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 | class OCIGenAI(LLM):
"""OCI大型语言模型。
要进行身份验证，OCI客户端使用https://docs.oracle.com/en-us/iaas/Content/API/Concepts/sdk_authentication_methods.htm中描述的方法。
身份验证方法通过auth_type传递，应为以下之一：
API_KEY（默认），SECURITY_TOKEN，INSTANCE_PRINCIPAL，RESOURCE_PRINCIPAL
确保您具有访问OCI生成AI服务所需的策略（配置文件/角色）。
如果使用特定的配置文件配置文件，则必须通过auth_profile传递配置文件的名称（来自~/.oci/config）。
要使用，必须提供区段ID
以及端点URL和模型ID
作为构造函数的命名参数。
例子：
.. 代码块:: python
from llama_index.llms.oci_genai import OCIGenAI
llm = OCIGenAI(
model="MY_MODEL_ID",
service_endpoint="https://inference.generativeai.us-chicago-1.oci.oraclecloud.com",
compartment_id="MY_OCID"
)
"""
model: str = Field(description="Id of the OCI Generative AI model to use.")
temperature: float = Field(description="The temperature to use for sampling.")
max_tokens: int = Field(description="The maximum number of tokens to generate.")
context_size: int = Field("The maximum number of tokens available for input.")
service_endpoint: str = Field(
default=None,
description="service endpoint url.",
)
compartment_id: str = Field(
default=None,
description="OCID of compartment.",
)
auth_type: Optional[str] = Field(
description="Authentication type, can be: API_KEY, SECURITY_TOKEN, INSTANCE_PRINCIPAL, RESOURCE_PRINCIPAL. If not specified, API_KEY will be used",
default="API_KEY",
)
auth_profile: Optional[str] = Field(
description="The name of the profile in ~/.oci/config. If not specified , DEFAULT will be used",
default="DEFAULT",
)
additional_kwargs: Dict[str, Any] = Field(
default_factory=dict,
description="Additional kwargs for the OCI Generative AI request.",
)
_client: Any = PrivateAttr()
_provider: str = PrivateAttr()
_serving_mode: str = PrivateAttr()
_completion_generator: str = PrivateAttr()
_chat_generator: str = PrivateAttr()
def __init__(
self,
model: str,
temperature: Optional[float] = DEFAULT_TEMPERATURE,
max_tokens: Optional[int] = 512,
context_size: Optional[int] = None,
service_endpoint: str = None,
compartment_id: str = None,
auth_type: Optional[str] = "API_KEY",
auth_profile: Optional[str] = "DEFAULT",
client: Optional[Any] = None,
provider: Optional[str] = None,
additional_kwargs: Optional[Dict[str, Any]] = None,
callback_manager: Optional[CallbackManager] = None,
system_prompt: Optional[str] = None,
messages_to_prompt: Optional[Callable[[Sequence[ChatMessage]], str]] = None,
completion_to_prompt: Optional[Callable[[str], str]] = None,
pydantic_program_mode: PydanticProgramMode = PydanticProgramMode.DEFAULT,
output_parser: Optional[BaseOutputParser] = None,
) -> None:
"""初始化OCIGenAI类。
Args:
model (str): 用于生成嵌入的模型的ID，例如 "meta.llama-2-70b-chat"。
temperature (Optional[float]): 用于采样的温度。默认值在lama_index.core.constants.DEFAULT_TEMPERATURE中指定。
max_tokens (Optional[int]): 要生成的最大标记数。默认值为512。
context_size (Optional[int]): 输入可用的最大标记数。如果未指定，将使用模型的默认上下文大小。
service_endpoint (str): 服务端点URL，例如 "https://inference.generativeai.us-chicago-1.oci.oraclecloud.com"。
compartment_id (str): 专用区的OCID。
auth_type (Optional[str]): 认证类型，可以是: API_KEY (默认值), SECURITY_TOKEN, INSTANCEAL, RESOURCE_PRINCIPAL。
如果未指定，将使用API_KEY。
auth_profile (Optional[str]): ~/.oci/config中配置文件的配置文件名。如果未指定，将使用DEFAULT。
client (Optional[Any]): 可选的OCI客户端对象。如果未提供，将使用提供的服务端点和认证方法创建客户端。
provider (Optional[str]): 模型的提供商名称。如果未指定，将从模型名称中推导出提供商。
additional_kwargs (Optional[Dict[str, Any]]): 用于LLM的其他kwargs。
"""
self._client = client or create_client(
auth_type, auth_profile, service_endpoint
)
self._provider = get_provider(model, provider)
self._serving_mode = get_serving_mode(model)
self._completion_generator = get_completion_generator()
self._chat_generator = get_chat_generator()
context_size = get_context_size(model, context_size)
additional_kwargs = additional_kwargs or {}
callback_manager = callback_manager or CallbackManager([])
super().__init__(
model=model,
temperature=temperature,
max_tokens=max_tokens,
context_size=context_size,
service_endpoint=service_endpoint,
compartment_id=compartment_id,
auth_type=auth_type,
auth_profile=auth_profile,
additional_kwargs=additional_kwargs,
callback_manager=callback_manager,
system_prompt=system_prompt,
messages_to_prompt=messages_to_prompt,
completion_to_prompt=completion_to_prompt,
