File size: 36,478 Bytes
5374a2d |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 |
import os
import json
import asyncio
from uuid import uuid4
from datetime import datetime
from typing import List, Union, Optional, Sequence, Dict, Any, Tuple
from llama_index.core.schema import NodeWithScore, TextNode, ImageNode, RelatedNodeInfo
from .rag_config import RAGConfig
from .readers import LLamaIndexReader, MultimodalReader
from .indexings import IndexFactory, BaseIndexWrapper
from .chunkers import ChunkFactory
from .embeddings import EmbeddingFactory, EmbeddingProvider
from .retrievers import RetrieverFactory, BaseRetrieverWrapper
from .postprocessors import PostprocessorFactory
from .indexings.base import IndexType
from .retrievers.base import RetrieverType
from .schema import Corpus, ChunkMetadata, IndexMetadata, Query, RagResult, ImageChunk, TextChunk
from evoagentx.storages.base import StorageHandler
from evoagentx.storages.schema import IndexStore
from evoagentx.models.base_model import BaseLLM
from evoagentx.core.logging import logger
class RAGEngine:
def __init__(self, config: RAGConfig, storage_handler: StorageHandler, llm: Optional[BaseLLM] = None):
self.config = config
self.storage_handler = storage_handler # Maybe reinit the vector_store by the load funcion.
self.embedding_factory = EmbeddingFactory()
self.index_factory = IndexFactory()
self.chunk_factory = ChunkFactory()
self.retriever_factory = RetrieverFactory()
self.postprocessor_factory = PostprocessorFactory()
self.llm = llm # LLM for entity extractor
# Set chunk class based on modality config
logger.info(f"RAGEngine modality config: {self.config.modality}")
if self.config.modality == "multimodal":
self.chunk_class = ImageChunk
else:
self.chunk_class = TextChunk
# Initialize reader based on modality
if self.config.modality == "multimodal":
self.reader = MultimodalReader(
recursive=self.config.reader.recursive,
exclude_hidden=self.config.reader.exclude_hidden,
num_files_limits=self.config.reader.num_files_limit,
errors=self.config.reader.errors
)
else:
self.reader = LLamaIndexReader(
recursive=self.config.reader.recursive,
exclude_hidden=self.config.reader.exclude_hidden,
num_workers=self.config.num_workers,
num_files_limits=self.config.reader.num_files_limit,
custom_metadata_function=self.config.reader.custom_metadata_function,
extern_file_extractor=self.config.reader.extern_file_extractor,
errors=self.config.reader.errors,
encoding=self.config.reader.encoding
)
# Initialize embedding model.
self.embed_model = self.embedding_factory.create(
provider=self.config.embedding.provider,
model_config=self.config.embedding.model_dump(exclude_unset=True),
)
# Dynamic Check the dimensions in StorageHandler
if (self.storage_handler.vector_store is not None) and (self.embed_model.dimensions is not None):
if self.storage_handler.storageConfig.vectorConfig.dimensions != self.embed_model.dimensions:
logger.warning("The dimensions in vector_store is not equal with embed_model. Reiniliaze vector_store.")
self.storage_handler.storageConfig.vectorConfig.dimensions = self.embed_model.dimensions
self.storage_handler._init_vector_store()
# Initialize chunker (skip for multimodal)
if self.config.modality == "multimodal":
self.chunker = None # No chunking for images
else:
self.chunker = self.chunk_factory.create(
strategy=self.config.chunker.strategy,
embed_model=self.embed_model.get_embedding_model(),
chunker_config={
"chunk_size": self.config.chunker.chunk_size,
"chunk_overlap": self.config.chunker.chunk_overlap,
"max_chunks": self.config.chunker.max_chunks
}
)
# Initialize indices and retrievers
self.indices: Dict[str, Dict[str, BaseIndexWrapper]] = {} # Nested: {corpus_id: {index_type: index}}
self.retrievers: Dict[str, Dict[str, BaseRetrieverWrapper]] = {} # Nested: {corpus_id: {index_type: retriever}}
def read(self, file_paths: Union[Sequence[str], str],
exclude_files: Optional[Union[str, List, Tuple, Sequence]] = None,
filter_file_by_suffix: Optional[Union[str, List, Tuple, Sequence]] = None,
merge_by_file: bool = False,
show_progress: bool = False,
corpus_id: str = None) -> Corpus:
"""Load and chunk documents from files.
