#بسم الله الرحمن الرحيم
import gradio as gr
import asyncio
import base64
from client import run_fistal
import asyncio
import os
# from dotenv import load_dotenv
# load_dotenv()
REQUIRED_SECRETS = [
"GOOGLE_API_KEY_1",
"GOOGLE_API_KEY_2",
"GOOGLE_API_KEY_3",
"GROQ_API_KEY",
"GEMINI_API_KEY",
"HF_TOKEN",
"MODAL_TOKEN_ID",
"MODAL_TOKEN_SECRET"
]
missing = [s for s in REQUIRED_SECRETS if not os.getenv(s)]
if missing:
raise ValueError(f"Missing secrets in HF Space: {', '.join(missing)}\nAdd them in Settings → Variables and secrets")
def image_to_base64(filepath):
try:
with open(filepath, "rb") as image_file:
encoded_string = base64.b64encode(image_file.read()).decode('utf-8')
mime_type = "image/jpeg" if filepath.lower().endswith((".jpg", ".jpeg")) else "image/png"
return f"data:{mime_type};base64,{encoded_string}"
except FileNotFoundError:
print(f"Error: Image file not found at {filepath}")
return ""
image_data_url = image_to_base64("static/new.jpg")
full_img = image_to_base64("static/fullnew.jpg")
css=f"""
.gradio-container {{
background: url('{full_img}') !important;
background-size: cover !important;
}}
.gradio-container .block {{
background-color: none !important;
}}
.gradio-container .wrap {{
background-color: none !important;
border: !important;
box-shadow: none !important;
outline: none !important;
}}
.features-box {{
padding: 10px;
color: white !important;
background-color: #white !important;
}}
.features-box .block {{
border: blue 1px !important;
background-color: yellow !important;
}}
#tuner {{
background: linear-gradient(to right, #008DDA, #6A1AAB, #C71585, #F56C40) !important;
color: white !important;
margin-top: 5px;
}}
#flow {{
padding: 8px !important;
color: white !important;
}}
#flow .markdown-text {{
color: white !important;
}}
.drop li {{
background-color: #bcb9cf !important;
color: black !important;
}}
.drop input {{
background-color: #bcb9cf !important;
background-size: cover !important;
color: black !important;
border: none !important;
padding: 6px 10px !important;
border-radius: 4px !important;
}}
.out {{
padding: 10px !important;
font-size: 16px !important;
color: white !important;
background: linear-gradient(90deg, rgba(102,126,234,0.3), rgba(106,26,180,0.3), rgba(245,108,64,0.3)) !important;
border-radius: 10px !important;
}}
.log-container {{
max-height: 600px !important;
overflow-y: auto !important;
background: rgba(14, 15, 15, 0.5) !important;
border-radius: 10px !important;
padding: 20px !important;
border: 1px solid #3a3a3a !important;
}}
#stat {{
min-height: 60px !important;
}}
#stat input, #stat textarea {{
padding: 12px 10px !important;
line-height: 1.5 !important;
color: black !important;
min-height: 60px !important;
height: auto !important;
display: flex !important;
align-items: center !important;
}}
#stat .wrap {{
min-height: 60px !important;
}}
.mod {{
background: linear-gradient(to right, #008DDA, #6A1AAB, #C71585, #F56C40) !important;
color: white !important;
}}
.log-container::-webkit-scrollbar {{
width: 10px;
}}
.log-container::-webkit-scrollbar-track {{
background: #2a2a2a;
border-radius: 5px;
}}
.log-container::-webkit-scrollbar-thumb {{
background: linear-gradient(to bottom, #008DDA, #6A1AAB, #C71585, #F56C40);
border-radius: 5px;
}}
.log-container::-webkit-scrollbar-thumb:hover {{
background: linear-gradient(to bottom, #0099ee, #7722bb, #dd1595, #ff7750);
}}
#copy-btn {{
background: linear-gradient(to right, #008DDA, #6A1AAB, #C71585, #F56C40) !important;
color: white !important;
margin-top: 10px !important;
}}
:root, .gradio-container * {{
--block-background-fill: #0e0f0f !important;
--panel-background-fill: none !important;
--input-background-fill: #bcb9cf !important;
--color-background-primary: #0e0f0f !important;
--block-border-width: 0px !important;
--block-border-color: #27272a !important;
--panel-border-width: 0px !important;
--input-text-color: #000000 !important;
--input-placeholder-color: #27272a !important;
--panel-border-color: linear-gradient(to right, #008DDA, #6A1AAB, #C71585, #F56C40) !important;
--neutral-50: #27272a !important;
}}
"""
def app():
with gr.Blocks(title="Fistal AI 🚀") as demo:
# Header section with background image
gr.HTML(f"""
Fistal AI 🚀
Seamlessly fine-tune LLMs with an Agentic AI powered by MCP, Modal, and Unsloth.
""")
gr.HTML("""
How does Fistal AI revolutionize LLM fine-tuning?
🧩 MCP + Agentic AI :
Automates workflows using MCP while running autonomous data and training pipelines through Agentic AI.
🦥 Unsloth :
Speeds up training with optimized kernels and memory-efficient 4-bit fine-tuning.
