File size: 12,471 Bytes
437a1c7 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 | import os
import traceback
from typing import Any
import gradio as gr
from openai import OpenAI
GENERATION_MODELS = [
"gpt-4.1-mini",
"gpt-4.1",
"gpt-4o-mini",
"gpt-5.5",
]
REASONING_MODELS = [
"gpt-5.5",
"o4-mini",
"o3-mini",
]
def get_client() -> OpenAI | None:
"""
Hugging Face Spaces exposes Secrets as environment variables.
Add your OpenAI key in Space Settings as OPENAI_API_KEY.
The lowercase fallback is included only to help during local testing.
"""
api_key = os.getenv("OPENAI_API_KEY") or os.getenv("openai_api_key")
if not api_key:
return None
return OpenAI(api_key=api_key)
def extract_output_text(response: Any) -> str:
"""Robustly extract text from an OpenAI Responses API response."""
output_text = getattr(response, "output_text", None)
if output_text:
return output_text.strip()
chunks: list[str] = []
for item in getattr(response, "output", []) or []:
content = getattr(item, "content", None)
if content is None and isinstance(item, dict):
content = item.get("content", [])
for part in content or []:
if isinstance(part, dict):
text = part.get("text") or part.get("output_text")
else:
text = getattr(part, "text", None) or getattr(part, "output_text", None)
if text:
chunks.append(str(text))
return "\n".join(chunks).strip() if chunks else str(response)
def is_gpt5_family(model: str) -> bool:
"""
GPT-5 family models may reject custom sampling controls such as temperature.
To avoid the common 400 error, this app does not send those controls to GPT-5.x models.
"""
return model.strip().lower().startswith("gpt-5")
def format_settings(title: str, settings: dict[str, Any]) -> str:
lines = [f"--- {title} ---"]
for key, value in settings.items():
lines.append(f"{key}: {value}")
lines.append("------------------------\n")
return "\n".join(lines)
def run_generation(
prompt: str,
model: str,
system_message: str,
temperature: float,
top_p: float,
max_output_tokens: int,
frequency_penalty: float,
presence_penalty: float,
show_settings: bool,
) -> str:
client = get_client()
if client is None:
return (
"Missing API key.\n\n"
"In Hugging Face Spaces, go to Settings → Secrets and add:\n"
"Name: OPENAI_API_KEY\n"
"Value: your OpenAI API key"
)
if not prompt or not prompt.strip():
return "Please enter a prompt."
params: dict[str, Any] = {
"model": model,
"instructions": system_message or "You are a helpful assistant.",
"input": prompt,
"max_output_tokens": int(max_output_tokens),
}
settings_note = ""
if is_gpt5_family(model):
settings_note = (
"Note: GPT-5 family models can reject custom sampling controls. "
"Temperature, top_p, frequency_penalty, and presence_penalty were not sent.\n\n"
)
else:
params.update(
{
"temperature": float(temperature),
"top_p": float(top_p),
"frequency_penalty": float(frequency_penalty),
"presence_penalty": float(presence_penalty),
}
)
try:
response = client.responses.create(**params)
text = extract_output_text(response)
if show_settings:
settings = {
"model": model,
"system_message": system_message,
"max_output_tokens": max_output_tokens,
}
if is_gpt5_family(model):
settings.update(
{
"sampling_controls": "not sent for GPT-5 family model",
}
)
else:
settings.update(
{
"temperature": temperature,
"top_p": top_p,
"frequency_penalty": frequency_penalty,
"presence_penalty": presence_penalty,
}
)
return settings_note + format_settings("Generation Settings", settings) + text
return settings_note + text
except Exception as exc:
return (
"OpenAI API error:\n"
f"{exc}\n\n"
"Tip: If you selected a GPT-5 family model, try keeping generation controls at default "
"or use the Reasoning Controls tab.\n\n"
f"Technical details:\n{traceback.format_exc()}"
)
def run_reasoning(
prompt: str,
model: str,
reasoning_effort: str,
max_output_tokens: int,
show_settings: bool,
) -> str:
client = get_client()
if client is None:
return (
"Missing API key.\n\n"
"In Hugging Face Spaces, go to Settings → Secrets and add:\n"
"Name: OPENAI_API_KEY\n"
"Value: your OpenAI API key"
)
if not prompt or not prompt.strip():
return "Please enter a prompt."
