Update app.py

app.py

@@ -9,7 +9,10 @@ import gradio as gr
 import os
 import pytesseract
 from PIL import Image
-import pickle 
+import pickle
+from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint
+# Load the CSV data as a DataFrame
+df = pd.read_csv("hf://datasets/kshitij230/Indian-Law/Indian-Law.csv")
 model = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')
 index = faiss.read_index('IPC_index.faiss')
 index2 = faiss.read_index('CrpC_index.faiss')
@@ -26,6 +29,13 @@ with open('CrPC_F.pkl', 'rb') as f:
 with open('CrPC_C.pkl', 'rb') as f:
     chunk_indices2 = pickle.load(f)
 # Step 3: Retrieval with Citations using PDF filename
+def retrieve_faq(query):
+    relevant_rows = df[df['Instruction'].str.contains(query, case=False)]
+    if not relevant_rows.empty:
+        response = relevant_rows.iloc[0]['Response']
+        return response
+    else:
+        return "Sorry, I couldn't find relevant FAQs for your query."
 def retrieve_info_with_citation(query, top_k=5):
     query_embedding = model.encode([query])
     D, I = index.search(query_embedding, k=top_k)
@@ -319,9 +329,13 @@ doj_tool=Tool(
     func=doj_info,
     description="Provides Summarized Information about Department of Justice."
 )
-llm = ChatGoogleGenerativeAI(
-    model="gemini-1.5-pro",
-    temperature=0.25,
+faq_tool=Tool(
+    name="Commonly Asked Questions",
+    func=retrieve_faq,
+    description="Provides Answers to commonly asked questions related to query keyword(s)"
+)
+llm = HuggingFaceEndpoint(
+    model="meta-llama/Meta-Llama-3-8B",
     max_tokens=None,
     timeout=None,
     max_retries=2,
@@ -345,7 +359,7 @@ llm = ChatGoogleGenerativeAI(
     """
 )
 
-agent_tools = [ipc_tool,crpc_tool,doj_tool]
+agent_tools = [ipc_tool,crpc_tool,doj_tool,faq_tool]
 
 agent = initialize_agent(
     tools=agent_tools,