File size: 14,218 Bytes
db4a26f |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 |
{
"cells": [
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# Copyright (c) 2024 Microsoft Corporation.\n",
"# Licensed under the MIT License."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"import os\n",
"\n",
"import pandas as pd\n",
"import tiktoken\n",
"\n",
"from graphrag.query.context_builder.entity_extraction import EntityVectorStoreKey\n",
"from graphrag.query.indexer_adapters import (\n",
" read_indexer_covariates,\n",
" read_indexer_entities,\n",
" read_indexer_relationships,\n",
" read_indexer_reports,\n",
" read_indexer_text_units,\n",
")\n",
"from graphrag.query.input.loaders.dfs import (\n",
" store_entity_semantic_embeddings,\n",
")\n",
"from graphrag.query.llm.oai.chat_openai import ChatOpenAI\n",
"from graphrag.query.llm.oai.embedding import OpenAIEmbedding\n",
"from graphrag.query.llm.oai.typing import OpenaiApiType\n",
"from graphrag.query.question_gen.local_gen import LocalQuestionGen\n",
"from graphrag.query.structured_search.local_search.mixed_context import (\n",
" LocalSearchMixedContext,\n",
")\n",
"from graphrag.query.structured_search.local_search.search import LocalSearch\n",
"from graphrag.vector_stores.lancedb import LanceDBVectorStore"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Local Search Example\n",
"\n",
"Local search method generates answers by combining relevant data from the AI-extracted knowledge-graph with text chunks of the raw documents. This method is suitable for questions that require an understanding of specific entities mentioned in the documents (e.g. What are the healing properties of chamomile?)."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Load text units and graph data tables as context for local search\n",
"\n",
"- In this test we first load indexing outputs from parquet files to dataframes, then convert these dataframes into collections of data objects aligning with the knowledge model."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Load tables to dataframes"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"INPUT_DIR = \"./inputs/operation dulce\"\n",
"LANCEDB_URI = f\"{INPUT_DIR}/lancedb\"\n",
"\n",
"COMMUNITY_REPORT_TABLE = \"create_final_community_reports\"\n",
"ENTITY_TABLE = \"create_final_nodes\"\n",
"ENTITY_EMBEDDING_TABLE = \"create_final_entities\"\n",
"RELATIONSHIP_TABLE = \"create_final_relationships\"\n",
"COVARIATE_TABLE = \"create_final_covariates\"\n",
"TEXT_UNIT_TABLE = \"create_final_text_units\"\n",
"COMMUNITY_LEVEL = 2"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"#### Read entities"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# read nodes table to get community and degree data\n",
"entity_df = pd.read_parquet(f\"{INPUT_DIR}/{ENTITY_TABLE}.parquet\")\n",
"entity_embedding_df = pd.read_parquet(f\"{INPUT_DIR}/{ENTITY_EMBEDDING_TABLE}.parquet\")\n",
"\n",
"entities = read_indexer_entities(entity_df, entity_embedding_df, COMMUNITY_LEVEL)\n",
"\n",
"# load description embeddings to an in-memory lancedb vectorstore\n",
"# to connect to a remote db, specify url and port values.\n",
"description_embedding_store = LanceDBVectorStore(\n",
" collection_name=\"entity_description_embeddings\",\n",
")\n",
"description_embedding_store.connect(db_uri=LANCEDB_URI)\n",
"entity_description_embeddings = store_entity_semantic_embeddings(\n",
" entities=entities, vectorstore=description_embedding_store\n",
")\n",
"\n",
"print(f\"Entity count: {len(entity_df)}\")\n",
"entity_df.head()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"#### Read relationships"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"relationship_df = pd.read_parquet(f\"{INPUT_DIR}/{RELATIONSHIP_TABLE}.parquet\")\n",
"relationships = read_indexer_relationships(relationship_df)\n",
"\n",
"print(f\"Relationship count: {len(relationship_df)}\")\n",
"relationship_df.head()"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"covariate_df = pd.read_parquet(f\"{INPUT_DIR}/{COVARIATE_TABLE}.parquet\")\n",
"\n",
"claims = read_indexer_covariates(covariate_df)\n",
"\n",
"print(f\"Claim records: {len(claims)}\")\n",
"covariates = {\"claims\": claims}"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"#### Read community reports"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"report_df = pd.read_parquet(f\"{INPUT_DIR}/{COMMUNITY_REPORT_TABLE}.parquet\")\n",
"reports = read_indexer_reports(report_df, entity_df, COMMUNITY_LEVEL)\n",
"\n",
"print(f\"Report records: {len(report_df)}\")\n",
"report_df.head()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"#### Read text units"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"text_unit_df = pd.read_parquet(f\"{INPUT_DIR}/{TEXT_UNIT_TABLE}.parquet\")\n",
"text_units = read_indexer_text_units(text_unit_df)\n",
"\n",
"print(f\"Text unit records: {len(text_unit_df)}\")\n",
"text_unit_df.head()"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"api_key = os.environ[\"GRAPHRAG_API_KEY\"]\n",
"llm_model = os.environ[\"GRAPHRAG_LLM_MODEL\"]\n",
"embedding_model = os.environ[\"GRAPHRAG_EMBEDDING_MODEL\"]\n",
"\n",
"llm = ChatOpenAI(\n",
" api_key=api_key,\n",
" model=llm_model,\n",
" api_type=OpenaiApiType.OpenAI, # OpenaiApiType.OpenAI or OpenaiApiType.AzureOpenAI\n",
" max_retries=20,\n",
")\n",
"\n",
"token_encoder = tiktoken.get_encoding(\"cl100k_base\")\n",
"\n",
"text_embedder = OpenAIEmbedding(\n",
" api_key=api_key,\n",
" api_base=None,\n",
" api_type=OpenaiApiType.OpenAI,\n",
" model=embedding_model,\n",
" deployment_name=embedding_model,\n",
" max_retries=20,\n",
