feat: temperature adjustment

InnovationHub/llm/chain.py

@@ -11,6 +11,71 @@ from langchain import OpenAI, ConversationChain, LLMChain, PromptTemplate
 from langchain.chains.conversation.memory import ConversationalBufferWindowMemory
 from EdgeGPT import Chatbot
 
+
+class VehicleManualChatbot:
+    def __init__(self, db_paths, vehicle_options, embeddings):
+        self.db_paths = db_paths
+        self.vehicle_options = vehicle_options
+        self.embeddings = embeddings
+        self.chatgpt_chain = self._create_chatgpt_chain()
+
+
+    def _create_chatgpt_chain(self, temperature=0.5):
+        template = """
+        {history}
+        Human: {human_input}
+        Assistant:"""
+
+        prompt = PromptTemplate(
+            input_variables=["history", "human_input"],
+            template=template
+        )
+
+        return LLMChain(
+            llm=OpenAI(temperature=temperature),
+            prompt=prompt,
+            verbose=True,
+            memory=ConversationalBufferWindowMemory(k=2),
+        )
+
+
+    def _get_prompt(self, question, vehicle, k=4):
+        prompt = f"""
+        I need information from my {vehicle} manual.
+        I will provide an excerpt from the manual. Use the excerpt and nothing else to answer the question.
+        You must refer to the excerpt as "{vehicle} Manual" in your response. Here is the excerpt:
+        """
+
+        index = FAISS.load_local(
+            folder_path=self.db_paths[vehicle], embeddings=self.embeddings)
+        similar_docs = index.similarity_search(query=question, k=k)
+        context = []
+        for d in similar_docs:
+            content = d.page_content
+            context.append(content)
+        user_input = prompt + '\n[EXCERPT]' + '\n' + \
+            '\n'.join(context[:k]) + '\n' + 'Question:\n' + question
+        return user_input
+
+
+    def _ask_question(self, question, vehicle, k=2):
+        index = FAISS.load_local(
+            folder_path=self.db_paths[vehicle], embeddings=self.embeddings)
+
+        prompt = self._get_prompt(question=question, vehicle=vehicle,
+                                  k=k)
+        response = self.chatgpt_chain.predict(human_input=prompt)
+        return response
+
+
+    def chat(self, question, vehicle, k=2, temperature=0.5):
+        self.chatgpt_chain = self._create_chatgpt_chain(temperature=temperature)
+        response = self._ask_question(question=question, vehicle=vehicle,
+                                    k=k)
+        return response
+
+
+
 db_paths = {
     "S-Class": "data/s-class-manual",
     "EQS": "data/eqs-manual"
@@ -18,84 +83,27 @@ db_paths = {
 
 embeddings = HuggingFaceEmbeddings()
 
-
-template = """
-{history}
-Human: {human_input}
-Assistant:"""
-
-prompt = PromptTemplate(
-    input_variables=["history", "human_input"],
-    template=template
-)
-
-chatgpt_chain = LLMChain(
-    llm=OpenAI(temperature=0),
-    prompt=prompt,
-    verbose=True,
-    memory=ConversationalBufferWindowMemory(k=2),
-)
-human_input = """I want you to act as a voice assistant for a Mercedes-Benz vehicle. 
-    I will provide you with excerpts from a vehicle manual. 
-    You must use the excerpts to answer the user's question as best as you can. 
-    If you are unsure about the answer, you will truthfully say "not sure".
-    Let's think step by step.
-"""
-bot_response = chatgpt_chain.predict(human_input=human_input)
-
-
-def get_prompt(question, vehicle, embeddings, k=4):
-    prompt = f"""
-    I need information from my {vehicle} manual. 
-    I will provide an excerpt from the manual. Use the excerpt and nothing else to answer the question. 
-    You must refer to the excerpt as "{vehicle} Manual" in your response. Here is the excerpt:
-    """
-    
-    index = FAISS.load_local(folder_path=db_paths[vehicle], embeddings=embeddings)
-    similar_docs = index.similarity_search(query=question, k=k)
-    context = []
-    for d in similar_docs:
-        content = d.page_content
-        context.append(content)
-    user_input = prompt + '\n[EXCERPT]' + '\n' + \
-        '\n'.join(context[:k]) + '\n' + 'Question:\n' + question
-    return user_input
-
-
-def ask_question(question, vehicle, embeddings, chatgpt_chain, k=2):
-    index = FAISS.load_local(
-        folder_path=db_paths[vehicle], embeddings=embeddings)
-    
-    prompt = get_prompt(question=question, vehicle=vehicle,
-                        embeddings=embeddings, k=k)
-    response = chatgpt_chain.predict(human_input=prompt)
-    return response
-
-
-async def chatbot(question, vehicle, k=2):    
-    response = ask_question(question=question, vehicle=vehicle,
-                        embeddings=embeddings, chatgpt_chain=chatgpt_chain, k=2)
-    return response    
-
+vehicle_options = ["S-Class", "EQS"]
+chatbot = VehicleManualChatbot(db_paths=db_paths,
+                               vehicle_options=vehicle_options,
+                               embeddings=embeddings)
 
