Spaces:
Running
Running
Update main.py
Browse files
main.py
CHANGED
@@ -1,4 +1,4 @@
|
|
1 |
-
from fastapi import FastAPI, UploadFile, File, Form, Request
|
2 |
from fastapi.middleware.cors import CORSMiddleware
|
3 |
from fastapi.responses import HTMLResponse, JSONResponse, FileResponse
|
4 |
from fastapi.staticfiles import StaticFiles
|
@@ -72,4 +72,301 @@ async def predict(
|
|
72 |
else:
|
73 |
return JSONResponse({"error": "Invalid option"}, status_code=400)
|
74 |
except Exception as e:
|
75 |
-
return JSONResponse({"error": f"Prediction failed: {str(e)}"}, status_code=500)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
"""from fastapi import FastAPI, UploadFile, File, Form, Request
|
2 |
from fastapi.middleware.cors import CORSMiddleware
|
3 |
from fastapi.responses import HTMLResponse, JSONResponse, FileResponse
|
4 |
from fastapi.staticfiles import StaticFiles
|
|
|
72 |
else:
|
73 |
return JSONResponse({"error": "Invalid option"}, status_code=400)
|
74 |
except Exception as e:
|
75 |
+
return JSONResponse({"error": f"Prediction failed: {str(e)}"}, status_code=500) """
|
76 |
+
from fastapi import FastAPI, UploadFile, File, Form, Request, HTTPException
|
77 |
+
from fastapi.responses import HTMLResponse, JSONResponse, FileResponse
|
78 |
+
from fastapi.staticfiles import StaticFiles
|
79 |
+
from fastapi.templating import Jinja2Templates
|
80 |
+
from fastapi.middleware.cors import CORSMiddleware
|
81 |
+
from transformers import pipeline, AutoTokenizer, AutoModelForSeq2SeqLM, AutoProcessor, AutoModelForCausalLM
|
82 |
+
from PIL import Image
|
83 |
+
import torch
|
84 |
+
import fitz # PyMuPDF
|
85 |
+
import docx
|
86 |
+
import pptx
|
87 |
+
import openpyxl
|
88 |
+
import re
|
89 |
+
import nltk
|
90 |
+
from nltk.tokenize import sent_tokenize
|
91 |
+
from gtts import gTTS
|
92 |
+
from fpdf import FPDF
|
93 |
+
import tempfile
|
94 |
+
import os
|
95 |
+
import shutil
|
96 |
+
import datetime
|
97 |
+
import hashlib
|
98 |
+
import easyocr
|
99 |
+
from typing import Optional
|
100 |
+
|
101 |
+
# Initialize app
|
102 |
+
app = FastAPI()
|
103 |
+
|
104 |
+
# CORS Configuration
|
105 |
+
app.add_middleware(
|
106 |
+
CORSMiddleware,
|
107 |
+
allow_origins=["*"],
|
108 |
+
allow_credentials=True,
|
109 |
+
allow_methods=["*"],
|
110 |
+
allow_headers=["*"],
|
111 |
+
)
|
112 |
+
|
113 |
+
# Static assets
|
114 |
+
app.mount("/static", StaticFiles(directory="static"), name="static")
|
115 |
+
app.mount("/resources", StaticFiles(directory="resources"), name="resources")
|
116 |
+
|
117 |
+
# Templates
|
118 |
+
templates = Jinja2Templates(directory="templates")
|
119 |
+
|
120 |
+
# Initialize models
|
121 |
+
nltk.download('punkt', quiet=True)
|
122 |
+
|
123 |
+
# Document processing models
|
124 |
+
try:
|
125 |
+
tokenizer = AutoTokenizer.from_pretrained("facebook/bart-large-cnn")
|
126 |
+
model = AutoModelForSeq2SeqLM.from_pretrained("facebook/bart-large-cnn")
|
127 |
+
model.eval()
|
128 |
+
