microsoft/TypeAgent
Publicmirrored from https://github.com/microsoft/TypeAgentAvailable
python/ta/test/fixtures.py
303lines · modecode
| 1 | # Copyright (c) Microsoft Corporation. |
| 2 | # Licensed under the MIT License. |
| 3 | |
| 4 | from collections.abc import AsyncGenerator, Iterator |
| 5 | import os |
| 6 | import tempfile |
| 7 | from typing import assert_never |
| 8 | |
| 9 | import pytest |
| 10 | import pytest_asyncio |
| 11 | |
| 12 | from typeagent.aitools import utils |
| 13 | from typeagent.aitools.embeddings import AsyncEmbeddingModel, TEST_MODEL_NAME |
| 14 | from typeagent.aitools.vectorbase import TextEmbeddingIndexSettings |
| 15 | from typeagent.knowpro import kplib |
| 16 | from typeagent.knowpro.collections import ( |
| 17 | MemoryMessageCollection, |
| 18 | MemorySemanticRefCollection, |
| 19 | ) |
| 20 | from typeagent.knowpro.convsettings import ConversationSettings |
| 21 | from typeagent.knowpro.interfaces import ( |
| 22 | DeletionInfo, |
| 23 | ICollection, |
| 24 | IConversation, |
| 25 | IConversationSecondaryIndexes, |
| 26 | IMessage, |
| 27 | IMessageCollection, |
| 28 | ISemanticRefCollection, |
| 29 | IStorageProvider, |
| 30 | ITermToSemanticRefIndex, |
| 31 | SemanticRef, |
| 32 | ScoredSemanticRefOrdinal, |
| 33 | TextLocation, |
| 34 | TextRange, |
| 35 | ) |
| 36 | from typeagent.knowpro.kplib import KnowledgeResponse |
| 37 | from typeagent.knowpro.messageindex import MessageTextIndexSettings |
| 38 | from typeagent.knowpro.reltermsindex import RelatedTermIndexSettings |
| 39 | from typeagent.knowpro.secindex import ConversationSecondaryIndexes |
| 40 | from typeagent.storage.memorystore import MemoryStorageProvider |
| 41 | from typeagent.storage.sqlitestore import SqliteStorageProvider |
| 42 | |
| 43 | |
| 44 | @pytest.fixture(scope="session") |
| 45 | def needs_auth() -> None: |
| 46 | utils.load_dotenv() |
| 47 | |
| 48 | |
| 49 | @pytest.fixture(scope="session") |
| 50 | def embedding_model() -> AsyncEmbeddingModel: |
| 51 | """Fixture to create a test embedding model with small embedding size for faster tests.""" |
| 52 | return AsyncEmbeddingModel(model_name=TEST_MODEL_NAME) |
| 53 | |
| 54 | |
| 55 | @pytest.fixture |
| 56 | def temp_dir() -> Iterator[str]: |
| 57 | with tempfile.TemporaryDirectory() as dir: |
| 58 | yield dir |
| 59 | |
| 60 | |
| 61 | @pytest.fixture |
| 62 | def temp_db_path() -> Iterator[str]: |
| 63 | """Create a temporary SQLite database file for testing.""" |
| 64 | fd, path = tempfile.mkstemp(suffix=".sqlite") |
| 65 | os.close(fd) |
| 66 | yield path |
| 67 | if os.path.exists(path): |
| 68 | os.remove(path) |
| 69 | |
| 70 | |
| 71 | @pytest_asyncio.fixture |
| 72 | async def memory_storage(embedding_model: AsyncEmbeddingModel) -> MemoryStorageProvider: |
| 73 | """Create a MemoryStorageProvider for testing.""" |
| 74 | embedding_settings = TextEmbeddingIndexSettings(embedding_model) |
| 75 | message_text_settings = MessageTextIndexSettings(embedding_settings) |
| 76 | related_terms_settings = RelatedTermIndexSettings(embedding_settings) |
| 77 | |
| 78 | return await MemoryStorageProvider.create( |
| 79 | message_text_settings=message_text_settings, |
| 80 | related_terms_settings=related_terms_settings, |
| 81 | ) |
| 82 | |
| 83 | |
| 84 | @pytest_asyncio.fixture |
| 85 | async def sqlite_storage( |
| 86 | temp_db_path: str, embedding_model: AsyncEmbeddingModel |
| 87 | ) -> AsyncGenerator[SqliteStorageProvider, None]: |
| 88 | """Create a SqliteStorageProvider for testing.""" |
| 89 | embedding_settings = TextEmbeddingIndexSettings(embedding_model) |
| 90 | message_text_settings = MessageTextIndexSettings(embedding_settings) |
| 91 | related_terms_settings = RelatedTermIndexSettings(embedding_settings) |
| 92 | |
| 93 | provider = await SqliteStorageProvider.create( |
| 94 | message_text_settings, related_terms_settings, temp_db_path |
| 95 | ) |
| 96 | yield provider |
| 97 | await provider.close() |
| 98 | |
| 99 | |
| 100 | @pytest_asyncio.fixture(params=["memory", "sqlite"]) |
