File size: 1,099 Bytes
2e5aca2 a8ab4cd 2e5aca2 a8ab4cd 2e5aca2 a8ab4cd 2e5aca2 a8ab4cd |
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 |
from utils.prompts import GRADER_SYSTEM_PROMPT, GRADER_PROMPT
from utils.config import GRADER_MODEL
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.pydantic_v1 import BaseModel, Field
from langchain_openai import ChatOpenAI
## Grading model
class DocumentGrade(BaseModel):
"""Binary score for relevance check on retrieved documents"""
binary_score: str = Field(
description='Document is relevant to the question, "yes" or "no"'
)
reasoning: str = Field(description="Reasoning for the score")
# Grader Prompts
grade_prompt = ChatPromptTemplate.from_messages(
[
("system", GRADER_SYSTEM_PROMPT),
("human", GRADER_PROMPT),
]
)
# LLM with function call
llm = ChatOpenAI(model=GRADER_MODEL, temperature=0, streaming=True)
structured_llm_grader = llm.with_structured_output(DocumentGrade)
retrieval_grader = grade_prompt | structured_llm_grader
# question = "agent memory"
# docs = retriever.invoke(question)
# doc_txt = docs[1].page_content
# print(retrieval_grader.invoke({"question": question, "document": doc_txt}))
|