")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Create local search context builder"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"context_builder = LocalSearchMixedContext(\n",
" community_reports=reports,\n",
" text_units=text_units,\n",
" entities=entities,\n",
" relationships=relationships,\n",
" covariates=covariates,\n",
" entity_text_embeddings=description_embedding_store,\n",
" embedding_vectorstore_key=EntityVectorStoreKey.ID, # if the vectorstore uses entity title as ids, set this to EntityVectorStoreKey.TITLE\n",
" text_embedder=text_embedder,\n",
" token_encoder=token_encoder,\n",
")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Create local search engine"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# text_unit_prop: proportion of context window dedicated to related text units\n",
"# community_prop: proportion of context window dedicated to community reports.\n",
"# The remaining proportion is dedicated to entities and relationships. Sum of text_unit_prop and community_prop should be <= 1\n",
"# conversation_history_max_turns: maximum number of turns to include in the conversation history.\n",
"# conversation_history_user_turns_only: if True, only include user queries in the conversation history.\n",
"# top_k_mapped_entities: number of related entities to retrieve from the entity description embedding store.\n",
"# top_k_relationships: control the number of out-of-network relationships to pull into the context window.\n",
"# include_entity_rank: if True, include the entity rank in the entity table in the context window. Default entity rank = node degree.\n",
"# include_relationship_weight: if True, include the relationship weight in the context window.\n",
"# include_community_rank: if True, include the community rank in the context window.\n",
"# return_candidate_context: if True, return a set of dataframes containing all candidate entity/relationship/covariate records that\n",
"# could be relevant. Note that not all of these records will be included in the context window. The \"in_context\" column in these\n",
"# dataframes indicates whether the record is included in the context window.\n",
"# max_tokens: maximum number of tokens to use for the context window.\n",
"\n",
"\n",
"local_context_params = {\n",
" \"text_unit_prop\": 0.5,\n",
" \"community_prop\": 0.1,\n",
" \"conversation_history_max_turns\": 5,\n",
" \"conversation_history_user_turns_only\": True,\n",
" \"top_k_mapped_entities\": 10,\n",
" \"top_k_relationships\": 10,\n",
" \"include_entity_rank\": True,\n",
" \"include_relationship_weight\": True,\n",
" \"include_community_rank\": False,\n",
" \"return_candidate_context\": False,\n",
" \"embedding_vectorstore_key\": EntityVectorStoreKey.ID, # set this to EntityVectorStoreKey.TITLE if the vectorstore uses entity title as ids\n",
" \"max_tokens\": 12_000, # change this based on the token limit you have on your model (if you are using a model with 8k limit, a good setting could be 5000)\n",
"}\n",
"\n",
"llm_params = {\n",
" \"max_tokens\": 2_000, # change this based on the token limit you have on your model (if you are using a model with 8k limit, a good setting could be 1000=1500)\n",
" \"temperature\": 0.0,\n",
"}"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"search_engine = LocalSearch(\n",
" llm=llm,\n",
" context_builder=context_builder,\n",
" token_encoder=token_encoder,\n",
" llm_params=llm_params,\n",
" context_builder_params=local_context_params,\n",
" response_type=\"multiple paragraphs\", # free form text describing the response type and format, can be anything, e.g. prioritized list, single paragraph, multiple paragraphs, multiple-page report\n",
")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Run local search on sample queries"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"result = await search_engine.asearch(\"Tell me about Agent Mercer\")\n",
"print(result.response)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"question = \"Tell me about Dr. Jordan Hayes\"\n",
"result = await search_engine.asearch(question)\n",
"print(result.response)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"#### Inspecting the context data used to generate the response"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"result.context_data[\"entities\"].head()"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"result.context_data[\"relationships\"].head()"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"result.context_data[\"reports\"].head()"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"result.context_data[\"sources\"].head()"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"if \"claims\" in result.context_data:\n",
" print(result.context_data[\"claims\"].head())"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Question Generation"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"This function takes a list of user queries and generates the next candidate questions."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"question_generator = LocalQuestionGen(\n",
" llm=llm,\n",
" context_builder=context_builder,\n",
" token_encoder=token_encoder,\n",
" llm_params=llm_params,\n",
" context_builder_params=local_context_params,\n",
")"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"question_history = [\n",
" \"Tell me about Agent Mercer\",\n",
" \"What happens in Dulce military base?\",\n",
"]\n",
"candidate_questions = await question_generator.agenerate(\n",
" question_history=question_history, context_data=None, question_count=5\n",
")\n",
"print(candidate_questions.response)"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.10.12"
}
},
"nbformat": 4,
"nbformat_minor": 2
}
|