Spaces:
Sleeping
Sleeping
File size: 6,457 Bytes
478dec6 | 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 | from pydantic import BaseModel
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import AzureChatOpenAI
from tenacity import (
retry,
stop_after_attempt,
wait_exponential,
retry_if_exception_type
)
from typing import Dict
from externals.observability.langfuse import langfuse_handler, langfuse
from services.llms.LLM import model_5mini, model_4omini
from utils.decorator import trace_runtime
from utils.logger import get_logger
logger = get_logger("base generator")
class MetadataObservability(BaseModel):
fullname: str
task_id: str
agent: str
class BaseAIGenerator:
"""
Args:
name:str,
prompt: ChatPromptTemplate,
input_llm: Dict,
metadata_observability: MetadataObservability,
output_model: BaseModel,
llm:AzureChatOpenAI = model_5mini | model_4omini,
"""
def __init__(self,
task_name:str,
prompt: ChatPromptTemplate,
input_llm: Dict,
metadata_observability: MetadataObservability,
llm:AzureChatOpenAI = model_5mini | model_4omini,
):
self.name = task_name
self.llm = llm
self.prompt = prompt
self.input_llm = input_llm
self.metadata_observability = metadata_observability
@retry(
reraise=True,
stop=stop_after_attempt(2), # retry max 3 times
wait=wait_exponential(multiplier=1, min=1, max=5),
retry=retry_if_exception_type(Exception) # retry on any exception from LLM
)
async def _asafe_invoke(self, chain, input_llm, config):
"""private helper for retries"""
return await chain.ainvoke(input_llm, config=config)
@retry(
reraise=True,
stop=stop_after_attempt(2), # retry max 3 times
wait=wait_exponential(multiplier=1, min=1, max=5),
retry=retry_if_exception_type(Exception) # retry on any exception from LLM
)
async def _safe_invoke(self, chain, input_llm, config):
"""private helper for retries"""
return chain.invoke(input_llm, config=config)
# @trace_runtime
# async def agenerate(self):
# try:
# chain = self.prompt | self.llm
# config = {"callbacks": [langfuse_handler]}
# with langfuse.start_as_current_observation(
# as_type='generation',
# name=self.name,
# input=self.input_llm,
# ) as trace:
# trace.update_trace(user_id=self.metadata_observability.fullname,
# session_id=self.metadata_observability.task_id,
# metadata=self.metadata_observability.model_dump()
# )
# output = await self._asafe_invoke(chain=chain,
# input_llm=self.input_llm,
# config=config)
# trace.update_trace(output=output)
# return output
# except Exception as E:
# logger.error(f"❌ BaseGenerator, agenerate error, {E}")
# return None
# @trace_runtime
# async def generate(self):
# try:
# chain = self.prompt | self.llm
# config = {"callbacks": [langfuse_handler]}
# with langfuse.start_as_current_observation(
# as_type='generation',
# name=self.name,
# input=self.input_llm,
# ) as trace:
# trace.update_trace(user_id=self.metadata_observability.fullname,
# session_id=self.metadata_observability.task_id,
# metadata=self.metadata_observability.model_dump()
# )
# output = self._safe_invoke(chain=chain,
# input_llm=self.input_llm,
# config=config)
# trace.update_trace(output=output)
# return output
# except Exception as E:
# logger.error(f"❌ BaseGenerator, generate error, {E}")
# return None
@trace_runtime
async def agenerate(self):
trace = None
try:
chain = self.prompt | self.llm
config = {"callbacks": [langfuse_handler]}
# ✅ Create trace (no context manager, no end())
trace = langfuse.trace(
name=self.name,
input=self.input_llm,
)
trace.update(
user_id=self.metadata_observability.fullname,
session_id=self.metadata_observability.task_id,
metadata=self.metadata_observability.model_dump(),
)
output = await self._asafe_invoke(
chain=chain,
input_llm=self.input_llm,
config=config,
)
trace.update(output=output)
return output
except Exception as e:
logger.exception("❌ BaseGenerator agenerate error")
if trace:
trace.update(
status="error",
error=str(e),
)
return None
@trace_runtime
async def generate(self):
trace = None
try:
chain = self.prompt | self.llm
config = {"callbacks": [langfuse_handler]}
trace = langfuse.trace(
name=self.name,
input=self.input_llm,
)
trace.update(
user_id=self.metadata_observability.fullname,
session_id=self.metadata_observability.task_id,
metadata=self.metadata_observability.model_dump(),
)
output = self._safe_invoke(
chain=chain,
input_llm=self.input_llm,
config=config,
)
trace.update(output=output)
return output
except Exception as e:
logger.exception("❌ BaseGenerator generate error")
if trace:
trace.update(
status="error",
error=str(e),
)
return None
|