Made Chatbot(has buttons/images)

app.py

@@ -1,46 +1,85 @@
 import gradio as gr
-from huggingface_hub import InferenceClient
+from huggingface_hub import InferenceClient #imports huggingface models
+from sentence_transformers import SentenceTransformer
+import torch
+import numpy as np
 
+
+# Load and process the knowledge base text file
+with open("/content/Knowledge.txt", "r", encoding="utf-8") as f:
+    knowledge_text = f.read()
+
+
+# Split the text into chunks (for example, by paragraphs)
+chunks = [chunk.strip() for chunk in knowledge_text.split("\n\n") if chunk.strip()]
+
+
+# Load an embedding model (this one is light and fast)
+embedder = SentenceTransformer('all-MiniLM-L6-v2')
+
+
+# Precompute embeddings for all chunks (as a tensor for fast similarity search)
+chunk_embeddings = embedder.encode(chunks, convert_to_tensor=True)
+def get_relevant_context(query, top_k=3):
+    """
+     Compute the embedding for the query, compare it against all chunk embeddings,
+     and return the top_k most similar chunks concatenated into a context string.
+    """
+    # Compute and normalize the query embedding
+    query_embedding = embedder.encode(query, convert_to_tensor=True)
+    query_embedding = query_embedding / query_embedding.norm()
+    
+    # Normalize chunk embeddings along the embedding dimension
+    norm_chunk_embeddings = chunk_embeddings / chunk_embeddings.norm(dim=1, keepdim=True)
+    
+    # Compute cosine similarity between the query and each chunk
+    similarities = torch.matmul(norm_chunk_embeddings, query_embedding)
+    
+    # Get the indices of the top_k most similar chunks
+    top_k_indices = torch.topk(similarities, k=top_k).indices.cpu().numpy()
+    
+    # Concatenate the top chunks into a single context string
+    context = "\n\n".join([chunks[i] for i in top_k_indices])
+    return context
+
+    
 client = InferenceClient("google/gemma-2-2b-it")
 
+
 def respond(message, history):
-    messages = [{"role": "system", "content": "I am a mean chatbot."}]
+    messages = [{"role": "system", "content": ""You are ChaChingas, an AI financial advisor for students and low-income families. Only answer questions about budgeting, saving, debt, or personal finance. If a user asks about unrelated topics like recipes, sports, or entertainment, politely say: 'I'm here to help with money and budgeting—ask me anything about that!' Speak clearly, keep answers short, and use simple language. When asked about budgeting, explain the 50/30/20 rule: 50% for needs, 30% for wants, 20% for savings or debt. Be supportive, practical, and easy to understand. Avoid giving investment, tax, or legal advice, and never ask for or handle sensitive personal financial information.”
+"}]
+    context = get_relevant_context(message, top_k=3)
 
-    # add all previous messages to the messages list
+    
+# add all previous messages to the messages list
     if history:
         for user_msg, assistant_msg in history:
             messages.append({"role": "user", "content": user_msg})
             messages.append({"role": "assistant", "content": assistant_msg})
 
+            
     # add the current user's message to the messages list
     messages.append({"role": "user", "content": message})
 
+    
     # makes the chat completion API call,
     # sending the messages and other parameters to the model
     # implements streaming, where one word/token appears at a time
-
+    
     response = ""
 
+    
+    # iterate through each message in the method
     for message in client.chat_completion(
         messages,
         max_tokens=500,
         temperature=.1,
         stream=True):
+        # add the tokens to the output content
+          token = message.choices[0].delta.content # capture the most recent toke
+          response += token # Add it to the response
+          yield response # yield the response:
 
-          token = message.choices[0].delta.content 
-          response += token 
-          yield response 
-
-chatbot = gr.ChatInterface(respond, examples = ["Learn About Stocks", "Help Me Budget"], title = "ChaChingas", description = "This is a financial literacy bot!")
-
-with gr.Blocks() as chatbot:
-  gr.Image(
-      value = "/content/Banner.png",
-      show_label = False,
-      show_share_button= False,
-      show_download_button= False )
-  gr.ChatInterface(respond, type = "messages")
-
-
-chatbot.launch()
-
+chatbot = gr.ChatInterface(respond, examples = ["Learn About Stocks", "Help me Budget"], title = "ChaChingus", description = "This is a financial literacy chatbot")
+chatbot.launch()