| 101 | async def storage_provider_type( |
| 102 | request: pytest.FixtureRequest, |
| 103 | embedding_model: AsyncEmbeddingModel, |
| 104 | temp_db_path: str, |
| 105 | ) -> AsyncGenerator[tuple[IStorageProvider, str], None]: |
| 106 | """Parameterized fixture that provides both memory and sqlite storage providers.""" |
| 107 | embedding_settings = TextEmbeddingIndexSettings(embedding_model) |
| 108 | message_text_settings = MessageTextIndexSettings(embedding_settings) |
| 109 | related_terms_settings = RelatedTermIndexSettings(embedding_settings) |
| 110 | |
| 111 | match request.param: |
| 112 | case "memory": |
| 113 | provider = await MemoryStorageProvider.create( |
| 114 | message_text_settings=message_text_settings, |
| 115 | related_terms_settings=related_terms_settings, |
| 116 | ) |
| 117 | yield provider, request.param |
| 118 | case "sqlite": |
| 119 | provider = await SqliteStorageProvider.create( |
| 120 | message_text_settings, related_terms_settings, temp_db_path |
| 121 | ) |
| 122 | yield provider, request.param |
| 123 | await provider.close() |
| 124 | case _: |
| 125 | assert_never(request.param) |
| 126 | |
| 127 | |
| 128 | # Unified fake message and conversation classes for testing |
| 129 | |
| 130 | |
| 131 | class FakeMessage(IMessage): |
| 132 | """Unified message implementation for testing purposes.""" |
| 133 | |
| 134 | def __init__( |
| 135 | self, text_chunks: list[str] | str, message_ordinal: int | None = None |
| 136 | ): |
| 137 | if isinstance(text_chunks, str): |
| 138 | self.text_chunks = [text_chunks] |
| 139 | else: |
| 140 | self.text_chunks = text_chunks |
| 141 | |
| 142 | # Handle timestamp - for compatibility with mock message pattern |
| 143 | if message_ordinal is not None: |
| 144 | self.ordinal = message_ordinal |
| 145 | self.timestamp = f"2020-01-01T{message_ordinal:02d}:00:00" |
| 146 | else: |
| 147 | self.timestamp = None |
| 148 | |
| 149 | self.tags: list[str] = [] |
| 150 | self.deletion_info: DeletionInfo | None = None |
| 151 | self.text_location = TextLocation(0, 0) |
| 152 | |
| 153 | def get_knowledge(self) -> KnowledgeResponse: |
| 154 | return KnowledgeResponse( |
| 155 | entities=[], |
| 156 | actions=[], |
| 157 | inverse_actions=[], |
| 158 | topics=[], |
| 159 | ) |
| 160 | |
| 161 | def get_text(self) -> str: |
| 162 | return " ".join(self.text_chunks) |
| 163 | |
| 164 | def get_text_location(self) -> TextLocation: |
| 165 | return self.text_location |
| 166 | |
| 167 | |
| 168 | class FakeMessageCollection(MemoryMessageCollection[FakeMessage]): |
| 169 | """Message collection for testing.""" |
| 170 | |
| 171 | pass |
| 172 | |
| 173 | |
| 174 | class FakeTermIndex(ITermToSemanticRefIndex): |
| 175 | """Simple term index for testing.""" |
| 176 | |
| 177 | def __init__( |
| 178 | self, term_to_refs: dict[str, list[ScoredSemanticRefOrdinal]] | None = None |
| 179 | ): |
| 180 | self.term_to_refs = term_to_refs or {} |
| 181 | |
| 182 | async def size(self) -> int: |
| 183 | return len(self.term_to_refs) |
| 184 | |
| 185 | async def get_terms(self) -> list[str]: |
| 186 | return list(self.term_to_refs.keys()) |
| 187 | |
| 188 | async def add_term( |
| 189 | self, |
| 190 | term: str, |
| 191 | semantic_ref_ordinal: int | ScoredSemanticRefOrdinal, |
| 192 | ) -> str: |
| 193 | if term not in self.term_to_refs: |
| 194 | self.term_to_refs[term] = [] |
| 195 | if isinstance(semantic_ref_ordinal, int): |
| 196 | scored_ref = ScoredSemanticRefOrdinal(semantic_ref_ordinal, 1.0) |
| 197 | else: |
| 198 | scored_ref = semantic_ref_ordinal |
| 199 | self.term_to_refs[term].append(scored_ref) |
| 200 | return term |
| 201 | |
| 202 | async def remove_term(self, term: str, semantic_ref_ordinal: int) -> None: |
| 203 | if term in self.term_to_refs: |
| 204 | self.term_to_refs[term] = [ |
| 205 | ref |
| 206 | for ref in self.term_to_refs[term] |
| 207 | if ref.semantic_ref_ordinal != semantic_ref_ordinal |
| 208 | ] |
| 209 | if not self.term_to_refs[term]: |