params: dict[str, Any] = {
"model": model,
"input": prompt,
"reasoning": {"effort": reasoning_effort},
"max_output_tokens": int(max_output_tokens),
}
try:
response = client.responses.create(**params)
text = extract_output_text(response)
if show_settings:
settings = {
"model": model,
"reasoning_effort": reasoning_effort,
"max_output_tokens": max_output_tokens,
"api": "OpenAI Responses API",
}
return format_settings("Reasoning Settings", settings) + text
return text
except Exception as exc:
return (
"OpenAI API error:\n"
f"{exc}\n\n"
"Tip: Make sure your account has access to the selected model, or try another model "
"from the dropdown.\n\n"
f"Technical details:\n{traceback.format_exc()}"
)
custom_css = """
.gradio-container {
max-width: 1180px !important;
margin: auto !important;
}
#main-title {
text-align: center;
}
.output-box textarea {
font-family: ui-monospace, SFMono-Regular, Menlo, Monaco, Consolas, "Liberation Mono", monospace;
}
"""
with gr.Blocks(
title="OpenAI LLM Controls",
theme=gr.themes.Soft(),
css=custom_css,
) as demo:
gr.Markdown(
"""
# OpenAI LLM Controls
Experiment with generation settings and reasoning effort using the OpenAI Responses API.
Add your key in Hugging Face Spaces as the secret `OPENAI_API_KEY`.
""",
elem_id="main-title",
)
with gr.Tab("Generation Controls"):
gr.Markdown(
"""
Use this tab to test practical writing and completion tasks.
For GPT-5 family models, the app avoids sending custom sampling controls to prevent unsupported-parameter errors.
"""
)
with gr.Row():
with gr.Column(scale=1):
gen_prompt = gr.Textbox(
lines=7,
label="Prompt",
value="Write a short LinkedIn post explaining why business leaders should learn AI. Maximum 120 words.",
)
gen_model = gr.Dropdown(
GENERATION_MODELS,
label="Model",
value="gpt-4.1-mini",
)
system_message = gr.Textbox(
lines=3,
label="System Message",
value="You are a helpful AI instructor. Keep answers clear and practical.",
)
with gr.Accordion("Advanced Generation Settings", open=True):
temperature = gr.Slider(
minimum=0.0,
maximum=2.0,
step=0.01,
value=0.7,
label="Temperature",
)
top_p = gr.Slider(
minimum=0.0,
maximum=1.0,
step=0.01,
value=1.0,
label="Top P",
)
max_output_tokens_gen = gr.Slider(
minimum=50,
maximum=4000,
step=10,
value=300,
label="Max Output Tokens",
)
frequency_penalty = gr.Slider(
minimum=-2.0,
maximum=2.0,
step=0.01,
value=0.0,
label="Frequency Penalty",
)
presence_penalty = gr.Slider(
minimum=-2.0,
maximum=2.0,
step=0.01,
value=0.0,
label="Presence Penalty",
)
show_settings_gen = gr.Checkbox(value=True, label="Show Settings")
gen_button = gr.Button("Generate", variant="primary")
with gr.Column(scale=1):
gen_output = gr.Textbox(
lines=22,
label="Output",
elem_classes=["output-box"],
show_copy_button=True,
)
gen_button.click(
fn=run_generation,
inputs=[
gen_prompt,
gen_model,
system_message,
temperature,
top_p,
max_output_tokens_gen,
frequency_penalty,
presence_penalty,
show_settings_gen,
],
outputs=gen_output,
)
with gr.Tab("Reasoning Controls"):
gr.Markdown(
"""
Use this tab for analysis, recommendations, technical trade-offs, planning, and decision-making tasks.
"""
)
with gr.Row():
with gr.Column(scale=1):
reason_prompt = gr.Textbox(
lines=9,
label="Prompt",
value=(
"A telecom company wants to build an AI customer support assistant. "
"They have 50,000 past support tickets, a FAQ website, billing policies, "
"and a small developer team. Should they start with: "
"1. Simple prompt-based chatbot 2. RAG chatbot 3. Fine-tuning "
"4. Agent with tools. Give a practical recommendation with trade-offs."
),
)
reason_model = gr.Dropdown(
REASONING_MODELS,
label="Model",
value="gpt-5.5",
)
reasoning_effort = gr.Radio(
["low", "medium", "high"],
label="Reasoning Effort",
value="medium",
)
max_output_tokens_reason = gr.Slider(
minimum=100,
maximum=8000,
step=50,
value=900,
label="Max Output Tokens",
)
show_settings_reason = gr.Checkbox(value=True, label="Show Settings")
reason_button = gr.Button("Reason", variant="primary")
with gr.Column(scale=1):
reason_output = gr.Textbox(
lines=22,
label="Output",
elem_classes=["output-box"],
show_copy_button=True,
)
reason_button.click(
fn=run_reasoning,
inputs=[
reason_prompt,
reason_model,
reasoning_effort,
max_output_tokens_reason,
show_settings_reason,
],
outputs=reason_output,
)
if __name__ == "__main__":
demo.queue()
demo.launch(
server_name="0.0.0.0",
server_port=int(os.getenv("PORT", "7860")),
)
|