Reads files from specified paths, processes them into documents, and chunks them into a Corpus.
Args:
file_paths (Union[Sequence[str], str]): Path(s) to files or directories.
exclude_files (Optional[Union[str, List, Tuple, Sequence]]): Files to exclude.
filter_file_by_suffix (Optional[Union[str, List, Tuple, Sequence]]): Filter files by suffix (e.g., '.pdf').
merge_by_file (bool): Merge documents by file.
show_progress (bool): Show loading progress.
corpus_id (Optional[str]): Identifier for the corpus. Defaults to a UUID if None.
Returns:
Corpus: The chunked corpus containing processed document chunks.
Raises:
Exception: If document reading or chunking fails.
"""
try:
corpus_id = corpus_id or str(uuid4())
documents = self.reader.load(
file_paths=file_paths,
exclude_files=exclude_files,
filter_file_by_suffix=filter_file_by_suffix,
merge_by_file=merge_by_file,
show_progress=show_progress
)
if self.config.modality == "multimodal":
# No chunking - convert ImageDocuments directly to ImageChunks
image_chunks = []
for doc in documents:
# Get image path from document - try multiple possible attributes
image_path = getattr(doc, 'image_path', None) or doc.metadata.get('file_path')
image_mimetype = getattr(doc, 'image_mimetype', None)
image_chunk = self.chunk_class(
image_path=image_path,
image_mimetype=image_mimetype,
chunk_id=doc.metadata.get('file_name', f'img_{len(image_chunks)}'),
metadata=ChunkMetadata(
doc_id=doc.metadata.get('file_name', f'doc_{len(image_chunks)}'),
corpus_id=corpus_id,
**doc.metadata
)
)
image_chunks.append(image_chunk)
corpus = Corpus(chunks=image_chunks, corpus_id=corpus_id)
logger.info(f"Read {len(documents)} multimodal documents (no chunking) for corpus {corpus_id}")
else:
corpus = self.chunker.chunk(documents)
corpus.corpus_id = corpus_id
logger.info(f"Read {len(documents)} documents and created {len(corpus.chunks)} chunks for corpus {corpus_id}")
return corpus
except Exception as e:
logger.error(f"Failed to read documents for corpus {corpus_id}: {str(e)}")
raise
def add(self, index_type: str, nodes: Union[Corpus, List[NodeWithScore], List[TextNode], List[ImageNode]],
corpus_id: str = None) -> None:
"""Add nodes to an index for a specific corpus.
Initializes an index if it doesn't exist and inserts nodes, updating metadata with corpus_id and index_type.
Args:
index_type (str): Type of index (e.g., VECTOR, GRAPH).
nodes (Union[Corpus, List[NodeWithScore], List[TextNode]]): Nodes or Corpus to add.
corpus_id (str, optional): Identifier for the corpus. Defaults to a UUID if None.
Return:
return a sequence with id of each added node.
Raises:
Exception: If index creation or node insertion fails.
