, |
| 28 | + DEFAULT_ANTHROPIC_PLACEHOLDERS, |
| 29 | + DEFAULT_ANTHROPIC_PROMPT, |
| 30 | + DEFAULT_ANTHROPIC_RAG_PLACEHOLDERS, |
| 31 | + DEFAULT_ANTHROPIC_RAG_PROMPT, |
| 32 | + KENDRA_INDEX_ID_ENV_VAR, |
| 33 | + MEMORY_CONFIG, |
| 34 | + RAG_KEY, |
| 35 | + USER_ID_EVENT_KEY, |
| 36 | +) |
| 37 | +from utils.enum_types import ConversationMemoryTypes, LLMProviderTypes |
| 38 | + |
| 39 | + |
| 40 | +@pytest.mark.parametrize( |
| 41 | + "is_streaming, rag_enabled, llm_type, prompt, placeholders, rag_key", |
| 42 | + [ |
| 43 | + (False, False, AnthropicLLM, DEFAULT_ANTHROPIC_PROMPT, DEFAULT_ANTHROPIC_PLACEHOLDERS, ""), |
| 44 | + (True, False, AnthropicLLM, DEFAULT_ANTHROPIC_PROMPT, DEFAULT_ANTHROPIC_PLACEHOLDERS, ""), |
| 45 | + ( |
| 46 | + False, |
| 47 | + True, |
| 48 | + AnthropicRetrievalLLM, |
| 49 | + DEFAULT_ANTHROPIC_RAG_PROMPT, |
| 50 | + DEFAULT_ANTHROPIC_RAG_PLACEHOLDERS, |
| 51 | + RAG_KEY, |
| 52 | + ), |
| 53 | + (True, True, AnthropicRetrievalLLM, DEFAULT_ANTHROPIC_RAG_PROMPT, DEFAULT_ANTHROPIC_RAG_PLACEHOLDERS, RAG_KEY), |
| 54 | + ], |
| 55 | +) |
| 56 | +def test_set_llm_model( |
| 57 | + is_streaming, |
| 58 | + rag_enabled, |
| 59 | + llm_type, |
| 60 | + prompt, |
| 61 | + placeholders, |
| 62 | + rag_key, |
| 63 | + chat_event, |
| 64 | + llm_config, |
| 65 | + setup_environment, |
| 66 | + setup_secret, |
| 67 | +): |
| 68 | + config = json.loads(llm_config["Parameter"]["Value"]) |
| 69 | + chat_event_body = json.loads(chat_event["body"]) |
| 70 | + builder = AnthropicBuilder( |
| 71 | + llm_config=config, |
| 72 | + rag_enabled=rag_enabled, |
| 73 | + connection_id="fake-connection-id", |
| 74 | + conversation_id="fake-conversation-id", |
| 75 | + ) |
| 76 | + user_id = chat_event.get("requestContext", {}).get("authorizer", {}).get(USER_ID_EVENT_KEY, {}) |
| 77 | + |
| 78 | + # Assign all the values to the builder attributes required to construct the LLMChat object |
| 79 | + builder.set_knowledge_base() |
| 80 | + builder.set_memory_constants(LLMProviderTypes.ANTHROPIC.value) |
| 81 | + builder.set_conversation_memory(user_id, chat_event_body[CONVERSATION_ID_EVENT_KEY]) |
| 82 | + builder.set_api_key() |
| 83 | + |
| 84 | + if is_streaming: |
| 85 | + with patch( |
| 86 | + "clients.builders.llm_builder.WebsocketStreamingCallbackHandler", |
| 87 | + return_value=AsyncIteratorCallbackHandler(), |
| 88 | + ): |
| 89 | + builder.set_llm_model() |
| 90 | + else: |
| 91 | + builder.set_llm_model() |
| 92 | + |
| 93 | + assert type(builder.llm_model) == llm_type |
| 94 | + assert builder.llm_model.model == config["LlmParams"]["ModelId"] |
| 95 | + assert builder.llm_model.prompt_template.template == prompt |
| 96 | + assert set(builder.llm_model.prompt_template.input_variables) == set(placeholders) |
| 97 | + assert builder.llm_model.model_params == { |
| 98 | + "max_length": 100, |
| 99 | + "temperature": 0.2, |
| 100 | + "top_p": 0.2, |
| 101 | + } |
| 102 | + assert builder.llm_model.api_token == "fake-secret-value" |
| 103 | + assert builder.llm_model.streaming == config["LlmParams"]["Streaming"] |
| 104 | + assert builder.llm_model.verbose == config["LlmParams"]["Verbose"] |
| 105 | + if rag_enabled: |
| 106 | + assert builder.llm_model.knowledge_base.kendra_index_id == os.getenv(KENDRA_INDEX_ID_ENV_VAR) |
