Upload 4 files

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+vector-db/index.faiss filter=lfs diff=lfs merge=lfs -text

app.py

@@ -0,0 +1,121 @@
+from langchain_community.document_loaders import PyPDFLoader,DirectoryLoader
+from langchain.embeddings import HuggingFaceEmbeddings
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain_community.vectorstores import FAISS
+import os
+from langchain.prompts import PromptTemplate
+from langchain_together import Together
+from langchain.chat_models import ChatOpenAI
+from langchain.embeddings import openai
+from langchain.memory import ConversationBufferWindowMemory
+from langchain.chains import ConversationalRetrievalChain
+import streamlit as st
+import time
+
+os.environ["OPENAI_API_KEY"] = "sk-proj-Ht5xjDYvIipDt2wohlMCT3BlbkFJUCTqmrg8TRufd642B0cc"
+openai.api_key = os.environ["OPENAI_API_KEY"]
+from langchain.embeddings.openai import OpenAIEmbeddings
+embeddings=OpenAIEmbeddings()
+
+#embeddings = HuggingFaceEmbeddings(model_name="nomic-ai/nomic-embed-text-v1",model_kwargs={"trust_remote_code":True,"revision":"289f532e14dbbbd5a04753fa58739e9ba766f3c7"})
+#vectordb = Chroma.from_documents(texts, embedding=embeddings, persist_directory="./data")
+#db_retriever =vectordb.as_retriever(search_type="similarity",search_kwargs={'k':4})
+
+db = FAISS.load_local("vector-db", embeddings, allow_dangerous_deserialization=True)
+db_retriever = db.as_retriever(search_type="similarity",search_kwargs={"k": 4})
+
+
+st.set_page_config(page_title="Qanoon-Bot")
+col1, col2, col3 = st.columns([1,4,1])
+with col2:
+    st.image("https://s3.ap-south-1.amazonaws.com/makerobosfastcdn/cms-assets/Legal_AI_Chatbot.png")
+
+st.markdown(
+    """
+    <style>
+div.stButton > button:first-child {
+    background-color: #ffd0d0;
+}
+div.stButton > button:active {
+    background-color: #ff6262;
+}
+   div[data-testid="stStatusWidget"] div button {
+        display: none;
+        }
+    
+    .reportview-container {
+            margin-top: -2em;
+        }
+        #MainMenu {visibility: hidden;}
+        .stDeployButton {display:none;}
+        footer {visibility: hidden;}
+        #stDecoration {display:none;}
+    button[title="View fullscreen"]{
+    visibility: hidden;}
+        </style>
+""",
+    unsafe_allow_html=True,
+)
+
+def reset_conversation():
+  st.session_state.messages = []
+  st.session_state.memory.clear()
+
+if "messages" not in st.session_state:
+    st.session_state.messages = []
+
+if "memory" not in st.session_state:
+    st.session_state.memory = ConversationBufferWindowMemory(k=2, memory_key="chat_history",return_messages=True) 
+
+##embeddings = HuggingFaceEmbeddings(model_name="nomic-ai/nomic-embed-text-v1",model_kwargs={"trust_remote_code":True,"revision":"289f532e14dbbbd5a04753fa58739e9ba766f3c7"})
+#db=FAISS.load_local("/content/ipc_vector_db", embeddings, allow_dangerous_deserialization=True)
+
+prompt_template = """<s>[INST]This is a chat template and As a legal chat bot specializing in pakistan Property laws and rights queries and , your primary objective is to provide accurate and concise information based on the user's questions. offer relevant context from the knowledge base while avoiding unnecessary details. Your responses will be brief, to the point, and in compliance with the established format. If a question falls outside the given context, you will refrain from utilizing the chat history and instead rely on your own knowledge base to generate an appropriate response. You will prioritize the user's query and refrain from posing additional questions. The aim is to deliver professional, precise, and contextually relevant information pertaining to the Indian Penal Code.
+CONTEXT: {context}
+CHAT HISTORY: {chat_history}
+QUESTION: {question}
+ANSWER:
+</s>[INST]
+"""
+
+prompt = PromptTemplate(template=prompt_template,
+                        input_variables=['context', 'question', 'chat_history'])
+
+# You can also use other LLMs options from https://python.langchain.com/docs/integrations/llms. Here I have used TogetherAI API
+
+
+llm=ChatOpenAI(temperature=0.2,model_name='gpt-4-turbo')
+qa = ConversationalRetrievalChain.from_llm(
+    llm=llm,
+    memory=st.session_state.memory,
+    retriever=db_retriever,
+    combine_docs_chain_kwargs={'prompt': prompt}
+)
+
+for message in st.session_state.messages:
+    with st.chat_message(message.get("role")):
+        st.write(message.get("content"))
+
+input_prompt = st.chat_input("Say something")
+
+if input_prompt:
+    with st.chat_message("user"):
+        st.write(input_prompt)
+
+    st.session_state.messages.append({"role":"user","content":input_prompt})
+
+    with st.chat_message("assistant"):
+        with st.status("Thinking 💡...",expanded=True):
+            result = qa.invoke(input=input_prompt)
+
+            message_placeholder = st.empty()
+
+            full_response = "**_Note: Information provided by Qanoon-Bot may be inaccurate. ** \n\n\n"
+        for chunk in result["answer"]:
+            full_response+=chunk
+            time.sleep(0.02)
+            
+            message_placeholder.markdown(full_response+" ▌")
+        st.button('Reset All Chat 🗑️', on_click=reset_conversation)
+
+    st.session_state.messages.append({"role":"assistant","content":result["answer"]})

requirements.txt

@@ -0,0 +1,12 @@
+langchain
+pypdf
+transformers
+sentence-transformers
+accelerate
+faiss-cpu
+streamlit
+tiktoken
+openai
+langchain-fireworks
+einops
+langchain-together

vector-db/index.faiss

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0261ac5964a31be34506606a99a0c181920835c14a42592fe41ce9e78e12f31b
+size 11870253

vector-db/index.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:314117d8fcda7744e1a7185ed972f047f85a4579a55d64a02ffe1e5f16b0db43
+size 1876337