import gradio as gr import os from openai import OpenAI # Initialize the OpenAI Client with your API key and endpoint api_key = os.environ.get("RUNPOD_API_KEY") # Ensure your API key is correctly loaded from environment variables client = OpenAI( api_key=api_key, base_url="https://api.runpod.ai/v2/vllm-k0g4c60zor9xuu/openai/v1", ) def predict(message, history=None): # Ensure history is initialized as an empty list if it's None if history is None: history = [] # Append the system role at the start if history is empty if not history: history.append(("system", "You are a history assistant, that provides the best possible answers to any historical questions asked about American History. Be helpful and specific, providing any detailed nuance needed to have a full understanding of the question.")) # Prepare messages in the format required by OpenAI history_openai_format = [] for human, assistant in history: history_openai_format.append({"role": "user", "content": human}) history_openai_format.append({"role": "assistant", "content": assistant}) history_openai_format.append({"role": "user", "content": message}) # Make the API call response_stream = client.chat.completions.create( model="ambrosfitz/llama-3-history", messages=history_openai_format, temperature=0, max_tokens=150, stream=True, ) # Accumulate response chunks to form the full message full_message = "" for chunk in response_stream: if chunk.choices[0].delta.content is not None: full_message += chunk.choices[0].delta.content yield full_message # Update history with the latest exchange history.append((message, full_message)) # Set up the Gradio interface iface = gr.Interface( fn=predict, inputs=[gr.Textbox(label="Type your question here..."), gr.State()], outputs=[gr.Textbox(), gr.State()], title="HistoryBot Chat", description="Interact with HistoryBot, a specialized assistant for American History. Ask any historical questions to get detailed and nuanced answers.", allow_flagging="never" ) iface.launch()