Transformers
English
code
Inference Endpoints
Canstralian's picture
Create main.py
de09b1c verified
import gradio as gr
from src.utils.dataset_loader import load_dataset
from src.models.model import load_model, run_inference
# Load dataset
df = load_dataset()
# List of available models (example)
models = ["Canstralian/text2shellcommands", "Canstralian/RabbitRedux", "Canstralian/CySec_Known_Exploit_Analyzer"]
prompt_ids = df.index.tolist() if df is not None else []
# Function to simulate conversation with model selection
def chat_interface(user_input, selected_model, prompt_id=None):
if df is not None and prompt_id is not None:
prompt = df.iloc[prompt_id]["prompt_text"] # Replace with the actual column name
# Run inference on the selected model
response = run_inference(user_input, selected_model, prompt)
else:
response = f"[{selected_model}] says: You entered '{user_input}'. This is a simulated response."
return response
# Gradio Interface
with gr.Blocks(css="./static/styles.css") as demo:
with gr.Row():
gr.Markdown("### Retro Hacker Chat with Debugging Prompts", elem_classes="retro-terminal")
with gr.Row():
user_input = gr.Textbox(
label="Enter your message:",
placeholder="Type your message here...",
elem_classes="retro-terminal"
)
model_selector = gr.Dropdown(
choices=models,
label="Select Model",
value=models[0],
elem_classes="retro-terminal"
)
if prompt_ids:
prompt_selector = gr.Dropdown(
choices=prompt_ids,
label="Select Debugging Prompt ID",
value=prompt_ids[0],
elem_classes="retro-terminal"
)
else:
prompt_selector = None
with gr.Row():
response_box = gr.Textbox(
label="Model Response:",
placeholder="The model's response will appear here...",
elem_classes="retro-terminal"
)
with gr.Row():
send_button = gr.Button("Send", elem_classes="retro-terminal")
# Link input and output
if prompt_selector:
send_button.click(chat_interface, inputs=[user_input, model_selector, prompt_selector], outputs=response_box)
else:
send_button.click(chat_interface, inputs=[user_input, model_selector], outputs=response_box)
# Launch the interface
demo.launch()