Add AI Debate App Demo

AI Debate app.md

@@ -0,0 +1,37 @@
+# AI Debate App
+
+## Introduction
+Debate-App is a web application that enables a back-and-forth conversation between two Language Models (LLMs) on a topic chosen by the user. Users can select any two models, input a query, and visualize the dialogue between the LLMs in real-time. Built using Unify and deployed with Streamlit, this application provides a platform for users to witness AI-generated debates and explore the capabilities of different language models.
+
+## Quick Demo
+https://github.com/Sanjay8602/Debate-App/assets/121057369/a81d4b3b-bf6f-44b6-bdba-f316b8054a77
+
+## Repository and Deployment
+Repo Link:
+```commandline 
+https://github.com/Sanjay8602/Debate-App
+```
+Guide to run locally:
+```commandline 
+git clone https://github.com/Sanjay8602/Debate-App.git
+```
+Installing Dependencies:
+```commandline 
+pip install -r requirements.txt
+```
+Run Locally:
+```commandline 
+streamlit run app.py
+```
+Deployed app link:
+```commandline 
+https://sanjay8602-debate-app-app-kt5o9f.streamlit.app/
+```
+
+## Contributors
+Contributors to the project:
+
+|     Name      |               GitHub Profile                   |
+|---------------|------------------------------------------------|
+| Sanjay Suthar | [Sanjay8602](https://github.com/Sanjay8602)    |
+| Ogban Ugot    | [ogbanugot](https://github.com/ogbanugot)      |

debate/README.md

@@ -0,0 +1,31 @@
+# LLM Debate App built using Unify and Streamlit
+
+## Overview
+
+This application enables a back and forth conversation between two LLMs on a topic chosen by the user.  
+Users should be able to:
+- choose any two models
+- input a query
+- visualize the dialogue between the LLMs
+
+Built using [Unify](https://unify.ai/docs/index.html).
+
+## Requirements for running locally
+```commandline 
+git clone https://github.com/Sanjay8602/Debate-App.git
+```
+
+```commandline 
+pip install -r requirements.txt
+```
+
+```commandline 
+streamlit run app.py
+```
+## Contributors
+Contributors to the project:
+
+|     Name      |               GitHub Profile                   |
+|---------------|------------------------------------------------|
+| Sanjay Suthar | [Sanjay8602](https://github.com/Sanjay8602)    |
+| Ogban Ugot    | [ogbanugot](https://github.com/ogbanugot)      |