summarizer = pipeline("summarization", model=model, tokenizer=tokenizer, device=-1)
|
129 |
+
reader = easyocr.Reader(['en'], gpu=torch.cuda.is_available())
|
130 |
+
except Exception as e:
|
131 |
+
print(f"Error loading summarization models: {e}")
|
132 |
+
summarizer = None
|
133 |
+
|
134 |
+
# Image captioning models
|
135 |
+
try:
|
136 |
+
processor = AutoProcessor.from_pretrained("microsoft/git-large-coco")
|
137 |
+
git_model = AutoModelForCausalLM.from_pretrained("microsoft/git-large-coco")
|
138 |
+
git_model.eval()
|
139 |
+
USE_GIT = True
|
140 |
+
except Exception as e:
|
141 |
+
print(f"Error loading GIT model, falling back to ViT: {e}")
|
142 |
+
captioner = pipeline("image-to-text", model="nlpconnect/vit-gpt2-image-captioning")
|
143 |
+
USE_GIT = False
|
144 |
+
|
145 |
+
# Helper functions
|
146 |
+
def clean_text(text: str) -> str:
|
147 |
+
text = re.sub(r'\s+', ' ', text)
|
148 |
+
text = re.sub(r'\u2022\s*|\d\.\s+', '', text)
|
149 |
+
text = re.sub(r'\[.*?\]|\(.*?\)', '', text)
|
150 |
+
text = re.sub(r'\bPage\s*\d+\b', '', text, flags=re.IGNORECASE)
|
151 |
+
return text.strip()
|
152 |
+
|
153 |
+
def extract_text(file_path: str, file_extension: str):
|
154 |
+
try:
|
155 |
+
if file_extension == "pdf":
|
156 |
+
with fitz.open(file_path) as doc:
|
157 |
+
text = "\n".join(page.get_text("text") for page in doc)
|
158 |
+
if len(text.strip()) < 50:
|
159 |
+
images = [page.get_pixmap() for page in doc]
|
160 |
+
temp_img = tempfile.NamedTemporaryFile(suffix=".png", delete=False)
|
161 |
+
images[0].save(temp_img.name)
|
162 |
+
ocr_result = reader.readtext(temp_img.name, detail=0)
|
163 |
+
os.unlink(temp_img.name)
|
164 |
+
text = "\n".join(ocr_result) if ocr_result else text
|
165 |
+
return clean_text(text), ""
|
166 |
+
|
167 |
+
elif file_extension == "docx":
|
168 |
+
doc = docx.Document(file_path)
|
169 |
+
return clean_text("\n".join(p.text for p in doc.paragraphs)), ""
|
170 |
+
|
171 |
+
elif file_extension == "pptx":
|
172 |
+
prs = pptx.Presentation(file_path)
|
173 |
+
text = [shape.text for slide in prs.slides for shape in slide.shapes if hasattr(shape, "text")]
|
174 |
+
return clean_text("\n".join(text)), ""
|
175 |
+
|
176 |
+
elif file_extension == "xlsx":
|
177 |
+
wb = openpyxl.load_workbook(file_path, read_only=True)
|
178 |
+
text = [" ".join(str(cell) for cell in row if cell) for sheet in wb.sheetnames for row in wb[sheet].iter_rows(values_only=True)]
|
179 |
+
return clean_text("\n".join(text)), ""
|
180 |
+
|
181 |
+
return "", "Unsupported file format"
|
182 |
+
except Exception as e:
|
183 |
+
return "", f"Error reading {file_extension.upper()} file: {str(e)}"
|
184 |
+
|
185 |
+
def chunk_text(text: str, max_tokens: int = 950):
|
186 |
+
try:
|
187 |
+
sentences = sent_tokenize(text)
|
188 |
+
except:
|
189 |
+
words = text.split()
|
190 |
+
sentences = [' '.join(words[i:i+20]) for i in range(0, len(words), 20)]
|
191 |
+
|
192 |
+
chunks = []
|
193 |
+
current_chunk = ""
|
194 |
+