| 107 | + else: |
| 108 | + assert builder.llm_model.knowledge_base == None |
| 109 | + assert builder.llm_model.conversation_memory.memory_type == ConversationMemoryTypes.DynamoDB.value |
| 110 | + assert type(builder.llm_model.conversation_memory) == DynamoDBChatMemory |
| 111 | + assert ( |
| 112 | + builder.llm_model.conversation_memory.memory_key |
| 113 | + == MEMORY_CONFIG[LLMProviderTypes.ANTHROPIC.value + rag_key]["history"] |
| 114 | + ) |
| 115 | + assert ( |
| 116 | + builder.llm_model.conversation_memory.input_key |
| 117 | + == MEMORY_CONFIG[LLMProviderTypes.ANTHROPIC.value + rag_key]["input"] |
| 118 | + ) |
| 119 | + assert ( |
| 120 | + builder.llm_model.conversation_memory.output_key |
| 121 | + == MEMORY_CONFIG[LLMProviderTypes.ANTHROPIC.value + rag_key]["output"] |
| 122 | + ) |
| 123 | + assert ( |
| 124 | + builder.llm_model.conversation_memory.human_prefix |
| 125 | + == MEMORY_CONFIG[LLMProviderTypes.ANTHROPIC.value + rag_key]["human_prefix"] |
| 126 | + ) |
| 127 | + assert ( |
| 128 | + builder.llm_model.conversation_memory.ai_prefix |
| 129 | + == MEMORY_CONFIG[LLMProviderTypes.ANTHROPIC.value + rag_key]["ai_prefix"] |
| 130 | + ) |
| 131 | + |
| 132 | + if is_streaming: |
| 133 | + assert builder.callbacks |
| 134 | + else: |
| 135 | + assert builder.callbacks is None |
| 136 | + |
| 137 | + |
| 138 | +@pytest.mark.parametrize( |
| 139 | + "prompt, is_streaming, rag_enabled", |
| 140 | + [ |
| 141 | + (DEFAULT_ANTHROPIC_PROMPT, True, False), |
| 142 | + (DEFAULT_ANTHROPIC_PROMPT, False, False), |
| 143 | + (DEFAULT_ANTHROPIC_RAG_PROMPT, True, True), |
| 144 | + (DEFAULT_ANTHROPIC_RAG_PROMPT, False, True), |
| 145 | + ], |
| 146 | +) |
| 147 | +def test_set_llm_model_throws_error_missing_memory(llm_config, chat_event, setup_environment, setup_secret): |
| 148 | + config = json.loads(llm_config["Parameter"]["Value"]) |
| 149 | + builder = AnthropicBuilder( |
| 150 | + llm_config=config, |
| 151 | + rag_enabled=False, |
| 152 | + connection_id="fake-connection-id", |
| 153 | + conversation_id="fake-conversation-id", |
| 154 | + ) |
| 155 | + |
| 156 | + builder.set_knowledge_base() |
| 157 | + builder.set_api_key() |
| 158 | + with patch( |
| 159 | + "clients.builders.llm_builder.WebsocketStreamingCallbackHandler", |
| 160 | + return_value=AsyncIteratorCallbackHandler(), |
| 161 | + ): |
| 162 | + with pytest.raises(ValueError) as error: |
| 163 | + builder.set_llm_model() |
| 164 | + |
| 165 | + assert error.value.args[0] == "Conversation Memory was set to null." |
| 166 | + |
| 167 | + |
| 168 | +@pytest.mark.parametrize( |
| 169 | + "prompt, is_streaming, rag_enabled", |
| 170 | + [ |
| 171 | + (DEFAULT_ANTHROPIC_PROMPT, True, False), |
| 172 | + (DEFAULT_ANTHROPIC_PROMPT, False, False), |
| 173 | + (DEFAULT_ANTHROPIC_RAG_PROMPT, True, True), |
| 174 | + (DEFAULT_ANTHROPIC_RAG_PROMPT, False, True), |
| 175 | + ], |
| 176 | +) |
| 177 | +def test_set_llm_model_with_errors(llm_config): |
| 178 | + parsed_config = json.loads(llm_config["Parameter"]["Value"]) |
| 179 | + builder = AnthropicBuilder( |
| 180 | + llm_config=parsed_config, |
| 181 | + rag_enabled=False, |
| 182 | + connection_id="fake-connection-id", |
| 183 | + conversation_id="fake-conversation-id", |
| 184 | + ) |
