Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -31,13 +31,19 @@ class ModelRegistry:
|
|
| 31 |
self.groq_models = self._fetch_groq_models()
|
| 32 |
|
| 33 |
def _fetch_groq_models(self) -> Dict[str, str]:
|
| 34 |
-
"""Fetch available Groq models"""
|
| 35 |
try:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
headers = {
|
| 37 |
-
"Authorization": f"Bearer {
|
| 38 |
"Content-Type": "application/json"
|
| 39 |
}
|
| 40 |
response = requests.get("https://api.groq.com/openai/v1/models", headers=headers)
|
|
|
|
| 41 |
if response.status_code == 200:
|
| 42 |
models = response.json().get("data", [])
|
| 43 |
return {model["id"]: model["id"] for model in models}
|
|
@@ -142,6 +148,38 @@ def build_prompts(snippets: List[str], prompt_instruction: str, custom_prompt: O
|
|
| 142 |
|
| 143 |
return "\n\n".join(prompts)
|
| 144 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 145 |
def send_to_hf_inference(prompt: str, model_name: str, api_key: str) -> str:
|
| 146 |
"""Send prompt to HuggingFace using Inference API"""
|
| 147 |
try:
|
|
@@ -179,26 +217,27 @@ def send_to_groq(prompt: str, model_name: str, api_key: str) -> str:
|
|
| 179 |
return f"Error with Groq API: {e}"
|
| 180 |
|
| 181 |
def copy_to_clipboard(text: str) -> str:
|
| 182 |
-
"""Copy text to clipboard"""
|
| 183 |
-
return "
|
| 184 |
-
|
| 185 |
-
|
| 186 |
-
|
| 187 |
-
|
| 188 |
-
|
| 189 |
-
|
| 190 |
-
|
| 191 |
-
|
| 192 |
-
|
| 193 |
-
|
|
|
|
| 194 |
try:
|
| 195 |
if not pdf:
|
| 196 |
-
return "Please upload a PDF file.", "",
|
| 197 |
|
| 198 |
# Extract text
|
| 199 |
text = extract_text_from_pdf(pdf.name)
|
| 200 |
if text.startswith("Error"):
|
| 201 |
-
return text, "",
|
| 202 |
|
| 203 |
# Format content
|
| 204 |
formatted_text = format_content(text, fmt)
|
|
@@ -211,42 +250,17 @@ def process_pdf(pdf, fmt, ctx_size, snippet_num, prompt, model_selection,
|
|
| 211 |
full_prompt = build_prompts(snippets, default_prompt, prompt, snippet_num)
|
| 212 |
|
| 213 |
if isinstance(full_prompt, str) and full_prompt.startswith("Error"):
|
| 214 |
-
return full_prompt, "",
|
| 215 |
-
|
| 216 |
-
# Process with selected model
|
| 217 |
-
if model_selection == "HuggingFace Inference":
|
| 218 |
-
if not hf_api_key:
|
| 219 |
-
return "HuggingFace API key required.", full_prompt, "", []
|
| 220 |
-
|
| 221 |
-
model_id = hf_custom_model if hf_model_choice == "Custom Model" else model_registry.hf_models[hf_model_choice]
|
| 222 |
-
summary = send_to_hf_inference(full_prompt, model_id, hf_api_key)
|
| 223 |
-
|
| 224 |
-
elif model_selection == "Groq API":
|
| 225 |
-
if not groq_api_key:
|
| 226 |
-
return "Groq API key required.", full_prompt, "", []
|
| 227 |
-
|
| 228 |
-
summary = send_to_groq(full_prompt, groq_model_choice, groq_api_key)
|
| 229 |
-
|
| 230 |
-
else: # OpenAI ChatGPT
|
| 231 |
-
summary = "Please use the Copy Prompt button and paste into ChatGPT."