"""
try:
corpus_id = corpus_id or str(uuid4())
if corpus_id not in self.indices:
self.indices[corpus_id] = {}
self.retrievers[corpus_id] = {}
if index_type not in self.indices[corpus_id]:
index = self.index_factory.create(
index_type=index_type,
embed_model=self.embed_model.get_embedding_model(),
storage_handler=self.storage_handler,
index_config=self.config.index.model_dump(exclude_unset=True) if self.config.index else {},
llm=self.llm,
)
self.indices[corpus_id][index_type] = index
self.retrievers[corpus_id][index_type] = self.retriever_factory.create(
retriever_type=self.config.retrieval.retrivel_type,
llm=self.llm,
index=index.get_index(),
graph_store=index.get_index().storage_context.graph_store,
embed_model=self.embed_model.get_embedding_model(),
query=Query(query_str="", top_k=self.config.retrieval.top_k if self.config.retrieval else 5),
storage_handler=self.storage_handler,
chunk_class=self.chunk_class
)
nodes_to_insert = nodes.to_llama_nodes() if isinstance(nodes, Corpus) else nodes
for node in nodes_to_insert:
node.metadata.update({"corpus_id": corpus_id, "index_type": index_type})
nodes_ids = self.indices[corpus_id][index_type].insert_nodes(nodes_to_insert)
logger.info(f"Added {len(nodes_to_insert)} nodes to {index_type} index for corpus {corpus_id}")
return nodes_ids
except Exception as e:
logger.error(f"Failed to add nodes to {index_type} index for corpus {corpus_id}: {str(e)}")
return []
def delete(self, corpus_id: str, index_type: Optional[str] = None,
node_ids: Optional[Union[str, List[str]]] = None,
metadata_filters: Optional[Dict[str, Any]] = None) -> None:
"""Delete nodes or an entire index from a corpus.
Removes specific nodes by ID or metadata filters, or deletes the entire index if no filters are provided.
Args:
corpus_id (str): Identifier for the corpus.
index_type (Optional[IndexType]): Specific index type to delete from. If None, affects all indices.
node_ids (Union[str, Optional[List[str]]]): List of node IDs to delete.
metadata_filters (Optional[Dict[str, Any]]): Metadata filters to select nodes for deletion.
Raises:
Exception: If deletion fails.
"""
try:
if corpus_id not in self.indices:
logger.warning(f"No indices found for corpus {corpus_id}")
return
target_indices = [index_type] if index_type else self.indices[corpus_id].keys()
for idx_type in list(target_indices): # Use list to avoid runtime modification issues
if idx_type not in self.indices[corpus_id]:
logger.warning(f"Index type {idx_type} not found for corpus {corpus_id}")
continue
index = self.indices[corpus_id][idx_type]
if node_ids or metadata_filters:
# Convert single node_id to list for consistency
node_ids_list = [node_ids] if isinstance(node_ids, str) else node_ids
index.delete_nodes(node_ids=node_ids_list, metadata_filters=metadata_filters)
logger.info(f"Deleted nodes from {idx_type} index for corpus {corpus_id}")
else:
# Delete entire index
index.clear()
del self.indices[corpus_id][idx_type]
del self.retrievers[corpus_id][idx_type]
logger.info(f"Deleted entire {idx_type} index for corpus {corpus_id}")
# Clean up corpus if no indices remain
if not self.indices[corpus_id]:
del self.indices[corpus_id]
del self.retrievers[corpus_id]
logger.info(f"Removed empty corpus {corpus_id}")
except Exception as e:
logger.error(f"Failed to delete from corpus {corpus_id}, index {index_type}: {str(e)}")
raise
def clear(self, corpus_id: Optional[str] = None) -> None:
"""Clear all indices for a specific corpus or all corpora.
Args:
corpus_id (Optional[str]): Specific corpus to clear. If None, clears all corpora.
Raises:
Exception: If clearing fails.
"""
try:
target_corpora = [corpus_id] if corpus_id else list(self.indices.keys())
for cid in target_corpora:
if cid not in self.indices:
logger.warning(f"No indices found for corpus {cid}")
continue
for idx_type in list(self.indices[cid].keys()):
index = self.indices[cid][idx_type]
index.clear()
del self.indices[cid][idx_type]
del self.retrievers[cid][idx_type]
logger.info(f"Cleared {idx_type} index for corpus {cid}")
# Clean up corpus if no indices remain
del self.indices[cid]
del self.retrievers[cid]
logger.info(f"Cleared corpus {cid}")
except Exception as e:
logger.error(f"Failed to clear indices for corpus {corpus_id or 'all'}: {str(e)}")
raise
def save(self, output_path: Optional[str] = None, corpus_id: Optional[str] = None,
index_type: Optional[str] = None, table: Optional[str] = None,
graph_exported: bool = False) -> None:
"""Save indices to files or database.
Serializes corpus chunks to JSONL files and metadata to JSON files if output_path is provided,
or saves to the SQLite database via StorageHandler if output_path is None.