pydantic_program_mode=pydantic_program_mode,
output_parser=output_parser,
)
@classmethod
def class_name(cls) -> str:
"""获取类名。"""
return "OCIGenAI_LLM"
@property
def metadata(self) -> LLMMetadata:
return LLMMetadata(
context_window=self.context_size,
num_output=self.max_tokens,
is_chat_model=self.model in CHAT_MODELS,
model_name=self.model,
)
@property
def _model_kwargs(self) -> Dict[str, Any]:
base_kwargs = {
"temperature": self.temperature,
"max_tokens": self.max_tokens,
}
return {
**base_kwargs,
**self.additional_kwargs,
}
def _get_all_kwargs(self, **kwargs: Any) -> Dict[str, Any]:
return {
**self._model_kwargs,
**kwargs,
}
@llm_completion_callback()
def complete(
self, prompt: str, formatted: bool = False, **kwargs: Any
) -> CompletionResponse:
inference_params = self._get_all_kwargs(**kwargs)
inference_params["is_stream"] = False
inference_params["prompt"] = prompt
request = self._completion_generator(
compartment_id=self.compartment_id,
serving_mode=self._serving_mode,
inference_request=self._provider.oci_completion_request(**inference_params),
)
response = self._client.generate_text(request)
return CompletionResponse(
text=self._provider.completion_response_to_text(response),
raw=response.__dict__,
)
@llm_completion_callback()
def stream_complete(
self, prompt: str, formatted: bool = False, **kwargs: Any
) -> CompletionResponseGen:
inference_params = self._get_all_kwargs(**kwargs)
inference_params["is_stream"] = True
inference_params["prompt"] = prompt
request = self._completion_generator(
compartment_id=self.compartment_id,
serving_mode=self._serving_mode,
inference_request=self._provider.oci_completion_request(**inference_params),
)
response = self._client.generate_text(request)
def gen() -> CompletionResponseGen:
content = ""
for event in response.data.events():
content_delta = self._provider.completion_stream_to_text(
json.loads(event.data)
)
content += content_delta
yield CompletionResponse(
text=content, delta=content_delta, raw=event.__dict__
)
return gen()
@llm_chat_callback()
def chat(self, messages: Sequence[ChatMessage], **kwargs: Any) -> ChatResponse:
oci_params = self._provider.messages_to_oci_params(messages)
oci_params["is_stream"] = False
all_kwargs = self._get_all_kwargs(**kwargs)
chat_params = {**all_kwargs, **oci_params}
request = self._chat_generator(
compartment_id=self.compartment_id,
serving_mode=self._serving_mode,
chat_request=self._provider.oci_chat_request(**chat_params),
)
response = self._client.chat(request)
return ChatResponse(
message=ChatMessage(
role=MessageRole.ASSISTANT,
content=self._provider.chat_response_to_text(response),
),
raw=response.__dict__,
)
def stream_chat(
self, messages: Sequence[ChatMessage], **kwargs: Any
) -> ChatResponseGen:
oci_params = self._provider.messages_to_oci_params(messages)
oci_params["is_stream"] = True
all_kwargs = self._get_all_kwargs(**kwargs)
chat_params = {**all_kwargs, **oci_params}
request = self._chat_generator(
compartment_id=self.compartment_id,
serving_mode=self._serving_mode,
chat_request=self._provider.oci_chat_request(**chat_params),
)
response = self._client.chat(request)
def gen() -> ChatResponseGen:
content = ""
for event in response.data.events():
content_delta = self._provider.chat_stream_to_text(
json.loads(event.data)
)
content += content_delta
yield ChatResponse(
message=ChatMessage(role=MessageRole.ASSISTANT, content=content),
delta=content_delta,
raw=event.__dict__,
)
return gen()
async def acomplete(
self, prompt: str, formatted: bool = False, **kwargs: Any
) -> CompletionResponse:
# do synchronous complete for now
return self.complete(prompt, formatted=formatted, **kwargs)
async def achat(
self, messages: Sequence[ChatMessage], **kwargs: Any
) -> ChatResponse:
# do synchronous chat for now
return self.chat(messages, **kwargs)
async def astream_chat(
self, messages: Sequence[ChatMessage], **kwargs: Any
) -> ChatResponseAsyncGen:
# do synchronous stream chat for now
return self.stream_chat(messages, **kwargs)
async def astream_complete(
self, prompt: str, formatted: bool = False, **kwargs: Any
) -> CompletionResponseAsyncGen:
# do synchronous stream complete for now
return self.stream_complete(prompt, formatted, **kwargs)
|