File size: 4,027 Bytes
c2c5723 8359d12 c2c5723 8359d12 6386510 c2c5723 6386510 8359d12 62138c4 c2c5723 b70c257 c2c5723 8359d12 c2c5723 652620b 8359d12 7cb9567 c2c5723 8359d12 b70c257 c2c5723 b70c257 62138c4 8359d12 b70c257 c2c5723 f3bc24e c2c5723 f77fb99 c2c5723 b70c257 f3bc24e c2c5723 561ff95 c2c5723 51a7d9e 8359d12 c2c5723 8359d12 c2c5723 8359d12 51a7d9e 8359d12 51a7d9e 8359d12 |
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 |
from langchain_community.vectorstores import Qdrant
from langchain_huggingface import HuggingFaceEmbeddings
import os
from dotenv import load_dotenv
from langchain.prompts import ChatPromptTemplate
from langchain.schema.runnable import RunnablePassthrough
from langchain.schema.output_parser import StrOutputParser
from qdrant_client import QdrantClient, models
from langchain_qdrant import Qdrant
from langchain_qdrant import QdrantVectorStore
from langchain_huggingface import ChatHuggingFace
# Load environment variables
load_dotenv()
HF_TOKEN = os.environ["HF_TOKEN"] = os.getenv("HF_TOKEN")
# HuggingFace Embeddings
embeddings = HuggingFaceEmbeddings(model_name="BAAI/bge-large-en-v1.5")
# Qdrant Client Setup
client = QdrantClient(
url=os.getenv("QDRANT_URL"),
api_key=os.getenv("QDRANT_API_KEY"),
prefer_grpc=True
)
collection_name = "mawared"
# Try to create collection, handle if it already exists
try:
client.create_collection(
collection_name=collection_name,
vectors_config=models.VectorParams(
size=768, # GTE-large embedding size
distance=models.Distance.COSINE
),
)
print(f"Created new collection: {collection_name}")
except Exception as e:
if "already exists" in str(e):
print(f"Collection {collection_name} already exists, continuing...")
else:
raise e
# Create Qdrant vector store
db = Qdrant(
client=client,
collection_name=collection_name,
embeddings=embeddings,
)
# Create retriever
retriever = db.as_retriever(
search_type="similarity",
search_kwargs={"k": 5}
)
llm = ChatOpenAI(base_url="https://api-inference.huggingface.co/v1/", temperature=0 , api_key=HF_TOKEN , model="meta-llama/Llama-3.3-70B-Instruct")
# Create prompt template
template = """
You are an expert assistant specializing in the Mawared HR System. Your task is to answer the user's question strictly based on the provided context. If the context lacks sufficient information, ask focused clarifying questions to gather additional details.
To improve your responses, follow these steps:
Chain-of-Thought (COT): Break down complex queries into logical steps. Use tags like [Step 1], [Step 2], etc., to label each part of the reasoning process. This helps structure your thinking and ensure clarity. For example:
[Step 1] Identify the key details in the context relevant to the question.
[Step 2] Break down any assumptions or information gaps.
[Step 3] Combine all pieces to form the final, well-reasoned response.
Reasoning: Demonstrate a clear logical connection between the context and your answer at each step. If information is missing or unclear, indicate the gap using tags like [Missing Information] and ask relevant follow-up questions to fill that gap.
Clarity and Precision: Provide direct, concise answers focused only on the context. Avoid including speculative or unrelated information.
Follow-up Questions: If the context is insufficient, focus on asking specific, relevant questions. Label them as [Clarifying Question] to indicate they are needed to complete the response. For example:
[Clarifying Question] Could you specify which employee section you're referring to?
Context:
{context}
Question:
{question}
Answer
"""
prompt = ChatPromptTemplate.from_template(template)
# Create the RAG chain using LCEL with prompt printing and streaming output
rag_chain = (
{"context": retriever, "question": RunnablePassthrough()}
| prompt
| llm
| StrOutputParser()
)
# Function to ask questions
def ask_question(question):
print("Answer:\t", end=" ", flush=True)
for chunk in rag_chain.stream(question):
print(chunk, end="", flush=True)
print("\n")
# Example usage
if __name__ == "__main__":
while True:
user_question = input("\n \n \n Ask a question (or type 'quit' to exit): ")
if user_question.lower() == 'quit':
break
answer = ask_question(user_question)
# print("\nFull answer received.\n") |