LLMPlus/Programming_Assistant
0
1import gradio as gr2import os 3import json 4import requests5 6# Streaming endpoint 7API_URL = "https://api.openai.com/v1/chat/completions" # os.getenv("API_URL") + "/generate_stream"8 9# Inference function10def predict(openai_gpt4_key, system_msg, inputs, top_p, temperature, chat_counter, chatbot=[], history=[]): 11 12 headers = {13 "Content-Type": "application/json",14 "Authorization": f"Bearer {openai_gpt4_key}" # Users will provide their own OPENAI_API_KEY 15 }16 print(f"system message is ^^ {system_msg}")17 if system_msg.strip() == '':18 initial_message = [{"role": "user", "content": f"{inputs}"}]19 multi_turn_message = []20 else:21 initial_message = [{"role": "system", "content": system_msg},22 {"role": "user", "content": f"{inputs}"}]23 multi_turn_message = [{"role": "system", "content": system_msg}]24 25 if chat_counter == 0:26 payload = {27 "model": "gpt-4",28 "messages": initial_message, 29 "temperature": 1.0,30 "top_p": 1.0,31 "n": 1,32 "stream": True,33 "presence_penalty": 0,34 "frequency_penalty": 0,35 }36 print(f"chat_counter - {chat_counter}")37 else:38 messages = multi_turn_message # Of the type - [{"role": "system", "content": system_msg},]39 for data in chatbot:40 user = {"role": "user", "content": data[0]}41 assistant = {"role": "assistant", "content": data[1]}42 messages.append(user)43 messages.append(assistant)44 temp = {"role": "user", "content": inputs}45 messages.append(temp)46 payload = {47 "model": "gpt-4",48 "messages": messages, # [{"role": "user", "content": f"{inputs}"}],49 "temperature": temperature,50 "top_p": top_p,51 "n": 1,52 "stream": True,53 "presence_penalty": 0,54 "frequency_penalty": 0,55 }56 57 chat_counter += 158 history.append(inputs)59 print(f"Logging: payload is - {payload}")60 response = requests.post(API_URL, headers=headers, json=payload, stream=True)61 print(f"Logging: response code - {response.status_code}")62 token_counter = 0 63 partial_words = "" 64 65 counter = 066 for chunk in response.iter_lines():67 if counter == 0: # Skipping first chunk68 counter += 169 continue70 if chunk.decode():71 chunk = chunk.decode()72 if len(chunk) > 12 and "content" in json.loads(chunk[6:])['choices'][0]['delta']:73 partial_words += json.loads(chunk[6:])['choices'][0]["delta"]["content"]74 if token_counter == 0:75 history.append(" " + partial_words)76 else:77 history[-1] = partial_words78 chat = [(history[i], history[i + 1]) for i in range(0, len(history) - 1, 2)] # Convert to tuples of list79 token_counter += 180 yield chat, history, chat_counter, response # Resembles {chatbot: chat, state: history} 81 82# Resetting to blank83def reset_textbox():84 return gr.update(value='')85 86# To set a component as visible=False87def set_visible_false():88 return gr.update(visible=False)89 90# To set a component as visible=True91def set_visible_true():92 return gr.update(visible=True)93 94title = """<h1 align="center">🔥GPT4 using Chat-Completions API & 🚀Gradio-Streaming</h1>"""95theme_addon_msg = """<center>🌟 This Demo also introduces you to Gradio Themes. Discover more on Gradio website using our <a href="https://gradio.app/theming-guide/" target="_blank">Themeing-Guide🎨</a>! You can develop from scratch, modify an existing Gradio theme, and share your themes with the community by uploading them to the Hugging Face hub easily using <code>theme.push_to_hub()</code>.</center>""" 96system_msg_info = """A conversation could begin with a system message to gently instruct the assistant. System message helps set the behavior of the AI Assistant. For example, the assistant could be instructed with 'You are a helpful assistant.'"""97 98# Modifying existing Gradio Theme99theme = gr.themes.Soft(primary_hue="zinc", secondary_hue="green", neutral_hue="green", text_size=gr.themes.sizes.text_lg) 100 101with gr.Blocks(css="""#col_container { margin-left: auto; margin-right: auto;} #chatbot {height: 520px; overflow: auto;}""", theme=theme) as demo:102 gr.HTML(title)103 gr.HTML("""<h3 align="center">🔥This Huggingface Gradio Demo provides you access to GPT4 API with System Messages. Please note that you would be needing an OPENAI API key for GPT4 access🙌</h3>""")104 gr.HTML(theme_addon_msg)105 gr.HTML('''<center><a href="https://huggingface.co/spaces/ysharma/ChatGPT4?duplicate=true"><img src="https://bit.ly/3gLdBN6" alt="Duplicate Space"></a>Duplicate the Space and run securely with your OpenAI API Key</center>''')106 107 with gr.Column(elem_id="col_container"):108 with gr.Row():109 openai_gpt4_key = gr.Textbox(label="OpenAI GPT4 Key", value="", type="password", placeholder="sk..", info="You have to provide your own GPT4 keys for this app to function properly",)110 with gr.Accordion(label="System message:", open=False):111 system_msg = gr.Textbox(label="Instruct the AI Assistant to set its behavior", info=system_msg_info, value="", placeholder="Type here..")112 accordion_msg = gr.HTML(value="🚧 To set System message you will have to refresh the app", visible=False)113 114 chatbot = gr.Chatbot(label='GPT4', elem_id="chatbot")115 inputs = gr.Textbox(placeholder="Hi there!", label="Type an input and press Enter")116 state = gr.State([]) 117 with gr.Row():118 with gr.Column(scale=7):119 b1 = gr.Button("Submit") # Just