Upload app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,142 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
from openai import OpenAI, OpenAIError
|
| 3 |
+
|
| 4 |
+
# Global message history
|
| 5 |
+
history = []
|
| 6 |
+
|
| 7 |
+
# Main chatbot function
|
| 8 |
+
# Now accepts api_key provided by the user
|
| 9 |
+
def chatbot(user_input, api_key, temperature, top_p, max_tokens):
|
| 10 |
+
global history
|
| 11 |
+
# Ignore empty input
|
| 12 |
+
if not user_input:
|
| 13 |
+
return history, ''
|
| 14 |
+
|
| 15 |
+
# Instantiate OpenAI/NVIDIA client with user-provided key
|
| 16 |
+
client = OpenAI(
|
| 17 |
+
base_url="https://integrate.api.nvidia.com/v1",
|
| 18 |
+
api_key=api_key
|
| 19 |
+
)
|
| 20 |
+
|
| 21 |
+
# Add user message to history
|
| 22 |
+
history.append({"role": "user", "content": user_input})
|
| 23 |
+
# Ensure system message at start
|
| 24 |
+
if len(history) == 1:
|
| 25 |
+
history.insert(0, {
|
| 26 |
+
"role": "system",
|
| 27 |
+
"content": "You are a helpful assistant that explains complex topics clearly."
|
| 28 |
+
})
|
| 29 |
+
|
| 30 |
+
try:
|
| 31 |
+
# Stream response
|
| 32 |
+
response_stream = client.chat.completions.create(
|
| 33 |
+
model="meta/llama3-8b-instruct",
|
| 34 |
+
messages=history,
|
| 35 |
+
temperature=temperature,
|
| 36 |
+
top_p=top_p,
|
| 37 |
+
max_tokens=max_tokens,
|
| 38 |
+
stream=True
|
| 39 |
+
)
|
| 40 |
+
assistant_reply = ""
|
| 41 |
+
for chunk in response_stream:
|
| 42 |
+
delta = chunk.choices[0].delta
|
| 43 |
+
if delta and delta.content:
|
| 44 |
+
assistant_reply += delta.content
|
| 45 |
+
except OpenAIError as e:
|
| 46 |
+
assistant_reply = f"⚠️ API Error: {e.__class__.__name__}: {e}"
|
| 47 |
+
|
| 48 |
+
# Store assistant response and prepare display history
|
| 49 |
+
history.append({"role": "assistant", "content": assistant_reply})
|
| 50 |
+
display = [
|
| 51 |
+
{"role": msg["role"], "content": msg["content"]}
|
| 52 |
+
for msg in history if msg["role"] in ["user", "assistant"]
|
| 53 |
+
]
|
| 54 |
+
return display, ''
|
| 55 |
+
|
| 56 |
+
# Clear conversation history
|
| 57 |
+
def clear_history():
|
| 58 |
+
global history
|
| 59 |
+
history = []
|
| 60 |
+
return [], ''
|
| 61 |
+
|
| 62 |
+
# Custom CSS for cleaner, centered layout
|
| 63 |
+
custom_css = r"""
|
| 64 |
+
#header {
|
| 65 |
+
text-align: center;
|
| 66 |
+
margin-bottom: 1rem;
|
| 67 |
+
}
|
| 68 |
+
#title {
|
| 69 |
+
font-size: 2rem;
|
| 70 |
+
margin: 0;
|
| 71 |
+
}
|
| 72 |
+
#chatbot {
|
| 73 |
+
border: none;
|
| 74 |
+
background-color: #f9f9f9;
|
| 75 |
+
}
|
| 76 |
+
footer {
|
| 77 |
+
visibility: hidden;
|
| 78 |
+
}
|
| 79 |
+
"""
|
| 80 |
+
|
| 81 |
+
with gr.Blocks(css=custom_css, theme=gr.themes.Base()) as demo:
|
| 82 |
+
# Centered header
|
| 83 |
+
with gr.Row(elem_id="header"):
|
| 84 |
+
gr.Markdown("<h1 id='title'>🌐 GeoChat</h1>")
|
| 85 |
+
|
| 86 |
+
# Main layout: chat + settings
|
| 87 |
+
with gr.Row():
|
| 88 |
+
with gr.Column(scale=4, min_width=600):
|
| 89 |
+
chatbot_ui = gr.Chatbot(elem_id="chatbot", label="Assistant", height=500, type="messages")
|
| 90 |
+
with gr.Row():
|
| 91 |
+
txt = gr.Textbox(
|
| 92 |
+
placeholder="Type your question and press Send...",
|
| 93 |
+
show_label=False,
|
| 94 |
+
lines=2
|
| 95 |
+
)
|
| 96 |
+
btn = gr.Button("Send")
|
| 97 |
+
with gr.Row():
|
| 98 |
+
clear_btn = gr.Button("Clear")
|
| 99 |
+
|
| 100 |
+
with gr.Column(scale=1, min_width=200):
|
| 101 |
+
gr.Markdown(
|
| 102 |
+
"""
|
| 103 |
+
### 🔑 API Key
|
| 104 |
+
Get your NVIDIA API Key at [NVIDIA NGC API Keys](https://org.ngc.nvidia.com/setup/api-keys)
|
| 105 |
+
"""
|
| 106 |
+
)
|
| 107 |
+
api_key_input = gr.Textbox(
|
| 108 |
+
label="NVIDIA API Key",
|
| 109 |
+
placeholder="Enter your key here",
|
| 110 |
+
type="password",
|
| 111 |
+
show_label=True
|
| 112 |
+
)
|
| 113 |
+
gr.Markdown("### ⚙️ Settings")
|
| 114 |
+
temp_slider = gr.Slider(0, 1, value=0.6, label="Temperature")
|
| 115 |
+
top_p_slider = gr.Slider(0, 1, value=0.95, label="Top-p")
|
| 116 |
+
max_tokens_slider = gr.Slider(64, 2048, value=1024, step=64, label="Max Tokens")
|
| 117 |
+
gr.Markdown(
|
| 118 |
+
"""
|
| 119 |
+
**Temperature:** controls the randomness of the responses; lower values make output more deterministic.
|
| 120 |
+
|
| 121 |
+
**Top-p:** sets the cumulative probability for nucleus sampling; lower values focus on fewer tokens.
|
| 122 |
+
|
| 123 |
+
**Max Tokens:** maximum number of tokens the model can generate in the response.
|
| 124 |
+
"""
|
| 125 |
+
)
|
| 126 |
+
|
| 127 |
+
# Interaction events
|
| 128 |
+
btn.click(
|
| 129 |
+
fn=chatbot,
|
| 130 |
+
inputs=[txt, api_key_input, temp_slider, top_p_slider, max_tokens_slider],
|
| 131 |
+
outputs=[chatbot_ui, txt]
|
| 132 |
+
)
|
| 133 |
+
txt.submit(
|
| 134 |
+
fn=chatbot,
|
| 135 |
+
inputs=[txt, api_key_input, temp_slider, top_p_slider, max_tokens_slider],
|
| 136 |
+
outputs=[chatbot_ui, txt]
|
| 137 |
+
)
|
| 138 |
+
clear_btn.click(fn=clear_history, outputs=[chatbot_ui, txt])
|
| 139 |
+
|
| 140 |
+
# Run locally and open browser automatically
|
| 141 |
+
if __name__ == "__main__":
|
| 142 |
+
demo.launch()
|