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 | class CohereEmbedding(BaseEmbedding):
"""CohereEmbedding uses the Cohere API to generate embeddings for text."""
# Instance variables initialized via Pydantic's mechanism
cohere_client: Any = Field(description="CohereAI client")
truncate: str = Field(description="Truncation type - START/ END/ NONE")
input_type: Optional[str] = Field(
description="Model Input type. If not provided, search_document and search_query are used when needed."
)
def __init__(
self,
cohere_api_key: Optional[str] = None,
model_name: str = "embed-english-v3.0",
truncate: str = "END",
input_type: Optional[str] = None,
embed_batch_size: int = DEFAULT_EMBED_BATCH_SIZE,
callback_manager: Optional[CallbackManager] = None,
):
"""
A class representation for generating embeddings using the Cohere API.
Args:
cohere_client (Any): An instance of the Cohere client, which is used to communicate with the Cohere API.
truncate (str): A string indicating the truncation strategy to be applied to input text. Possible values
are 'START', 'END', or 'NONE'.
input_type (Optional[str]): An optional string that specifies the type of input provided to the model.
This is model-dependent and could be one of the following: 'search_query',
'search_document', 'classification', or 'clustering'.
model_name (str): The name of the model to be used for generating embeddings. The class ensures that
this model is supported and that the input type provided is compatible with the model.
"""
# Validate model_name and input_type
if model_name not in VALID_MODEL_INPUT_TYPES:
raise ValueError(f"{model_name} is not a valid model name")
if input_type not in VALID_MODEL_INPUT_TYPES[model_name]:
raise ValueError(
f"{input_type} is not a valid input type for the provided model."
)
if truncate not in VALID_TRUNCATE_OPTIONS:
raise ValueError(f"truncate must be one of {VALID_TRUNCATE_OPTIONS}")
super().__init__(
cohere_client=cohere.Client(cohere_api_key, client_name="llama_index"),
cohere_api_key=cohere_api_key,
model_name=model_name,
input_type=input_type,
truncate=truncate,
embed_batch_size=embed_batch_size,
callback_manager=callback_manager,
)
@classmethod
def class_name(cls) -> str:
return "CohereEmbedding"
def _embed(self, texts: List[str], input_type: str) -> List[List[float]]:
"""Embed sentences using Cohere."""
if self.model_name in [
CAMN.ENGLISH_V3,
CAMN.ENGLISH_LIGHT_V3,
CAMN.MULTILINGUAL_V3,
CAMN.MULTILINGUAL_LIGHT_V3,
]:
result = self.cohere_client.embed(
texts=texts,
input_type=self.input_type or input_type,
model=self.model_name,
truncate=self.truncate,
).embeddings
else:
result = self.cohere_client.embed(
texts=texts, model=self.model_name, truncate=self.truncate
).embeddings
return [list(map(float, e)) for e in result]
def _get_query_embedding(self, query: str) -> List[float]:
"""Get query embedding. For query embeddings, input_type='search_query'."""
return self._embed([query], input_type="search_query")[0]
async def _aget_query_embedding(self, query: str) -> List[float]:
"""Get query embedding async. For query embeddings, input_type='search_query'."""
return self._get_query_embedding(query)
def _get_text_embedding(self, text: str) -> List[float]:
"""Get text embedding."""
return self._embed([text], input_type="search_document")[0]
async def _aget_text_embedding(self, text: str) -> List[float]:
"""Get text embedding async."""
return self._get_text_embedding(text)
def _get_text_embeddings(self, texts: List[str]) -> List[List[float]]:
"""Get text embeddings."""
return self._embed(texts, input_type="search_document")
|