Spaces:
Build error
Build error
File size: 9,930 Bytes
d660b02 |
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 |
import uuid
from abc import ABC
from typing import Any, Callable, Dict, Generic, Type, TypeVar
from uuid import UUID
import numpy as np
from loguru import logger
from pydantic import UUID4, BaseModel, Field
from qdrant_client.http import exceptions
from qdrant_client.http.models import Distance, VectorParams
from qdrant_client.models import CollectionInfo, PointStruct, Record
from llm_engineering.application.networks.embeddings import EmbeddingModelSingleton
from llm_engineering.domain.exceptions import ImproperlyConfigured
from llm_engineering.domain.types import DataCategory
from llm_engineering.infrastructure.db.qdrant import connection
T = TypeVar("T", bound="VectorBaseDocument")
class VectorBaseDocument(BaseModel, Generic[T], ABC):
id: UUID4 = Field(default_factory=uuid.uuid4)
def __eq__(self, value: object) -> bool:
if not isinstance(value, self.__class__):
return False
return self.id == value.id
def __hash__(self) -> int:
return hash(self.id)
@classmethod
def from_record(cls: Type[T], point: Record) -> T:
_id = UUID(point.id, version=4)
payload = point.payload or {}
attributes = {
"id": _id,
**payload,
}
if cls._has_class_attribute("embedding"):
attributes["embedding"] = point.vector or None
return cls(**attributes)
def to_point(self: T, **kwargs) -> PointStruct:
exclude_unset = kwargs.pop("exclude_unset", False)
by_alias = kwargs.pop("by_alias", True)
payload = self.model_dump(exclude_unset=exclude_unset, by_alias=by_alias, **kwargs)
_id = str(payload.pop("id"))
vector = payload.pop("embedding", {})
if vector and isinstance(vector, np.ndarray):
vector = vector.tolist()
return PointStruct(id=_id, vector=vector, payload=payload)
def model_dump(self: T, **kwargs) -> dict:
dict_ = super().model_dump(**kwargs)
dict_ = self._uuid_to_str(dict_)
return dict_
def _uuid_to_str(self, item: Any) -> Any:
if isinstance(item, dict):
for key, value in item.items():
if isinstance(value, UUID):
item[key] = str(value)
elif isinstance(value, list):
item[key] = [self._uuid_to_str(v) for v in value]
elif isinstance(value, dict):
item[key] = {k: self._uuid_to_str(v) for k, v in value.items()}
return item
@classmethod
def bulk_insert(cls: Type[T], documents: list["VectorBaseDocument"]) -> bool:
try:
cls._bulk_insert(documents)
except exceptions.UnexpectedResponse:
logger.info(
f"Collection '{cls.get_collection_name()}' does not exist. Trying to create the collection and reinsert the documents."
)
cls.create_collection()
try:
cls._bulk_insert(documents)
except exceptions.UnexpectedResponse:
logger.error(f"Failed to insert documents in '{cls.get_collection_name()}'.")
return False
return True
@classmethod
def _bulk_insert(cls: Type[T], documents: list["VectorBaseDocument"]) -> None:
points = [doc.to_point() for doc in documents]
connection.upsert(collection_name=cls.get_collection_name(), points=points)
@classmethod
def bulk_find(cls: Type[T], limit: int = 10, **kwargs) -> tuple[list[T], UUID | None]:
try:
documents, next_offset = cls._bulk_find(limit=limit, **kwargs)
except exceptions.UnexpectedResponse:
logger.error(f"Failed to search documents in '{cls.get_collection_name()}'.")
documents, next_offset = [], None
return documents, next_offset
@classmethod
def _bulk_find(cls: Type[T], limit: int = 10, **kwargs) -> tuple[list[T], UUID | None]:
collection_name = cls.get_collection_name()
offset = kwargs.pop("offset", None)
offset = str(offset) if offset else None
records, next_offset = connection.scroll(
collection_name=collection_name,
limit=limit,
with_payload=kwargs.pop("with_payload", True),
with_vectors=kwargs.pop("with_vectors", False),
offset=offset,
**kwargs,
)
documents = [cls.from_record(record) for record in records]
if next_offset is not None:
next_offset = UUID(next_offset, version=4)
return documents, next_offset
@classmethod
def search(cls: Type[T], query_vector: list, limit: int = 10, **kwargs) -> list[T]:
try:
documents = cls._search(query_vector=query_vector, limit=limit, **kwargs)
except exceptions.UnexpectedResponse:
logger.error(f"Failed to search documents in '{cls.get_collection_name()}'.")