Args:
output_path (Optional[str]): Directory to save JSONL and JSON files. If None, saves to database.
corpus_id (Optional[str]): Specific corpus to save. If None, saves all corpora.
index_type (Optional[str]): Specific index type to save. If None, saves all indices.
table (Optional[str]): Database table name for index data. Defaults to 'indexing' if None.
graph_exported (bool): If True, export graph nodes and relations for graph indices. Defaults to False.
Raises:
Exception: If saving fails or file operations encounter errors.
"""
try:
target_corpora = [corpus_id] if corpus_id else list(self.indices.keys())
table = table or "indexing"
for cid in target_corpora:
if cid not in self.indices:
logger.warning(f"No indices found for corpus {cid}")
continue
target_indices = [index_type] if index_type and index_type in self.indices[cid] else self.indices[cid].keys()
for idx_type in target_indices:
index = self.indices[cid][idx_type]
# Skip saving for graph indices unless graph_exported is True
if idx_type == IndexType.GRAPH and not graph_exported:
logger.warning(f"Skipping save for graph index {idx_type} in corpus {cid} as graph_exported is False")
continue
# For graph indices, include kg_nodes and kg_rels if graph_exported is True
if idx_type == IndexType.GRAPH and graph_exported:
index.build_kv_store() # FIXME: Computer's memory may increase dramatically
# Convert index nodes to Corpus
chunks = [
self.chunk_class.from_llama_node(node_data)
for node_id, node_data in index.id_to_node.items()
]
corpus = Corpus(chunks=chunks, corpus_id=cid)
# Prepare metadata
vector_config = self.storage_handler.storageConfig.vectorConfig.model_dump() if self.storage_handler.storageConfig.vectorConfig else {}
graph_config = self.storage_handler.storageConfig.graphConfig.model_dump() if self.storage_handler.storageConfig.graphConfig else {}
metadata = IndexMetadata(
corpus_id=cid,
index_type=idx_type,
collection_name=vector_config.get("qdrant_collection_name", "default_collection"),
dimension=self.embed_model.dimensions,
vector_db_type=vector_config.get("vector_name", None),
graph_db_type=graph_config.get("graph_name", None),
embedding_model_name=self.config.embedding.model_name,
date=str(datetime.now()),
)
if output_path:
# File-based saving
os.makedirs(output_path, exist_ok=True)
safe_cid = "".join(c if c.isalnum() or c in ["-", "_"] else "_" for c in cid)
safe_idx_type = "".join(c if c.isalnum() or c in ["-", "_"] else "_" for c in idx_type)
nodes_file = os.path.join(output_path, f"{safe_cid}_{safe_idx_type}_nodes.jsonl")
metadata_file = os.path.join(output_path, f"{safe_cid}_{safe_idx_type}_metadata.json")
# Save corpus as JSONL
corpus.to_jsonl(nodes_file, indent=0)
logger.info(f"Saved {len(corpus.chunks)} chunks to {nodes_file}")
# Save metadata as JSON
with open(metadata_file, "w", encoding="utf-8") as f:
json.dump(metadata.model_dump(), f, indent=2, ensure_ascii=False)
logger.info(f"Saved metadata to {metadata_file}")
else:
# Database saving
index_data = {
"corpus_id": cid,
"content": corpus.model_dump(),
"date": str(datetime.now()),
"metadata": metadata.model_dump()
}
self.storage_handler.save_index(index_data, table=table)
logger.info(f"Saved {idx_type} index with {len(corpus.chunks)} chunks for corpus {cid} to database table {table}")
except Exception as e:
logger.error(f"Failed to save indices for corpus {corpus_id or 'all'}: {str(e)}")
raise
def load(self, source: Optional[str] = None, corpus_id: Optional[str] = None,
index_type: Optional[str] = None, table: Optional[str] = None) -> None:
"""Load indices from files or database.
Reconstructs indices and retrievers from JSONL/JSON files or SQLite database records.
Validates the embedding model name and dimension before reinitializing the embedding model.
