added first edition of working code

app.py

@@ -0,0 +1,87 @@
+import gradio as gr
+from huggingface_hub import InferenceClient #imports huggingface models
+
+# NEW LIBRARIES
+
+from sentence_transformers import SentenceTransformer
+import torch
+import numpy as np
+
+## START NEW CODE
+
+# 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
+
+## END OF NEW CODE
+
+client = InferenceClient("google/gemma-2-2b-it")
+
+def respond(message, history):
+    messages = [{"role": "system", "content": "I am a kind chatbot."}]
+
+    # NEW CODE
+    # Retrieve context relevant to the current user message
+    context = get_relevant_context(message, top_k=3)
+
+    # 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:
+
+chatbot = gr.ChatInterface(respond)
+
+chatbot.launch()