File size: 9,270 Bytes
f47ddc5 |
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 |
import requests
from typing import List, Optional, Sequence, Any, AsyncGenerator
from llama_index.legacy.llms import LLM, LLMMetadata
from llama_index.legacy.llms.types import ChatMessage
from llama_index.core.llms.callbacks import llm_chat_callback, llm_completion_callback
from llama_index.core.base.llms.types import ChatMessage, ChatResponse, CompletionResponseAsyncGen, ChatResponseAsyncGen, MessageRole, CompletionResponse, CompletionResponseGen
from llama_index.core import SimpleDirectoryReader, VectorStoreIndex
class Kognie(LLM):
"""
A custom LLM that calls a FastAPI server at /text endpoint.
"""
base_url: str = 'http://api2.kognie.com'
api_key: str
model: str
response_format: str = 'url'
@property
def metadata(self) -> LLMMetadata:
# Provide info about your model to LlamaIndex (adjust as needed)
return LLMMetadata(
model_name=self.model
)
def _generate_text(
self,
prompt: str,
model: Optional[str] = None,
**kwargs
) -> str:
"""
The single-turn text generation method.
LlamaIndex calls `_generate_text` internally whenever it needs a completion.
"""
# Decide on mode and model to use, falling back to defaults
selected_model = model if model else self.model
endpoint = f"{self.base_url}/text"
# Prepare GET request parameters
params = {
"question": prompt,
"model": selected_model
}
# Prepare HTTP headers
headers = {
"X-KEY": self.api_key
}
try:
# Send request
response = requests.get(endpoint, params=params, headers=headers)
response.raise_for_status()
except requests.HTTPError as exc:
raise ValueError(f"FastAPI /text endpoint error: {exc}") from exc
data = response.json()
text_output = data.get("response", "")
return text_output
def _generate_image(
self,
prompt: str,
model: str,
response_format: str,
**kwargs
) -> str:
"""
The single-turn text generation method.
LlamaIndex calls `_generate_text` internally whenever it needs a completion.
"""
# Decide on mode and model to use, falling back to defaults
selected_model = model if model else self.model
endpoint = f"{self.base_url}/image"
# Prepare GET request parameters
params = {
"question": prompt,
"model": selected_model,
"response_format": response_format
}
# Prepare HTTP headers
headers = {
"X-KEY": self.api_key
}
try:
# Send request
response = requests.get(endpoint, params=params, headers=headers)
response.raise_for_status()
except requests.HTTPError as exc:
raise ValueError(f"FastAPI /text endpoint error: {exc}") from exc
# Parse JSON
data = response.json()
text_output = data.get("response", "")
return text_output
def generate_img(
self,
prompt: str,
model: str,
response_format: str,
) -> ChatMessage:
img_output = self._generate_image(
prompt=prompt,
model=model,
response_format=response_format
)
return ChatMessage(role="assistant", content=img_output)
# (Optional) Multi-turn chat approach
def chat(
self,
messages: List[ChatMessage],
model: Optional[str] = None,
**kwargs
) -> ChatMessage:
"""
If you want to handle multi-turn chat style conversation, override this method.
In LlamaIndex, some indices or chat modules might call `chat(messages=...)`.
"""
# Merge messages into a single prompt
# e.g. if you want to pass a conversation log:
conversation_log = ""
for m in messages:
role = m.role # "system", "user", or "assistant"
content = m.content
if role == "user":
conversation_log += f"User: {content}\n"
else:
conversation_log += f"{role.capitalize()}: {content}\n"
# Now just call your single-turn generation on the entire conversation log
# This is simplistic; you can implement more advanced chat logic if needed
text_output = self._generate_text(
prompt=conversation_log,
model=model,
**kwargs
)
return ChatMessage(role="assistant", content=text_output)
@llm_chat_callback()
def messages_to_prompt(messages):
prompt = ""
for message in messages:
if message.role == MessageRole.SYSTEM:
prompt += f"<|system|>\n(message.content)</s>\n"
elif message.role == MessageRole.USER:
prompt += f"<|user|>\n{message.content}</s>\n"
elif message.role == MessageRole.ASSISTANT:
prompt += f"<|assistant|>\n{message.content}</s>\n"
# Ensure the prompt starts with a system message
if not prompt.startswith("<|system|>\n"):
prompt = "<|system|>\n</s>\n" + prompt
# Add a final assistant prompt
prompt += "<|assistant|>\n"
return prompt
async def stream_chat(self, messages: Sequence[ChatMessage], **kwargs: Any) -> AsyncGenerator[ChatResponse, None]:
# Assume `astream_complete` is an async method that yields CompletionResponse objects
async for completion_response in self.astream_complete(self.messages_to_prompt(messages), **kwargs):
# Here, you manually convert each CompletionResponse to a ChatResponse
chat_response = self.convert_completion_to_chat(
completion_response)
yield chat_response
async def astream_complete(self, prompt: str, **kwargs: Any) -> AsyncGenerator[CompletionResponse, None]:
# Implement your logic to asynchronously stream completion responses
pass
def convert_completion_to_chat(self, completion_response: CompletionResponse) -> ChatResponse:
# Implement your conversion logic here
# For simplicity, we're directly using the completion text as the chat content
return ChatResponse(message=ChatMessage(role="assistant", content=completion_response.text))
@llm_chat_callback()
async def achat(
self,
messages: Sequence[ChatMessage],
**kwargs: Any,
) -> ChatResponse:
return self.chat(messages, **kwargs)
@llm_chat_callback()
async def astream_chat(
self,
messages: Sequence[ChatMessage],
**kwargs: Any,
) -> ChatResponseAsyncGen:
async def gen() -> ChatResponseAsyncGen:
for message in self.stream_chat(messages, **kwargs):
yield message
# NOTE: convert generator to async generator
return gen()
@llm_completion_callback()
async def acomplete(
self, prompt: str, formatted: bool = False, **kwargs: Any
) -> CompletionResponse:
return self.complete(prompt, formatted=formatted, **kwargs)
@llm_completion_callback()
def complete(
self, prompt: str, formatted: bool = False, **kwargs: Any
) -> CompletionResponse:
return self.complete(prompt, formatted=formatted, **kwargs)
@llm_completion_callback()
async def astream_complete(
self, prompt: str, formatted: bool = False, **kwargs: Any
) -> CompletionResponseAsyncGen:
async def gen() -> CompletionResponseAsyncGen:
for message in self.stream_complete(prompt, formatted=formatted, **kwargs):
yield message
# NOTE: convert generator to async generator
return gen()
@llm_completion_callback()
def stream_complete(
self, prompt: str, formatted: bool = False, **kwargs: Any
) -> CompletionResponseGen:
def gen() -> CompletionResponseGen:
for message in self.stream_complete(prompt, formatted=formatted, **kwargs):
yield message
return gen()
@classmethod
def class_name(cls) -> str:
return "custom_llm"
# # 1) Initialize your custom LLM
# custom_llm = Kognie(
# api_key="kg-qnA0uVr4MbJmDtpuyQEmnZWnwe6RkZjF",
# model="gpt-4o-mini"
# )
# answer = custom_llm.chat(messages=[ChatMessage(role="user", content="Who was the first president of the United States?")])
# print(answer)
# answer = custom_llm.generate_img(prompt='a dog', model='flux-pro-1.1', response_format='url')
# documents = SimpleDirectoryReader("./data").load_data()
# vector_index = VectorStoreIndex.from_documents(documents)
# query_engine = vector_index.as_query_engine()
# answer = query_engine.query(
# "what is the documents about?"
# )
# print(answer)
|