Spaces:
Sleeping
Sleeping
File size: 4,612 Bytes
c7d967d |
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 |
import chromadb
import os
from langchain_chroma import Chroma
from chromadb.config import DEFAULT_DATABASE, DEFAULT_TENANT
import time
import transformers
from langchain_community.llms import CTransformers
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_core.prompts import PromptTemplate
from transformers import pipeline
from langchain_core.output_parsers import StrOutputParser
from langchain_ollama import ChatOllama
client = chromadb.HttpClient("http://localhost:8000")
def using_ollama_model(retriever, query, results,conversation_history):
history_text = ""
for item in conversation_history:
if "question" in item and item["question"]:
history_text += f"User: {item['question']}\n"
if "answer" in item and item["answer"]:
history_text += f"Assistant: {item['answer']}\n"
print("<<<<<< LLM MODEL STARTED >>>>>>")
print(" ========>", history_text)
# Ensure the prompt template is well-structured
prompt_template = """
You are a helpful assistant. Answer the following question using the provided context and previous conversation history.
If the context does not contain the answer, only then reply with: "Sorry, I don't have enough information."
Conversation History :{history}
Context:{results}
Question:{query}
"""
# Initialize the PromptTemplate
template = PromptTemplate(
input_variables=["history","results", "query"], template=prompt_template,
)
doc_texts = "\\n".join([doc.page_content for doc in results])
formatted_output = template.format(history=history_text,results=doc_texts, query=query)
print("<<<<<<<<<<< Formatted Output >>>>>>>>>>>")
print(formatted_output)
print("type of formatted output is ", type(formatted_output))
llm = ChatOllama(model="llama3.2", temperature=0.4, num_predict=512)
rag_chain = template | llm | StrOutputParser()
# results = retriever.invoke(query)
# doc_texts = "\\n".join([doc.page_content for doc in results])
answer = rag_chain.invoke({"history" : history_text,"results": doc_texts, "query": query})
return answer
# # Set up the RAG pipeline
# rag_pipeline = RetrievalQAWithSourcesChain.from_chain_type(
# llm=llm, chain_type="stuff", retriever=retriever
# )
#
# try:
# # # answer = rag_pipeline.run(formatted_output)
# answer = rag_pipeline.invoke(formatted_output)
# return answer
# except Exception as e:
# print(f"Error occurred during invocation: {e}")
# return None
def retrievingReponse(docId, query, conversation_history) :
model_kwargs = {"device": "mps"}
encode_kwargs = {"normalize_embeddings": True}
embeddings = HuggingFaceEmbeddings(
model_name="sentence-transformers/paraphrase-distilroberta-base-v1",
model_kwargs=model_kwargs,
encode_kwargs=encode_kwargs,
)
vectorDB = Chroma(
collection_name="embeddings",
embedding_function=embeddings, # Using the encode method to get embeddings
persist_directory="MM_CHROMA_DB",
)
# retriever = vectorDB.as_retriever(
# search_type="mmr",
# search_kwargs={
# "k": 6, # was 5 originally
# "lambda_mult": 1, # was 0.30 originally
# "filter": {"docId": docId}
# }
# )
retriever = vectorDB.as_retriever(
search_type="similarity",
search_kwargs={
"k": 4, # was 5 originally
# "lambda_mult": 1, # was 0.30 originally
"filter": {"docId": docId}
}
)
# retriever = vectorDB.as_retriever()
print("<<<<<<<<<<<<<<<< Retriever >>>>>>>>>>>>>>>>")
# print("d",retriever)
print("\n")
results = retriever.invoke(
query
)
unique_results = []
seen_texts = set()
for result in results:
print(result)
# If the result's content has not been seen before, process it
if result.page_content not in seen_texts:
ans = result.page_content
ans = ans.replace("\n", "") # Clean the content by removing newlines
unique_results.append(ans) # Add the cleaned answer to the results list
seen_texts.add(result.page_content) # Mark this text as seen
os.environ["TOKENIZERS_PARALLELISM"] = "false"
start = time.time()
# llm_result = using_llm_model(retriever, query, results)
llm_result = using_ollama_model(retriever, query, results, conversation_history)
end = time.time()
print("Inference Time:>>>>>>> ", end - start)
return llm_result
|