Args:
source (Optional[str]): Directory containing JSONL/JSON files. If None, loads from database.
corpus_id (Optional[str]): Specific corpus to load. If None, loads all corpora.
index_type (Optional[str]): Specific index type to load. If None, loads all indices.
table (Optional[str]): Database table name for index data. Defaults to 'indexing' if None.
Returns:
The Sequence with id of loaded chunk.
Raises:
Exception: If loading fails due to file or database errors, invalid data, or unsupported embedding model/dimension.
Warning:
Try to call this function may cause some Bugs, when you load the nodes from file or database storage systems at twice.
Because All the indexing share the same storage backend from storageHandler.
For example:
The vector database (.e.g Faiss) can insert again, even thougt there is a same node.
"""
try:
table = table or "indexing"
config_dimension = self.storage_handler.storageConfig.vectorConfig.dimensions
loaded_chunk_ids: List[str] = []
if source:
# File-based loading
if not os.path.exists(source):
logger.error(f"Source directory {source} does not exist")
raise FileNotFoundError(f"Source directory {source} does not exist")
for file_name in os.listdir(source):
if not file_name.endswith("_metadata.json"):
continue
parts = file_name.split("_")
if len(parts) < 3:
logger.warning(f"Skipping invalid metadata file: {file_name}")
continue
cid = "_".join(parts[:-2])
idx_type = parts[-2]
if (corpus_id and corpus_id != cid) or (index_type and index_type != idx_type):
continue
metadata_file = os.path.join(source, file_name)
nodes_file = os.path.join(source, f"{cid}_{idx_type}_nodes.jsonl")
# Load metadata
with open(metadata_file, "r", encoding="utf-8") as f:
metadata = IndexMetadata.model_validate(json.load(f))
# Validate embedding model
if not self.embed_model.validate_model(self.config.embedding.provider, metadata.embedding_model_name):
raise ValueError(
f"Embedding model '{metadata.embedding_model_name}' is not supported by provider '{self.config.embedding.provider}'. "
f"Supported models: {EmbeddingProvider.SUPPORTED_MODELS.get(self.config.embedding.provider, [])}"
)
# Validate dimension
if metadata.dimension != config_dimension:
raise ValueError(
f"Embedding dimension {metadata.dimension} in metadata does not match configured dimension {config_dimension}."
)
# Load corpus
if not os.path.exists(nodes_file):
logger.warning(f"Nodes file {nodes_file} not found for metadata {metadata_file}")
continue
corpus = Corpus.from_jsonl(nodes_file, corpus_id=cid)
# Reinitialize embedding model if needed
if metadata.embedding_model_name != self.config.embedding.model_name:
logger.info(f"Reinitializing embedding model to {metadata.embedding_model_name}")
self.embed_model = self.embedding_factory.create(
provider=self.config.embedding.provider,
model_config=self.config.embedding.model_dump(exclude_unset=True)
)
# Load index
chunk_ids = self._load_index(corpus, cid, idx_type)
loaded_chunk_ids.extend(chunk_ids)
logger.info(f"Loaded {idx_type} index with {len(corpus.chunks)} chunks for corpus {cid} from {nodes_file}")
else:
# Database loading
records = self.storage_handler.load(tables=[table]).get(table, [])
if not records:
logger.warning(f"No records found in table {table}")
return
for record in records:
parsed = self.storage_handler.parse_result(record, IndexStore)
cid = parsed["corpus_id"]
idx_type = parsed["metadata"]["index_type"]
if (corpus_id and corpus_id != cid) or (index_type and index_type != idx_type):
continue
# Reconstruct corpus
chunks = []
for chunk_data in parsed["content"]["chunks"]:
metadata = ChunkMetadata.model_validate(chunk_data["metadata"])
if self.config.modality == "multimodal":
# Create ImageChunk for multimodal mode
chunk = ImageChunk(
chunk_id=chunk_data["chunk_id"],
image_path=chunk_data["image_path"],
image_mimetype=chunk_data.get("image_mimetype"),
metadata=metadata,
embedding=chunk_data["embedding"],
excluded_embed_metadata_keys=chunk_data["excluded_embed_metadata_keys"],
excluded_llm_metadata_keys=chunk_data["excluded_llm_metadata_keys"],