|
| 232 |
-
|
| 233 |
-
# Save files for download
|
| 234 |
-
files_to_download = []
|
| 235 |
|
|
|
|
| 236 |
with tempfile.NamedTemporaryFile(delete=False, mode='w', suffix='.txt') as prompt_file:
|
| 237 |
prompt_file.write(full_prompt)
|
| 238 |
-
files_to_download.append(prompt_file.name)
|
| 239 |
-
|
| 240 |
-
if summary != "Please use the Copy Prompt button and paste into ChatGPT.":
|
| 241 |
-
with tempfile.NamedTemporaryFile(delete=False, mode='w', suffix='.txt') as summary_file:
|
| 242 |
-
summary_file.write(summary)
|
| 243 |
-
files_to_download.append(summary_file.name)
|
| 244 |
|
| 245 |
-
return "
|
| 246 |
|
| 247 |
except Exception as e:
|
| 248 |
logging.error(f"Error processing PDF: {e}")
|
| 249 |
-
return f"Error processing PDF: {str(e)}", "",
|
| 250 |
|
| 251 |
# Main Interface
|
| 252 |
with gr.Blocks(theme=gr.themes.Default()) as demo:
|
|
@@ -273,12 +287,18 @@ with gr.Blocks(theme=gr.themes.Default()) as demo:
|
|
| 273 |
label="π Output Format"
|
| 274 |
)
|
| 275 |
|
| 276 |
-
gr.Markdown("### Context
|
| 277 |
with gr.Row():
|
| 278 |
-
context_buttons = []
|
| 279 |
for size_name, size_value in CONTEXT_SIZES.items():
|
| 280 |
-
|
| 281 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 282 |
|
| 283 |
context_size = gr.Slider(
|
| 284 |
minimum=1000,
|
|
@@ -334,6 +354,13 @@ with gr.Blocks(theme=gr.themes.Default()) as demo:
|
|
| 334 |
type="password"
|
| 335 |
)
|
| 336 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 337 |
# Right Column - Output
|
| 338 |
with gr.Column(scale=1):
|
| 339 |
process_button = gr.Button("π Process PDF", variant="primary")
|
|
@@ -370,7 +397,8 @@ with gr.Blocks(theme=gr.themes.Default()) as demo:
|
|
| 370 |
def toggle_model_options(choice):
|
| 371 |
return (
|
| 372 |
gr.update(visible=choice == "HuggingFace Inference"),
|
| 373 |
-
gr.update(visible=choice == "Groq API")
|
|
|
|
| 374 |
)
|
| 375 |
|
| 376 |
def refresh_groq_models_list():
|
|
@@ -384,7 +412,7 @@ with gr.Blocks(theme=gr.themes.Default()) as demo:
|
|
| 384 |
model_choice.change(
|
| 385 |
toggle_model_options,
|
| 386 |
inputs=[model_choice],
|
| 387 |
-
outputs=[hf_options, groq_options]
|
| 388 |
)
|
| 389 |
|
| 390 |
for btn, size_value in context_buttons:
|
|
@@ -412,17 +440,30 @@ with gr.Blocks(theme=gr.themes.Default()) as demo:
|
|
| 412 |
format_type,
|
| 413 |
context_size,
|
| 414 |
snippet_number,
|
| 415 |
-
custom_prompt
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 416 |
model_choice,
|
| 417 |
hf_model,
|
| 418 |
hf_custom_model,
|
| 419 |
hf_api_key,
|
| 420 |
groq_model,
|
| 421 |
-
groq_api_key
|
|
|
|
| 422 |
],
|
| 423 |
outputs=[
|
| 424 |
-
progress_status,
|
| 425 |
-
generated_prompt,
|
| 426 |
summary_output,
|
| 427 |
download_files
|
| 428 |
]
|
|
@@ -451,22 +492,13 @@ with gr.Blocks(theme=gr.themes.Default()) as demo:
|
|
| 451 |
1. Upload a PDF document
|
| 452 |
2. Choose output format and context window size
|
| 453 |
3. Select snippet number (default: 1) or enter custom prompt
|
| 454 |
-
4. Select your preferred model:
|
| 455 |
- OpenAI ChatGPT: Manual copy/paste workflow
|
| 456 |
- HuggingFace Inference: Direct API integration
|
| 457 |
- Groq API: High-performance inference
|
| 458 |
5. Click 'Process PDF' to generate summary
|
| 459 |
-
6. Use 'Copy Prompt' and 'Open ChatGPT' for manual processing
|
| 460 |
7. Download generated files as needed
|
| 461 |
-
|
| 462 |
-
### βοΈ Features:
|
| 463 |
-
- Support for multiple PDF formats
|
| 464 |
-
- Flexible text formatting options
|
| 465 |
-
- Predefined context window sizes (4K to 200K)
|
| 466 |
-
- Multiple model integrations
|
| 467 |
-
- Copy to clipboard functionality
|
| 468 |
-
- Direct ChatGPT integration
|
| 469 |
-
- Downloadable outputs
|