| 210 | del self.term_to_refs[term] |
| 211 | |
| 212 | async def lookup_term(self, term: str) -> list[ScoredSemanticRefOrdinal] | None: |
| 213 | return self.term_to_refs.get(term) |
| 214 | |
| 215 | |
| 216 | class FakeConversation(IConversation[FakeMessage, FakeTermIndex]): |
| 217 | """Unified conversation implementation for testing purposes.""" |
| 218 | |
| 219 | def __init__( |
| 220 | self, |
| 221 | name_tag: str = "FakeConversation", |
| 222 | messages: list[FakeMessage] | None = None, |
| 223 | semantic_refs: list[SemanticRef] | None = None, |
| 224 | storage_provider: IStorageProvider | None = None, |
| 225 | has_secondary_indexes: bool = True, |
| 226 | ): |
| 227 | self.name_tag = name_tag |
| 228 | self.tags: list[str] = [] |
| 229 | |
| 230 | # Set up messages |
| 231 | if messages is None: |
| 232 | messages = [FakeMessage("Hello world")] |
| 233 | self.messages: IMessageCollection[FakeMessage] = FakeMessageCollection(messages) |
| 234 | |
| 235 | # Set up semantic refs |
| 236 | self.semantic_refs: ISemanticRefCollection = MemorySemanticRefCollection( |
| 237 | semantic_refs or [] |
| 238 | ) |
| 239 | |
| 240 | # Set up term index |
| 241 | self.semantic_ref_index: FakeTermIndex | None = FakeTermIndex() |
| 242 | |
| 243 | # Store settings with storage provider for access via conversation.settings.storage_provider |
| 244 | if storage_provider is None: |
| 245 | # Default storage provider will be created lazily in async context |
| 246 | self._needs_async_init = True |
| 247 | self.secondary_indexes = None |
| 248 | self._storage_provider = None |
| 249 | self._has_secondary_indexes = has_secondary_indexes |
| 250 | else: |
| 251 | # Create test model for settings |
| 252 | test_model = AsyncEmbeddingModel(model_name=TEST_MODEL_NAME) |
| 253 | self.settings = ConversationSettings(test_model, storage_provider) |
| 254 | self._needs_async_init = False |
| 255 | self._storage_provider = storage_provider |
| 256 | |
| 257 | if has_secondary_indexes: |
| 258 | # Set up secondary indexes |
| 259 | embedding_settings = TextEmbeddingIndexSettings(test_model) |
| 260 | related_terms_settings = RelatedTermIndexSettings(embedding_settings) |
| 261 | self.secondary_indexes: ( |
| 262 | IConversationSecondaryIndexes[FakeMessage] | None |
| 263 | ) = ConversationSecondaryIndexes( |
| 264 | storage_provider, related_terms_settings |
| 265 | ) |
| 266 | else: |
| 267 | self.secondary_indexes = None |
| 268 | |
| 269 | async def ensure_initialized(self): |
| 270 | """Ensure async initialization is complete.""" |
| 271 | if self._needs_async_init: |
| 272 | test_model = AsyncEmbeddingModel(model_name=TEST_MODEL_NAME) |
| 273 | self.settings = ConversationSettings(test_model) |
| 274 | storage_provider = await self.settings.get_storage_provider() |
| 275 | self._storage_provider = storage_provider |
| 276 | if self.semantic_ref_index is None: |
| 277 | self.semantic_ref_index = await storage_provider.get_semantic_ref_index() # type: ignore |
| 278 | |
| 279 | if self._has_secondary_indexes: |
| 280 | # Set up secondary indexes |
| 281 | embedding_settings = TextEmbeddingIndexSettings(test_model) |
| 282 | related_terms_settings = RelatedTermIndexSettings(embedding_settings) |
| 283 | self.secondary_indexes = ConversationSecondaryIndexes( |
| 284 | storage_provider, related_terms_settings |
| 285 | ) |
| 286 | else: |
| 287 | self.secondary_indexes = None |
| 288 | |
| 289 | self._needs_async_init = False |
| 290 | |
| 291 | |
| 292 | @pytest.fixture |
| 293 | def fake_conversation() -> FakeConversation: |
| 294 | """Fixture to create a FakeConversation instance.""" |
| 295 | return FakeConversation() |
| 296 | |
| 297 | |
| 298 | @pytest.fixture |
| 299 | async def fake_conversation_with_storage( |
| 300 | memory_storage: MemoryStorageProvider, |
| 301 | ) -> FakeConversation: |
| 302 | """Fixture to create a FakeConversation instance with storage provider.""" |
| 303 | return FakeConversation(storage_provider=memory_storage) |
| 304 | |