relationships={k: RelatedNodeInfo(**v) for k, v in chunk_data["relationships"].items()}
)
else:
# Create TextChunk for text mode
chunk = TextChunk(
chunk_id=chunk_data["chunk_id"],
text=chunk_data["text"],
metadata=metadata,
embedding=chunk_data["embedding"],
start_char_idx=chunk_data["start_char_idx"],
end_char_idx=chunk_data["end_char_idx"],
excluded_embed_metadata_keys=chunk_data["excluded_embed_metadata_keys"],
excluded_llm_metadata_keys=chunk_data["excluded_llm_metadata_keys"],
relationships={k: RelatedNodeInfo(**v) for k, v in chunk_data["relationships"].items()}
)
chunks.append(chunk)
corpus = Corpus(
chunks=chunks,
corpus_id=cid,
metadata=IndexMetadata.model_validate(parsed["metadata"])
)
# Validate embedding model
metadata = IndexMetadata.model_validate(parsed["metadata"])
if not self.embed_model.validate_model(self.config.embedding.provider, metadata.embedding_model_name):
raise ValueError(
f"Embedding model '{metadata.embedding_model_name}' is not supported by provider '{self.config.embedding.provider}'. "
f"Supported models: {EmbeddingProvider.SUPPORTED_MODELS.get(self.config.embedding.provider, [])}"
)
# Validate dimension
if metadata.dimension != config_dimension:
raise ValueError(
f"Embedding dimension {metadata.dimension} in metadata does not match configured dimension {config_dimension}."
)
# Reinitialize embedding model if needed
if metadata.embedding_model_name != self.config.embedding.model_name:
logger.info(f"Reinitializing embedding model to {metadata.embedding_model_name}")
self.embed_model = self.embedding_factory.create(
provider=self.config.embedding.provider,
model_config=self.config.embedding.model_dump(exclude_unset=True)
)
# Load index
chunk_ids = self._load_index(corpus, cid, idx_type)
loaded_chunk_ids.extend(chunk_ids)
logger.info(f"Loaded {idx_type} index with {len(corpus.chunks)} chunks for corpus {cid} from database table {table}")
return loaded_chunk_ids
except Exception as e:
logger.error(f"Failed to load indices: {str(e)}")
raise
def _load_index(self, corpus: Corpus, corpus_id: str, index_type: str) -> Sequence[str]:
"""Helper method to load an index and its retriever."""
try:
if corpus_id not in self.indices:
self.indices[corpus_id] = {}
self.retrievers[corpus_id] = {}
if index_type not in self.indices[corpus_id]:
index = self.index_factory.create(
index_type=index_type,
embed_model=self.embed_model.get_embedding_model(),
storage_handler=self.storage_handler,
index_config=self.config.index.model_dump(exclude_unset=True) if self.config.index else {},
llm=self.llm
)
self.indices[corpus_id][index_type] = index
retriever_type = RetrieverType.GRAPH if index_type == IndexType.GRAPH else RetrieverType.VECTOR
self.retrievers[corpus_id][index_type] = self.retriever_factory.create(
retriever_type=retriever_type,
llm=self.llm,
index=index.get_index(),
graph_store=index.get_index().storage_context.graph_store,
embed_model=self.embed_model.get_embedding_model(),
query=Query(query_str="", top_k=self.config.retrieval.top_k if self.config.retrieval else 5),
storage_handler=self.storage_handler
)
nodes = corpus.to_llama_nodes()
for node in nodes:
node.metadata.update({"corpus_id": corpus_id, "index_type": index_type})
chunk_ids = self.indices[corpus_id][index_type].load(nodes)
logger.info(f"Inserted {len(nodes)} nodes into {index_type} index for corpus {corpus_id}")
return chunk_ids
except Exception as e:
logger.error(f"Failed to load index for corpus {corpus_id}, index_type {index_type}: {str(e)}")
raise
async def aget(self, corpus_id: str, index_type: str, node_ids: List[str]) -> List[Union[TextChunk, ImageChunk]]:
"""Retrieve chunks by node_ids from the index."""
try:
chunks = await self.indices[corpus_id][index_type].get(node_ids=node_ids)
logger.info(f"Retrieved {len(chunks)} chunks for node_ids: {node_ids}")
return chunks
except Exception as e:
logger.error(f"Failed to get chunks: {str(e)}")
return []
async def query_async(self, query: Union[str, Query], corpus_id: Optional[str] = None,
query_transforms: Optional[List] = None) -> RagResult:
"""Execute a query across indices and return processed results asynchronously.