debate/app.py

@@ -0,0 +1,143 @@
+import streamlit as st
+from unify import Unify
+
+st.set_page_config(page_title="Debate App built with Unify")
+
+
+def start_interaction():
+    st.session_state.continue_interaction = True
+
+
+def stop_interaction():
+    st.session_state.continue_interaction = False
+
+
+def clear_history():
+    st.session_state.continue_interaction = False
+    st.session_state.model1_messages = []
+    st.session_state.model2_messages = []
+
+endpoints = ["llama-2-13b-chat@anyscale", "mistral-7b-instruct-v0.1@deepinfra", "gpt-4@deepinfra", "codellama-7b-instruct@octoai",
+             "gpt-3.5-turbo@openai", "pplx-70b-chat@perplexity-ai", "llama-3-8b-chat@together-ai", "gemma-2b-it@together-ai", "gpt-4-turbo@openai",
+             "deepseek-coder-33b-instruct@together-ai", "mistral-large@mistral-ai", "llama-3-8b-chat@fireworks-ai"]
+
+def input_fields():
+    with st.sidebar:
+        st.session_state.unify_key = st.text_input("UNIFY KEY", type="password")
+        st.image("robot_icon_green.png", width=20)
+        st.session_state.llm_1 = st.selectbox(
+            "Select LLM to debate supporting the topic",
+            endpoints,
+            key="endpoints_1_llm",
+
+        )
+        personas = ["factual", "funny", "silly", "serious", "angry"]
+        st.session_state.llm_1_persona = st.selectbox(
+            "Select the LLMs persona",
+            personas,
+            key="llm1_persona"
+        )
+
+        st.image("robot_icon_yellow.png", width=20)
+        st.session_state.llm_2 = st.selectbox(
+            "Select LLM to debate opposing the topic",
+            endpoints,
+            key= "endpoints_2_llm",
+
+        )
+
+        st.session_state.llm_2_persona = st.selectbox(
+            "Select the LLMs persona",
+            personas,
+            key="llm2_persona"
+        )
+        # Initialize stop button
+        if st.button("Stop debate", help="stop the debate at "
+                                         "the last complete "
+                                         "reply-response "
+                                         "cycle"):
+            stop_interaction()
+        else:
+            pass
+
+        # Clear history
+        if st.button("Clear chat history"):
+            clear_history()
+        else:
+            pass
+
+
+def initialize_model(llm_endpoint, unify_key):
+    model = Unify(
+        api_key=unify_key,
+        endpoint=llm_endpoint
+    )
+    return model
+
+
+# Function to generate response from a model given a prompt
+def generate_response(model, topic, position, persona, prompt):
+    messages = [
+        {"role": "system", "content": f"You are debating {position} the following topic: {topic}. "
+                                      f"Consider the opposing points and provide a response. "
+                                      f"Adopt a {persona} persona when responding."},
+    ]
+    messages.extend(prompt)
+    return model.generate(messages=messages, stream=True)
+
+
+def main():
+    st.title("Debate App built with Unify")
+    st.text("Choose two LLMs to debate each other on a given topic.")
+
+    input_fields()
+
+    if 'continue_interaction' not in st.session_state:
+        st.session_state.continue_interaction = True
+
+    if "model1_messages" not in st.session_state:
+        st.session_state.model1_messages = []
+
+    if "model2_messages" not in st.session_state:
+        st.session_state.model2_messages = []
+
+    with st.form(key='my_form'):
+        topic = st.text_input(label='Enter the debate topic here:')
+        submit = st.form_submit_button(label='Start debate')
+
+    if len(st.session_state.model1_messages) > 0 and len(st.session_state.model2_messages) > 0:
+        for _i, (model1_message, model2_message) in enumerate(
+                zip(st.session_state.model1_messages, st.session_state.model2_messages)):
+            with st.chat_message(name="model1", avatar="robot_icon_green.png"):
+                st.write(model1_message)
+            with st.chat_message(name="model2", avatar="robot_icon_yellow.png"):
+                st.write(model2_message)
+
+    model1 = initialize_model(st.session_state.llm_1, st.session_state.unify_key)
+    model2 = initialize_model(st.session_state.llm_2, st.session_state.unify_key)
+    if submit:
+        st.session_state.continue_interaction = True
+        model1_messages = []
+        model2_messages = []
+        while st.session_state.continue_interaction:
+            with st.chat_message(name="model1", avatar="robot_icon_green.png"):
+                if len(model1_messages) == 0:
+                    stream = generate_response(model1, topic, "for", st.session_state.llm_1_persona,
+                                               [{"role": "user", "content": "start debate."}])
+                else:
+                    model1_messages.append({"role": "user", "content": model2_response})
+                    stream = generate_response(model1, topic, "for", st.session_state.llm_1_persona, model1_messages)
+                model1_response = st.write_stream(stream)
+            model1_messages.append({"role": "assistant", "content": model1_response})
+            st.session_state.model1_messages.append(model1_response)
+
+            with st.chat_message(name="model2", avatar="robot_icon_yellow.png"):
+                model2_messages.append({"role": "user", "content": model1_response})
+                stream = generate_response(model2, topic, "against", st.session_state.llm_2_persona, model2_messages)
+                model2_response = st.write_stream(stream)
+            model2_messages.append({"role": "assistant", "content": model2_response})
+            st.session_state.model2_messages.append(model2_response)
+
+
+if __name__ == "__main__":
+    main()

debate/requirements.txt

@@ -0,0 +1,2 @@
+streamlit~=1.33.0
+unifyai~=0.8.2