import os from dotenv import find_dotenv, load_dotenv import streamlit as st from groq import Groq # Load environment variables load_dotenv(find_dotenv()) # Set up Streamlit page configuration st.set_page_config( page_icon="📃", layout="wide", page_title="Groq & LLaMA3x Chat Bot" ) # App Title st.title("Groq Chat with LLaMA3x") # Initialize the Groq client using the API key from the environment variables client = Groq(api_key=os.environ.get("GROQ_API_KEY")) # Cache the model fetching function to improve performance @st.cache_data def fetch_available_models(): """ Fetches the available models from the Groq API. Returns a list of models or an empty list if there's an error. """ try: models_response = client.models.list() return models_response.data except Exception as e: st.error(f"Error fetching models: {e}") return [] # Load available models and filter them available_models = fetch_available_models() filtered_models = [ model for model in available_models if model.id.startswith('llama-3') ] # Prepare a dictionary of model metadata models = { model.id: { "name": model.id, "tokens": 4000, "developer": model.owned_by, } for model in filtered_models } # Initialize session state variables if "messages" not in st.session_state: st.session_state.messages = [] if "selected_model" not in st.session_state: st.session_state.selected_model = None # Sidebar: Controls with st.sidebar: # Powered by Groq logo st.markdown( """ Powered by Groq for fast inference. """, unsafe_allow_html=True ) st.markdown("---") # Define a function to clear messages when the model changes def reset_chat_on_model_change(): st.session_state.messages = [] # Model selection dropdown if models: model_option = st.selectbox( "Choose a model:", options=list(models.keys()), format_func=lambda x: f"{models[x]['name']} ({models[x]['developer']})", on_change=reset_chat_on_model_change, # Reset chat when model changes ) else: st.warning("No available models to select.") model_option = None # Token limit slider if models: max_tokens_range = models[model_option]["tokens"] max_tokens = st.slider( "Max Tokens:", min_value=200, max_value=max_tokens_range, value=max(100, int(max_tokens_range * 0.5)), step=256, help=f"Adjust the maximum number of tokens for the response. Maximum for the selected model: {max_tokens_range}" ) else: max_tokens = 200 # Additional options stream_mode = st.checkbox("Enable Streaming", value=True) # Button to clear the chat if st.button("Clear Chat"): st.session_state.messages = [] st.markdown("### Usage Summary") usage_box = st.empty() # Disclaimer st.markdown( """ ----- ⚠️ **Important:** *The responses provided by this application are generated automatically using an AI model. Users are responsible for verifying the accuracy of the information before relying on it. Always cross-check facts and data for critical decisions.* """ ) # Main Chat Interface st.markdown("### Chat Interface") # Display the chat history for message in st.session_state.messages: avatar = "🔋" if message["role"] == "assistant" else "🧑‍💻" with st.chat_message(message["role"], avatar=avatar): st.markdown(message["content"]) # Capture user input user_input = st.chat_input("Enter your message here...") if user_input: # Append the user input to the session state st.session_state.messages.append({"role": "user", "content": user_input}) with st.chat_message("user", avatar="🧑‍💻"): st.markdown(user_input) # Generate a response using the selected model try: full_response = "" usage_summary = "" if stream_mode: # Generate a response with streaming enabled chat_completion = client.chat.completions.create( model=model_option, messages=[ {"role": m["role"], "content": m["content"]} for m in st.session_state.messages ], max_tokens=max_tokens, stream=True ) with st.chat_message("assistant", avatar="🔋"): response_placeholder = st.empty() for chunk in chat_completion: if chunk.choices[0].delta.content: full_response += chunk.choices[0].delta.content response_placeholder.markdown(full_response) else: # Generate a response without streaming chat_completion = client.chat.completions.create( model=model_option, messages=[ {"role": m["role"], "content": m["content"]} for m in st.session_state.messages ], max_tokens=max_tokens, stream=False ) response = chat_completion.choices[0].message.content usage_data = chat_completion.usage with st.chat_message("assistant", avatar="🔋"): st.markdown(response) full_response = response if usage_data: usage_summary = ( f"**Token Usage:**\n" f"- Prompt Tokens: {usage_data.prompt_tokens}\n" f"- Response Tokens: {usage_data.completion_tokens}\n" f"- Total Tokens: {usage_data.total_tokens}\n\n" f"**Timings:**\n" f"- Prompt Time: {round(usage_data.prompt_time,5)} secs\n" f"- Response Time: {round(usage_data.completion_time,5)} secs\n" f"- Total Time: {round(usage_data.total_time,5)} secs" ) if usage_summary: usage_box.markdown(usage_summary) # Append the assistant's response to the session state st.session_state.messages.append( {"role": "assistant", "content": full_response} ) except Exception as e: st.error(f"Error generating the response: {e}")