documents = []
return documents
@classmethod
def _search(cls: Type[T], query_vector: list, limit: int = 10, **kwargs) -> list[T]:
collection_name = cls.get_collection_name()
records = connection.search(
collection_name=collection_name,
query_vector=query_vector,
limit=limit,
with_payload=kwargs.pop("with_payload", True),
with_vectors=kwargs.pop("with_vectors", False),
**kwargs,
)
documents = [cls.from_record(record) for record in records]
return documents
@classmethod
def get_or_create_collection(cls: Type[T]) -> CollectionInfo:
collection_name = cls.get_collection_name()
try:
return connection.get_collection(collection_name=collection_name)
except exceptions.UnexpectedResponse:
use_vector_index = cls.get_use_vector_index()
collection_created = cls._create_collection(
collection_name=collection_name, use_vector_index=use_vector_index
)
if collection_created is False:
raise RuntimeError(f"Couldn't create collection {collection_name}") from None
return connection.get_collection(collection_name=collection_name)
@classmethod
def create_collection(cls: Type[T]) -> bool:
collection_name = cls.get_collection_name()
use_vector_index = cls.get_use_vector_index()
return cls._create_collection(collection_name=collection_name, use_vector_index=use_vector_index)
@classmethod
def _create_collection(cls, collection_name: str, use_vector_index: bool = True) -> bool:
if use_vector_index is True:
vectors_config = VectorParams(size=EmbeddingModelSingleton().embedding_size, distance=Distance.COSINE)
else:
vectors_config = {}
return connection.create_collection(collection_name=collection_name, vectors_config=vectors_config)
@classmethod
def get_category(cls: Type[T]) -> DataCategory:
if not hasattr(cls, "Config") or not hasattr(cls.Config, "category"):
raise ImproperlyConfigured(
"The class should define a Config class with"
"the 'category' property that reflects the collection's data category."
)
return cls.Config.category
@classmethod
def get_collection_name(cls: Type[T]) -> str:
if not hasattr(cls, "Config") or not hasattr(cls.Config, "name"):
raise ImproperlyConfigured(
"The class should define a Config class with" "the 'name' property that reflects the collection's name."
)
return cls.Config.name
@classmethod
def get_use_vector_index(cls: Type[T]) -> bool:
if not hasattr(cls, "Config") or not hasattr(cls.Config, "use_vector_index"):
return True
return cls.Config.use_vector_index
@classmethod
def group_by_class(
cls: Type["VectorBaseDocument"], documents: list["VectorBaseDocument"]
) -> Dict["VectorBaseDocument", list["VectorBaseDocument"]]:
return cls._group_by(documents, selector=lambda doc: doc.__class__)
@classmethod
def group_by_category(cls: Type[T], documents: list[T]) -> Dict[DataCategory, list[T]]:
return cls._group_by(documents, selector=lambda doc: doc.get_category())
@classmethod
def _group_by(cls: Type[T], documents: list[T], selector: Callable[[T], Any]) -> Dict[Any, list[T]]:
grouped = {}
for doc in documents:
key = selector(doc)
if key not in grouped:
grouped[key] = []
grouped[key].append(doc)
return grouped
@classmethod
def collection_name_to_class(cls: Type["VectorBaseDocument"], collection_name: str) -> type["VectorBaseDocument"]:
for subclass in cls.__subclasses__():
try:
if subclass.get_collection_name() == collection_name:
return subclass
except ImproperlyConfigured:
pass
try:
return subclass.collection_name_to_class(collection_name)
except ValueError:
continue
raise ValueError(f"No subclass found for collection name: {collection_name}")
@classmethod
def _has_class_attribute(cls: Type[T], attribute_name: str) -> bool:
if attribute_name in cls.__annotations__:
return True
for base in cls.__bases__:
if hasattr(base, "_has_class_attribute") and base._has_class_attribute(attribute_name):
return True
return False
|