for sentence in sentences:
|
195 |
+
token_length = len(tokenizer.encode(current_chunk + " " + sentence))
|
196 |
+
if token_length <= max_tokens:
|
197 |
+
current_chunk += " " + sentence
|
198 |
+
else:
|
199 |
+
chunks.append(current_chunk.strip())
|
200 |
+
current_chunk = sentence
|
201 |
+
|
202 |
+
if current_chunk:
|
203 |
+
chunks.append(current_chunk.strip())
|
204 |
+
|
205 |
+
return chunks
|
206 |
+
|
207 |
+
def generate_summary(text: str, length: str = "medium") -> str:
|
208 |
+
cache_key = hashlib.md5((text + length).encode()).hexdigest()
|
209 |
+
|
210 |
+
length_params = {
|
211 |
+
"short": {"max_length": 80, "min_length": 30},
|
212 |
+
"medium": {"max_length": 200, "min_length": 80},
|
213 |
+
"long": {"max_length": 300, "min_length": 210}
|
214 |
+
}
|
215 |
+
|
216 |
+
chunks = chunk_text(text)
|
217 |
+
try:
|
218 |
+
summaries = summarizer(
|
219 |
+
chunks,
|
220 |
+
max_length=length_params[length]["max_length"],
|
221 |
+
min_length=length_params[length]["min_length"],
|
222 |
+
do_sample=False,
|
223 |
+
truncation=True
|
224 |
+
)
|
225 |
+
summary_texts = [s['summary_text'] for s in summaries]
|
226 |
+
except Exception as e:
|
227 |
+
summary_texts = [f"[Error: {str(e)}"]
|
228 |
+
|
229 |
+
final_summary = " ".join(summary_texts)
|
230 |
+
final_summary = ". ".join(s.strip().capitalize() for s in final_summary.split(". ") if s.strip())
|
231 |
+
return final_summary if len(final_summary) > 25 else "Summary too short"
|
232 |
+
|
233 |
+
def text_to_speech(text: str):
|
234 |
+
try:
|
235 |
+
tts = gTTS(text)
|
236 |
+
temp_audio = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3")
|
237 |
+
tts.save(temp_audio.name)
|
238 |
+
return temp_audio.name
|
239 |
+
except Exception as e:
|
240 |
+
print(f"Error in text-to-speech: {e}")
|
241 |
+
return ""
|
242 |
+
|
243 |
+
def create_pdf(summary: str, original_filename: str):
|
244 |
+
try:
|
245 |
+
pdf = FPDF()
|
246 |
+
pdf.add_page()
|
247 |
+
pdf.set_font("Arial", 'B', 16)
|
248 |
+
pdf.cell(200, 10, txt="Document Summary", ln=1, align='C')
|
249 |
+
pdf.set_font("Arial", size=12)
|
250 |
+
pdf.cell(200, 10, txt=f"Original file: {original_filename}", ln=1)
|
251 |
+
pdf.cell(200, 10, txt=f"Generated on: {datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S')}", ln=1)
|
252 |
+
pdf.ln(10)
|
253 |
+
pdf.multi_cell(0, 10, txt=summary)
|
254 |
+
temp_pdf = tempfile.NamedTemporaryFile(delete=False, suffix=".pdf")
|
255 |
+
pdf.output(temp_pdf.name)
|
256 |
+
return temp_pdf.name
|
257 |
+
except Exception as e:
|
258 |
+
print(f"Error creating PDF: {e}")
|
259 |
+
return ""
|
260 |
+
|
261 |
+
def generate_caption(image_path: str) -> str:
|
262 |
+
try:
|
263 |
+
if USE_GIT:
|
264 |
+
image = Image.open(image_path).convert("RGB")
|
265 |
+
inputs = processor(images=image, return_tensors="pt")
|
266 |
+
outputs = git_model.generate(**inputs, max_length=50)
|
267 |
+
caption = processor.batch_decode(outputs, skip_special_tokens=True)[0]
|
268 |
+
else:
|