Performs query preprocessing, asynchronous retrieval, and post-processing.
Args:
query (Union[str, Query]): Query string or Query object.
corpus_id (Optional[str]): Specific corpus to query. If None, queries all corpora.
query_transforms (Optional[List]): Query Transforms is used to augment query in pre-processing.
Returns:
RagResult: Retrieved chunks with scores and metadata.
Raises:
Exception: If query processing fails.
"""
try:
if isinstance(query, str):
query = Query(query_str=query, top_k=self.config.retrieval.top_k)
if not self.indices or (corpus_id and corpus_id not in self.indices):
logger.warning(f"No indices found for corpus {corpus_id or 'any'}")
return RagResult(corpus=Corpus(chunks=[]), scores=[], metadata={"query": query.query_str})
# Pre-Processing
if query_transforms and query_transforms is not None:
for t in query_transforms:
query = t(query)
results = []
target_corpora = [corpus_id] if corpus_id else self.indices.keys()
# Create all retrieval tasks
tasks = []
for cid in target_corpora:
for idx_type, retriever in self.retrievers[cid].items():
if query.metadata_filters and query.metadata_filters.get("index_type") and \
query.metadata_filters["index_type"] != idx_type:
continue
task = retriever.aretrieve(
Query(
query_str=query.query_str,
top_k=query.top_k or self.config.retrieval.top_k,
similarity_cutoff=query.similarity_cutoff,
keyword_filters=query.keyword_filters,
metadata_filters=query.metadata_filters
)
)
tasks.append((task, cid, idx_type))
# Run all retrievals concurrently
retrieval_tasks = [task for task, _, _ in tasks]
retrieval_results = await asyncio.gather(*retrieval_tasks, return_exceptions=True)
# Process results
for (_, cid, idx_type), result in zip(tasks, retrieval_results):
if isinstance(result, Exception):
logger.error(f"Retrieval failed for {idx_type} in corpus {cid}: {str(result)}")
else:
results.append(result)
logger.info(f"Retrieved {len(result.corpus.chunks)} chunks from {idx_type} retriever for corpus {cid}")
if not results:
return RagResult(corpus=Corpus(chunks=[]), scores=[], metadata={"query": query.query_str})
# Check the 'similarity_cutoff' and 'keyword_filters' in query
query.similarity_cutoff = self.config.retrieval.similarity_cutoff if query.similarity_cutoff is None else query.similarity_cutoff
query.keyword_filters = self.config.retrieval.keyword_filters if query.keyword_filters is None else query.keyword_filters
postprocessor = self.postprocessor_factory.create(
self.config.retrieval.postprocessor_type,
query=query
)
final_result = postprocessor.postprocess(query, results)
if query.metadata_filters:
final_result.corpus.chunks = [
chunk for chunk in final_result.corpus.chunks
if all(chunk.metadata.model_dump().get(k) == v for k, v in query.metadata_filters.items())
]
final_result.scores = [chunk.metadata.similarity_score for chunk in final_result.corpus.chunks]
logger.info(f"Applied metadata filters, retained {len(final_result.corpus.chunks)} chunks")
logger.info(f"Query returned {len(final_result.corpus.chunks)} chunks after post-processing")
return final_result
except Exception as e:
logger.error(f"Query failed: {str(e)}")
raise
def query(self, query: Union[str, Query], corpus_id: Optional[str] = None,
query_transforms: Optional[List] = None) -> RagResult:
"""Synchronous wrapper for the async query method."""
return asyncio.run(self.query_async(query, corpus_id, query_transforms)) |