| 185 | + builder.errors = ["some-error-1", "some-error-2"] |
| 186 | + builder.conversation_memory = "" |
| 187 | + |
| 188 | + with patch( |
| 189 | + "clients.builders.llm_builder.WebsocketStreamingCallbackHandler", |
| 190 | + return_value=AsyncIteratorCallbackHandler(), |
| 191 | + ): |
| 192 | + with pytest.raises(ValueError) as error: |
| 193 | + builder.set_llm_model() |
| 194 | + |
| 195 | + assert ( |
| 196 | + error.value.args[0] == "There are errors in the following configuration parameters:\nsome-error-1\nsome-error-2" |
| 197 | + ) |
| 198 | + |
| 199 | + |
| 200 | +@pytest.mark.parametrize( |
| 201 | + "prompt, is_streaming, rag_enabled", |
| 202 | + [ |
| 203 | + (DEFAULT_ANTHROPIC_PROMPT, True, False), |
| 204 | + (DEFAULT_ANTHROPIC_PROMPT, False, False), |
| 205 | + (DEFAULT_ANTHROPIC_RAG_PROMPT, True, True), |
| 206 | + (DEFAULT_ANTHROPIC_RAG_PROMPT, False, True), |
| 207 | + ], |
| 208 | +) |
| 209 | +def test_set_llm_model_with_missing_config_fields(llm_config): |
| 210 | + parsed_config = deepcopy(json.loads(llm_config["Parameter"]["Value"])) |
| 211 | + del parsed_config["LlmParams"] |
| 212 | + builder = AnthropicBuilder( |
| 213 | + llm_config=parsed_config, |
| 214 | + rag_enabled=False, |
| 215 | + connection_id="fake-connection-id", |
| 216 | + conversation_id="fake-conversation-id", |
| 217 | + ) |
| 218 | + |
| 219 | + with pytest.raises(ValueError) as error: |
| 220 | + builder.set_llm_model() |
| 221 | + |
| 222 | + assert ( |
| 223 | + error.value.args[0] |
| 224 | + == "There are errors in the following configuration parameters:\nMissing required field (LlmParams) containing LLM configuration in the config which is required to construct the LLM." |
| 225 | + ) |
| 226 | + |
| 227 | + |
| 228 | +@pytest.mark.parametrize( |
| 229 | + "prompt, is_streaming, rag_enabled, model", |
| 230 | + [ |
| 231 | + (DEFAULT_ANTHROPIC_PROMPT, False, False, AnthropicLLM), |
| 232 | + (DEFAULT_ANTHROPIC_PROMPT, True, False, AnthropicLLM), |
| 233 | + (DEFAULT_ANTHROPIC_RAG_PROMPT, False, True, AnthropicRetrievalLLM), |
| 234 | + (DEFAULT_ANTHROPIC_RAG_PROMPT, True, True, AnthropicRetrievalLLM), |
| 235 | + ], |
| 236 | +) |
| 237 | +def test_returned_anthropic_model(llm_config, chat_event, rag_enabled, model, setup_environment, setup_secret): |
| 238 | + config = json.loads(llm_config["Parameter"]["Value"]) |
| 239 | + chat_event_body = json.loads(chat_event["body"]) |
| 240 | + builder = AnthropicBuilder( |
| 241 | + connection_id="fake-connection-id", |
| 242 | + conversation_id="fake-conversation-id", |
| 243 | + llm_config=config, |
| 244 | + rag_enabled=rag_enabled, |
| 245 | + ) |
| 246 | + user_id = chat_event.get("requestContext", {}).get("authorizer", {}).get(USER_ID_EVENT_KEY, {}) |
| 247 | + |
| 248 | + builder.set_knowledge_base() |
| 249 | + builder.set_memory_constants(LLMProviderTypes.ANTHROPIC.value) |
| 250 | + builder.set_conversation_memory(user_id, chat_event_body[CONVERSATION_ID_EVENT_KEY]) |
| 251 | + builder.set_api_key() |
| 252 | + with patch( |
| 253 | + "clients.builders.llm_builder.WebsocketStreamingCallbackHandler", |
| 254 | + return_value=AsyncIteratorCallbackHandler(), |
| 255 | + ): |
| 256 | + builder.set_llm_model() |
| 257 | + assert type(builder.llm_model) == model |
0 commit comments