create the button without attempting to style it here.120 with gr.Column(scale=3):121 server_status_code = gr.Textbox(label="Status code from OpenAI server")122 123 with gr.Accordion("Parameters", open=False):124 top_p = gr.Slider(minimum=0, maximum=1.0, value=1.0, step=0.05, interactive=True, label="Top-p (nucleus sampling)")125 temperature = gr.Slider(minimum=0, maximum=5.0, value=1.0, step=0.1, interactive=True, label="Temperature")126 chat_counter = gr.Number(value=0, visible=False, precision=0)127 128 # inputs.submit(predict, [openai_gpt4_key, system_msg, inputs, top_p, temperature, chat_counter, chatbot, state], [chatbot, state, chat_counter, server_status_code])129 # b1.click(predict, [openai_gpt4_key, system_msg, inputs, top_p, temperature, chat_counter, chatbot, state], [chatbot, state, chat_counter, server_status_code])130 131 inputs.submit(132 predict, 133 inputs=[openai_gpt4_key, system_msg, inputs, top_p, temperature, chat_counter, chatbot, state],134 outputs=[chatbot, state, chat_counter, server_status_code],135 concurrency_limit=20 # Apply concurrency limit here136 )137 b1.click(138 predict, 139 inputs=[openai_gpt4_key, system_msg, inputs, top_p, temperature, chat_counter, chatbot, state],140 outputs=[chatbot, state, chat_counter, server_status_code],141 concurrency_limit=20 # And here142 )143 144 inputs.submit(set_visible_false, inputs=[], outputs=[system_msg])145 b1.click(set_visible_false, inputs=[], outputs=[system_msg])146 inputs.submit(set_visible_true, inputs=[], outputs=[accordion_msg])147 b1.click(set_visible_true, inputs=[], outputs=[accordion_msg])148 149 b1.click(reset_textbox, inputs=[], outputs=[inputs])150 inputs.submit(reset_textbox, inputs=[], outputs=[inputs])151 152 153 #154 inputs.submit(set_visible_false, [], [system_msg])155 b1.click(set_visible_false, [], [system_msg])156 inputs.submit(set_visible_true, [], [accordion_msg])157 b1.click(set_visible_true, [], [accordion_msg])158 159 b1.click(reset_textbox, [], [inputs])160 inputs.submit(reset_textbox, [], [inputs])161 162 with gr.Accordion(label="Examples for System message:", open=False):163 gr.Examples(164 examples=[165 ["You are an AI programming assistant.\n\n- Follow the user's requirements carefully and to the letter.\n- First think step-by-step -- describe your plan for what to build in pseudocode, written out in great detail.\n- Then output the code in a single code block.\n- Minimize any other prose."],166 ["You are ComedianGPT who is a helpful assistant. You answer everything with a joke and witty replies."],167 ["You are ChefGPT, a helpful assistant who answers questions with culinary expertise and a pinch of humor."],168 ["You are FitnessGuruGPT, a fitness expert who shares workout tips and motivation with a playful twist."],169 ["You are SciFiGPT, an AI assistant who discusses science fiction topics with a blend of knowledge and wit."],170 ["You are PhilosopherGPT, a thoughtful assistant who responds to inquiries with philosophical insights and a touch of humor."],171 ["You are EcoWarriorGPT, a helpful assistant who shares environment-friendly advice with a lighthearted approach."],172 ["You are MusicMaestroGPT, a knowledgeable AI who discusses music and its history with a mix of facts and playful banter."],173 ["You are SportsFanGPT, an enthusiastic assistant who talks about sports and shares amusing anecdotes."],174 ["You are TechWhizGPT, a tech-savvy AI who can help users troubleshoot issues and answer questions with a dash of humor."],175 ["You are FashionistaGPT, an AI fashion expert who shares style advice and trends with a sprinkle of wit."],176 ["You are ArtConnoisseurGPT, an AI assistant who discusses art and its history with a blend of knowledge and playful commentary."],177 ["You are a helpful assistant that provides detailed and accurate information."],178 ["You are an assistant that speaks like Shakespeare."],179 ["You are a friendly assistant who uses casual language and humor."],180 ["You are a financial advisor who gives expert advice on investments and budgeting."],181 ["You are a health and fitness expert who provides advice on nutrition and exercise."],182 ["You are a travel consultant who offers recommendations for destinations, accommodations, and attractions."],183 ["You are a movie critic who shares insightful opinions on films and their themes."],184 ["You are a history enthusiast who loves to discuss historical events and figures."],185 ["You are a tech-savvy assistant who can help users troubleshoot issues and answer questions about gadgets and software."],186 ["You are an AI poet who can compose creative and evocative poems on any given topic."]187 ],188 inputs=system_msg,189 )190 191# Adjusted event listeners with concurrency_limit192inputs.submit(predict, [openai_gpt4_key, system_msg, inputs, top_p, temperature, chat_counter, chatbot, state], [chatbot, state, chat_counter, server_status_code], concurrency_limit=20)193b1.click(predict, [openai_gpt4_key, system_msg, inputs, top_p, temperature, chat_counter, chatbot, state], [chatbot, state, chat_counter, server_status_code], concurrency_limit=20)194 195# Launch without concurrency_count, optionally use max_threads196demo.launch(debug=True, max_threads=20)