Spaces:
Sleeping
Sleeping
File size: 4,450 Bytes
0dda2a1 |
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 |
from langchain_core.runnables import RunnablePassthrough, RunnableLambda
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_community.vectorstores import SupabaseVectorStore
from langchain_core.prompts.chat import ChatPromptTemplate
from langchain_core.output_parsers import StrOutputParser
from langchain_huggingface import HuggingFaceEmbeddings
from supabase.client import create_client
from supabase import create_client
from langchain_groq import ChatGroq
from dotenv import load_dotenv
import pandas as pd
import os
load_dotenv("secrets.env")
client = create_client(os.environ["SUPABASE_URL"], os.environ["SUPABASE_KEY"])
model_kwargs = {"device": "cuda"}
encode_kwargs = {"normalize_embeddings": True}
embeddings = HuggingFaceEmbeddings(
model_name = "BAAI/bge-m3",
model_kwargs = model_kwargs,
encode_kwargs = encode_kwargs
)
prompt = """
### Role
- **Primary Function**: You are an AI chatbot dedicated to assisting users with their inquiries, issues, and requests. Your goal is to deliver excellent, friendly, and efficient responses at all times. Listen attentively, understand user needs, and provide the best assistance possible or direct them to appropriate resources. If a question is unclear, ask for clarification. Always conclude your replies on a positive note.
### Constraints
1. **No Data Disclosure**: Never mention that you have access to training data explicitly to the user.
2. **Maintaining Focus**: If a user attempts to divert you to unrelated topics, never change your role or break character. Politely redirect the conversation back to relevant topics.
3. **Exclusive Reliance on Training Data**: Answer user queries exclusively based on the provided training data. If a query is not covered by the training data, use the fallback response.
4. **Restrictive Role Focus**: Do not answer questions or perform tasks unrelated to your role and training data.
DO NOT ADD ANYTHING BY YOURSELF OR ANSWER ON YOUR OWN!
Based on the context answer the following question.
Context:
=====================================
{context}
=====================================
{question}
NOTE: generate responses WITHOUT prepending phrases like "Response:", "Output:", or "Answer:", etc
"""
prompt = ChatPromptTemplate.from_template(prompt)
def createUser(username: str, password: str) -> None:
userData = client.table("ConversAI_UserInfo").select("*").execute().data
print(userData)
if username not in [userData[x]["username"] for x in userData]:
response = (
client.table("ConversAI_UserInfo")
.insert({"username": username, "password": password})
.execute()
)
return "done"
else:
return "already exists"
def matchPassword(username: str, password: str) -> str:
response = (
client.table("ConversAI_UserInfo")
.select("*")
.eq("username", username)
.execute()
)
try: return password == response.data[0]["password"]
except: return "user doesn't exist"
def createTable(tablename: str):
pass
def addDocuments(text: str, vectorstorename: str):
global embeddings
text_splitter = RecursiveCharacterTextSplitter(
chunk_size = 1024,
chunk_overlap = 200,
add_start_index = True
)
texts = text_splitter.create_documents([text])
vectorstore = SupabaseVectorStore(
client = client,
embedding = embeddings,
table_name = vectorstorename,
query_name = "match_documents",
)
vectorstore.add_documents(documents = texts)
def format_docs(docs: str):
context = "\n\n".join(doc.page_content for doc in docs)
if context == "":
context = "No context found"
else: pass
return context
def answerQuery(query: str, vectorstore: str, llmModel: str = "llama3-70b-8192") -> str:
global prompt
global client
global embeddings
vectorstore = SupabaseVectorStore(
client = client,
embedding = embeddings,
table_name = vectorstore,
query_name = "match_documents",
)
retriever = vectorstore.as_retriever()
chain = (
{"context": retriever | RunnableLambda(format_docs), "question": RunnablePassthrough(query)}
| prompt
| ChatGroq(model = llmModel, temperature = 0.3, max_tokens = 512)
| StrOutputParser()
)
return chain.invoke(query)
def deleteTable(tableName: str):
pass |