| 470 |
""")
|
| 471 |
|
| 472 |
# Launch the interface
|
|
|
|
| 31 |
self.groq_models = self._fetch_groq_models()
|
| 32 |
|
| 33 |
def _fetch_groq_models(self) -> Dict[str, str]:
|
| 34 |
+
"""Fetch available Groq models with proper error handling"""
|
| 35 |
try:
|
| 36 |
+
groq_api_key = os.getenv('GROQ_API_KEY')
|
| 37 |
+
if not groq_api_key:
|
| 38 |
+
logging.warning("No GROQ_API_KEY found in environment")
|
| 39 |
+
return self._get_default_groq_models()
|
| 40 |
+
|
| 41 |
headers = {
|
| 42 |
+
"Authorization": f"Bearer {groq_api_key}",
|
| 43 |
"Content-Type": "application/json"
|
| 44 |
}
|
| 45 |
response = requests.get("https://api.groq.com/openai/v1/models", headers=headers)
|
| 46 |
+
|
| 47 |
if response.status_code == 200:
|
| 48 |
models = response.json().get("data", [])
|
| 49 |
return {model["id"]: model["id"] for model in models}
|
|
|
|
| 148 |
|
| 149 |
return "\n\n".join(prompts)
|
| 150 |
|
| 151 |
+
def send_to_model(prompt, model_selection, hf_model_choice, hf_custom_model, hf_api_key,
|
| 152 |
+
groq_model_choice, groq_api_key, openai_api_key):
|
| 153 |
+
"""Send prompt to selected model"""
|
| 154 |
+
try:
|
| 155 |
+
if model_selection == "HuggingFace Inference":
|
| 156 |
+
if not hf_api_key:
|
| 157 |
+
return "HuggingFace API key required.", []
|
| 158 |
+
|
| 159 |
+
model_id = hf_custom_model if hf_model_choice == "Custom Model" else model_registry.hf_models[hf_model_choice]
|
| 160 |
+
summary = send_to_hf_inference(prompt, model_id, hf_api_key)
|
| 161 |
+
|
| 162 |
+
elif model_selection == "Groq API":
|
| 163 |
+
if not groq_api_key:
|
| 164 |
+
return "Groq API key required.", []
|
| 165 |
+
|
| 166 |
+
summary = send_to_groq(prompt, groq_model_choice, groq_api_key)
|
| 167 |
+
|
| 168 |
+
elif model_selection == "OpenAI ChatGPT":
|
| 169 |
+
if not openai_api_key:
|
| 170 |
+
return "OpenAI API key required.", []
|
| 171 |
+
# Implement OpenAI API call here
|
| 172 |
+
|
| 173 |
+
# Save summary for download
|
| 174 |
+
with tempfile.NamedTemporaryFile(delete=False, mode='w', suffix='.txt') as summary_file:
|
| 175 |
+
summary_file.write(summary)
|
| 176 |
+
|
| 177 |
+
return summary, [summary_file.name]
|
| 178 |
+
|
| 179 |
+
except Exception as e:
|
| 180 |
+
logging.error(f"Error sending to model: {e}")
|
| 181 |
+
return f"Error sending to model: {str(e)}", []
|
| 182 |
+
|
| 183 |
def send_to_hf_inference(prompt: str, model_name: str, api_key: str) -> str:
|
| 184 |
"""Send prompt to HuggingFace using Inference API"""
|
| 185 |
try:
|
|
|
|
| 217 |
return f"Error with Groq API: {e}"
|
| 218 |
|
| 219 |
def copy_to_clipboard(text: str) -> str:
|
| 220 |
+
"""Copy text to clipboard using JavaScript"""
|
| 221 |
+
return """
|
| 222 |
+
navigator.clipboard.writeText(text)
|
| 223 |
+
.then(() => gradioApp().querySelector('#progress_status').value = 'Copied to clipboard!')
|
| 224 |
+
.catch(() => gradioApp().querySelector('#progress_status').value = 'Failed to copy');
|
| 225 |
+
"""
|
| 226 |
+
|
| 227 |
+
def open_chatgpt() -> None:
|
| 228 |
+
"""Open ChatGPT in new browser tab"""
|
| 229 |
+
return """window.open('https://chat.openai.com/', '_blank');"""
|
| 230 |
+
|
| 231 |
+
def process_pdf(pdf, fmt, ctx_size, snippet_num, prompt):
|
| 232 |
+
"""Generate prompt from PDF without model processing"""
|
| 233 |
try:
|
| 234 |
if not pdf:
|
| 235 |
+
return "Please upload a PDF file.", "", []
|
| 236 |
|
| 237 |
# Extract text
|
| 238 |
text = extract_text_from_pdf(pdf.name)
|
| 239 |
if text.startswith("Error"):
|
| 240 |
+
return text, "", []
|
| 241 |
|
| 242 |
# Format content
|
| 243 |
formatted_text = format_content(text, fmt)
|
|
|
|
| 250 |
full_prompt = build_prompts(snippets, default_prompt, prompt, snippet_num)
|
| 251 |
|