 def start_ui():
-    vehicle_options = ["S-Class", "EQS"]
     chatbot_interface = gradio.Interface(
-        fn=chatbot,
-        inputs=[
-            "text",
-            gradio.inputs.Dropdown(vehicle_options, label="Select Vehicle Model"),
-            gradio.inputs.Slider(minimum=1, maximum=10, step=1, label="k")
+        fn=chatbot.chat,
+        inputs=["text",
+                gradio.inputs.Dropdown(
+                    vehicle_options, label="Select Vehicle Model"),
+                gradio.inputs.Slider(minimum=1, maximum=10, step=1, label="k")
         ],
         outputs="text",
         title="Owner's Manual",
         description="Ask your vehicle manual and get a response.",
-        examples=[
-            ["What are the different features of the dashboard console?", "S-Class", 2],
-            ["What is flacon?", "S-Class", 3],
-            ["What is hyperscreen?", "EQS", 2],
-            ["Where can I find my vin?", "EQS", 3]
+        examples=[["What are the different features of the dashboard console?", "S-Class", 2],
+                  ["What is flacon?", "S-Class", 3],
+                  ["What is hyperscreen?", "EQS", 2],
+                  ["Where can I find my vin?", "EQS", 3]
         ]
     )
 
-    chatbot_interface.launch()
+    chatbot_interface.launch()

InnovationHub/llm/vector_store.py

@@ -85,7 +85,7 @@ def convert_to_html(similar_docs):
     return html
 
 
-def create_similarity_plot(embeddings, labels, query_index, n_clusters=3):
+def create_similarity_plot(embeddings, labels, query, n_clusters=3):
     # Only include embeddings that have corresponding labels
     embeddings_with_labels = [
         embedding for i, embedding in enumerate(embeddings) if i < len(labels)]
@@ -100,16 +100,16 @@ def create_similarity_plot(embeddings, labels, query_index, n_clusters=3):
 
     # Create a trace for the query point
     query_trace = go.Scatter3d(
-        x=[pca_embeddings[query_index, 0]],
-        y=[pca_embeddings[query_index, 1]],
-        z=[pca_embeddings[query_index, 2]],
+        x=[pca_embeddings[-1, 0]],
+        y=[pca_embeddings[-1, 1]],
+        z=[pca_embeddings[-1, 2]],
         mode='markers',
         marker=dict(
             color='black',
             symbol='diamond',
             size=10
         ),
-        name='Query'
+        name=f"Query: '{query}'"
     )
 
     # Create a trace for the other points

app.py

@@ -2,30 +2,27 @@ from InnovationHub.llm.vector_store import *
 from InnovationHub.llm.chain import *
 
 
-"""
-# Create the vector index
-db_path = "./data/s-class-manual"
-embeddings = HuggingfaceEmbeddings()
-index = FAISS(docs=docs, folder_path=db_path, embeddings=embeddings)
-"""
-
-# Launch the Gradio UI
-def start_gradio():
+def start_ui():
     chatbot_interface = gradio.Interface(
-        fn=chatbot,
-        inputs=["text", gradio.inputs.Checkbox(label="Create bot"), gradio.inputs.Slider(
-            minimum=1, maximum=10, step=1, label="k")],
+        fn=chatbot.chat,
+        inputs=["text",
+                gradio.inputs.Dropdown(
+                    vehicle_options, label="Select Vehicle Model"),
+                gradio.inputs.Slider(minimum=1, maximum=10, step=1, label="k"),
+                gradio.inputs.Slider(
+                    minimum=0.0, maximum=1.0, step=0.01, label="Temperature")
+                ],
         outputs="text",
-        title="Mercedes-Benz S-Class Owner's Manual",
-        description="Ask your vehicle's manual questions and get answers",
-        examples=[
-            ["What are the different features of the dashboard console?", True, 2],
-            ["What do they do?", False, 3]
-        ]
+        title="Owner's Manual",
+        description="Ask your vehicle manual and get a response.",
+        examples=[["What are the different features of the dashboard console?", "S-Class", 2, 0.5],
+                  ["What is flacon?", "S-Class", 3, 0.8],
+                  ["What is hyperscreen?", "EQS", 2, 0.2],
+                  ["Where can I find my vin?", "EQS", 3, 0.3]
+                  ]
     )
-    chatbot_interface.launch()
 
+    chatbot_interface.launch()
 
 if __name__ == '__main__':
-    start_ui()
-
+    start_ui()