⚡ Modal Labs (with Volumes) :
Provides serverless GPU compute with persistent volumes for fast scaling and reproducible experiments.
💾 Hugging Face + Gradio :
Model stored securely in HF repositories, Gradio's polished UI makes Fistal AI better.
🔑 Gemini API Key :
Enables secure access to the Gemini API that performs model orchestration and automated workflows.
""")
gr.HTML("""
🚀 Start Fine-Tuning Now
Add your dataset topic, task type, number of samples, and your preferred Unsloth model.
Then sit back and watch Fistal AI automatically build datasets, fine-tune your LLMs, and deliver results like magic.
Note: The process may take 30-45 minutes, depending on the number of samples and model chosen.
""")
with gr.Group():
with gr.Row():
topic = gr.Textbox(label="📚 Dataset topic", placeholder="Python Questions, Return policy FAQS...")
samples = gr.Slider(label="📊 Number of samples", minimum=0, maximum=2000, step=5, value=1000)
task_type = gr.Dropdown(
label="🎯 Task Type",
choices=["text-generation", "summarization", "classification", "question-answering"], elem_classes="drop"
)
model_name = gr.Dropdown(
label="🤖 Model to Fine-tune",
choices=[
"unsloth/Llama-3.2-1B-Instruct-bnb-4bit",
"unsloth/Phi-3-mini-4k-instruct",
"unsloth/Phi-3-medium-4k-instruct",
"unsloth/Llama-3.2-3B-Instruct-bnb-4bit",
"unsloth/Qwen2.5-3B-Instruct-bnb-4bit",
"unsloth/Qwen2.5-1.5B-Instruct-bnb-4bit",
"unsloth/Qwen2.5-0.5B-Instruct-bnb-4bit",
"unsloth/Qwen2.5-Coder-3B-Instruct-bnb-4bit",
"unsloth/gemma-2-2b-it-bnb-4bit",
"unsloth/SmolLM2-1.7B-Instruct-bnb-4bit",
"unsloth/Phi-3.5-mini-instruct-bnb-4bit",
"unsloth/Granite-3.0-2b-instruct-bnb-4bit",
"unsloth/granite-4.0-h-1b-bnb-4bit"
], elem_classes="drop"
)
tuner = gr.Button("🚀 Start Finetuning", size="lg", elem_id="tuner")
gr.Markdown("""## 🔀 Agent Activity Flow""", elem_id="flow")
status = gr.Textbox(label="Status", value="Ready to start...", interactive=False, elem_id="stat")
with gr.Group(elem_classes="log-container"):
output = gr.Markdown(label="Output Log:", value="", elem_classes="out")
eval_report_storage = gr.Textbox(visible=False) # Hidden storage
copy_btn = gr.Button("Finetuning completed 🚀", visible=False, elem_id="copy-btn")
async def run_workflow(dataset_topic, samples, model, task):
output_log = "## Under the Hood\n\n"
output_log += "📋 **Configuration:**\n\n"
output_log += f" • Topic: {dataset_topic}\n\n"
output_log += f" • Samples: {samples}\n\n"
output_log += f" • Model: {model}\n\n"
output_log += f" • Task: {task}\n\n"
yield ("Starting workflow...", output_log, "", gr.Button(visible=False))
try:
in_eval_report = False
eval_report_buffer = ""
async for chunk in run_fistal(
dataset_topic=dataset_topic,
num_samples=samples,
model_name=model,
task_type=task
):
if "evaluating" in str(chunk).lower() or "llm_as_judge" in str(chunk).lower():
in_eval_report = True
if in_eval_report:
eval_report_buffer += str(chunk)
else:
output_log += str(chunk)
import re
urls = re.findall(r'https://huggingface\.co/[^\s\)]+', output_log + eval_report_buffer)
model_url = urls[0] if urls else ""
model_url = model_url.rstrip('.')
model_url = re.sub(r'[^a-zA-Z0-9:/._-].*$', '', model_url)
yield ("🟡 Processing...", output_log + eval_report_buffer, eval_report_buffer, gr.Button(visible=False))
await asyncio.sleep(0.1)
final_output = output_log
if eval_report_buffer:
final_output += "📊 **EVALUATION REPORT**\n\n"
final_output += eval_report_buffer
final_output += "\n\n✨ **Fistal AI has completed the process!**"
yield ("🟢 Complete!", final_output, eval_report_buffer, gr.Button(visible=True))
except Exception as e:
import traceback
error_log = output_log + f"\n\n❌ **ERROR:**\n```\n{str(e)}\n{traceback.format_exc()}\n```"
yield ("🔴 Error", error_log, "",gr.Button(visible=False))
tuner.click(
run_workflow,
[topic, samples, model_name, task_type],
[status, output, eval_report_storage, copy_btn]
)
def copy_report(report_text):
if report_text and report_text.strip():
return report_text
return ""
copy_btn.click(
copy_report,
inputs=[eval_report_storage],
outputs=[],
js="""
(text) => {
if (text && text.trim().length > 0) {
navigator.clipboard.writeText(text);
alert('✅ Evaluation report copied to clipboard!');
} else {
alert('Fistal has completed the process.');
}
}
"""
)
return demo
if __name__ == "__main__":
app().launch(theme= gr.themes.Ocean(), css=css) 