| 252 |
if isinstance(full_prompt, str) and full_prompt.startswith("Error"):
|
| 253 |
+
return full_prompt, "", []
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 254 |
|
| 255 |
+
# Save prompt for download
|
| 256 |
with tempfile.NamedTemporaryFile(delete=False, mode='w', suffix='.txt') as prompt_file:
|
| 257 |
prompt_file.write(full_prompt)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 258 |
|
| 259 |
+
return "Prompt generated!", full_prompt, [prompt_file.name]
|
| 260 |
|
| 261 |
except Exception as e:
|
| 262 |
logging.error(f"Error processing PDF: {e}")
|
| 263 |
+
return f"Error processing PDF: {str(e)}", "", []
|
| 264 |
|
| 265 |
# Main Interface
|
| 266 |
with gr.Blocks(theme=gr.themes.Default()) as demo:
|
|
|
|
| 287 |
label="π Output Format"
|
| 288 |
)
|
| 289 |
|
| 290 |
+
gr.Markdown("### Context Size")
|
| 291 |
with gr.Row():
|
|
|
|
| 292 |
for size_name, size_value in CONTEXT_SIZES.items():
|
| 293 |
+
gr.Button(
|
| 294 |
+
size_name,
|
| 295 |
+
size="sm", # Make buttons smaller
|
| 296 |
+
scale=1 # Equal scaling
|
| 297 |
+
).click(
|
| 298 |
+
lambda v=size_value: v,
|
| 299 |
+
None,
|
| 300 |
+
context_size
|
| 301 |
+
)
|
| 302 |
|
| 303 |
context_size = gr.Slider(
|
| 304 |
minimum=1000,
|
|
|
|
| 354 |
type="password"
|
| 355 |
)
|
| 356 |
|
| 357 |
+
# In the UI section, add OpenAI API key input:
|
| 358 |
+
with gr.Column(visible=False) as openai_options:
|
| 359 |
+
openai_api_key = gr.Textbox(
|
| 360 |
+
label="π OpenAI API Key",
|
| 361 |
+
type="password"
|
| 362 |
+
)
|
| 363 |
+
|
| 364 |
# Right Column - Output
|
| 365 |
with gr.Column(scale=1):
|
| 366 |
process_button = gr.Button("π Process PDF", variant="primary")
|
|
|
|
| 397 |
def toggle_model_options(choice):
|
| 398 |
return (
|
| 399 |
gr.update(visible=choice == "HuggingFace Inference"),
|
| 400 |
+
gr.update(visible=choice == "Groq API"),
|
| 401 |
+
gr.update(visible=choice == "OpenAI ChatGPT")
|
| 402 |
)
|
| 403 |
|
| 404 |
def refresh_groq_models_list():
|
|
|
|
| 412 |
model_choice.change(
|
| 413 |
toggle_model_options,
|
| 414 |
inputs=[model_choice],
|
| 415 |
+
outputs=[hf_options, groq_options, openai_options]
|
| 416 |
)
|
| 417 |
|
| 418 |
for btn, size_value in context_buttons:
|
|
|
|
| 440 |
format_type,
|
| 441 |
context_size,
|
| 442 |
snippet_number,
|
| 443 |
+
custom_prompt
|
| 444 |
+
],
|
| 445 |
+
outputs=[
|
| 446 |
+
progress_status,
|
| 447 |
+
generated_prompt,
|
| 448 |
+
download_files
|
| 449 |
+
]
|
| 450 |
+
)
|
| 451 |
+
|
| 452 |
+
# Add a new button for sending to model
|
| 453 |
+
send_button = gr.Button("π Send to Model", variant="primary")
|
| 454 |
+
send_button.click(
|
| 455 |
+
send_to_model,
|
| 456 |
+
inputs=[
|
| 457 |
+
generated_prompt,
|
| 458 |
model_choice,
|
| 459 |
hf_model,
|
| 460 |
hf_custom_model,
|
| 461 |
hf_api_key,
|
| 462 |
groq_model,
|
| 463 |
+
groq_api_key,
|
| 464 |
+
openai_api_key
|
| 465 |
],
|
| 466 |
outputs=[
|
|
|
|
|
|
|
| 467 |
summary_output,
|
| 468 |
download_files
|
| 469 |
]
|
|
|
|
| 492 |
1. Upload a PDF document
|
| 493 |
2. Choose output format and context window size
|
| 494 |
3. Select snippet number (default: 1) or enter custom prompt
|
| 495 |
+
4. Select your preferred model in case you want to proceed directly (or continue with 5):
|
| 496 |
- OpenAI ChatGPT: Manual copy/paste workflow
|
| 497 |
- HuggingFace Inference: Direct API integration
|
| 498 |
- Groq API: High-performance inference
|
| 499 |
5. Click 'Process PDF' to generate summary
|
| 500 |
+
6. Use 'Copy Prompt' and, optionally, 'Open ChatGPT' for manual processing
|
| 501 |
7. Download generated files as needed
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 502 |
""")
|
| 503 |
|
| 504 |
# Launch the interface
|