269 |
+
result = captioner(image_path)
|
270 |
+
caption = result[0]['generated_text']
|
271 |
+
return caption
|
272 |
+
except Exception as e:
|
273 |
+
raise Exception(f"Caption generation failed: {str(e)}")
|
274 |
+
|
275 |
+
# API Endpoints
|
276 |
+
@app.post("/summarize/")
|
277 |
+
async def summarize_document(file: UploadFile = File(...), length: str = Form("medium")):
|
278 |
+
valid_types = [
|
279 |
+
'application/pdf',
|
280 |
+
'application/vnd.openxmlformats-officedocument.wordprocessingml.document',
|
281 |
+
'application/vnd.openxmlformats-officedocument.presentationml.presentation',
|
282 |
+
'application/vnd.openxmlformats-officedocument.spreadsheetml.sheet'
|
283 |
+
]
|
284 |
+
|
285 |
+
if file.content_type not in valid_types:
|
286 |
+
raise HTTPException(
|
287 |
+
status_code=400,
|
288 |
+
detail="Please upload a valid document (PDF, DOCX, PPTX, or XLSX)"
|
289 |
+
)
|
290 |
+
|
291 |
+
try:
|
292 |
+
# Save temp file
|
293 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix=os.path.splitext(file.filename)[1]) as temp:
|
294 |
+
shutil.copyfileobj(file.file, temp)
|
295 |
+
temp_path = temp.name
|
296 |
+
|
297 |
+
# Process file
|
298 |
+
text, error = extract_text(temp_path, os.path.splitext(file.filename)[1][1:].lower())
|
299 |
+
if error:
|
300 |
+
raise HTTPException(status_code=400, detail=error)
|
301 |
+
|
302 |
+
if not text or len(text.split()) < 30:
|
303 |
+
raise HTTPException(status_code=400, detail="Document too short to summarize")
|
304 |
+
|
305 |
+
summary = generate_summary(text, length)
|
306 |
+
audio_path = text_to_speech(summary)
|
307 |
+
pdf_path = create_pdf(summary, file.filename)
|
308 |
+
|
309 |
+
return {
|
310 |
+
"summary": summary,
|
311 |
+
"audio_url": f"/files/{os.path.basename(audio_path)}" if audio_path else None,
|
312 |
+
"pdf_url": f"/files/{os.path.basename(pdf_path)}" if pdf_path else None
|
313 |
+
}
|
314 |
+
|
315 |
+
except HTTPException:
|
316 |
+
raise
|
317 |
+
except Exception as e:
|
318 |
+
raise HTTPException(
|
319 |
+
status_code=500,
|
320 |
+
detail=f"Summarization failed: {str(e)}"
|
321 |
+
)
|
322 |
+
finally:
|
323 |
+
if 'temp_path' in locals() and os.path.exists(temp_path):
|
324 |
+
os.unlink(temp_path)
|
325 |
+
|
326 |
+
@app.post("/imagecaption/")
|
327 |
+
async def caption_image(file: UploadFile = File(...)):
|
328 |
+
valid_types = ['image/jpeg', 'image/png', 'image/gif', 'image/webp']
|
329 |
+
if file.content_type not in valid_types:
|
330 |
+
raise HTTPException(
|
331 |
+
status_code=400,
|
332 |
+
detail="Please upload a valid image (JPEG, PNG, GIF, or WEBP)"
|
333 |
+
)
|
334 |
+
|
335 |
+
try:
|
336 |
+
# Save temp file
|
337 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix=os.path.splitext(file.filename)[1]) as temp:
|
338 |
+
shutil.copyfileobj(file.file, temp)
|
339 |
+
temp_path = temp.name
|
340 |
+
|
341 |
+
# Generate caption
|
342 |
+
caption = generate_caption(temp_path)
|
343 |
+
|
344 |
+
# Generate audio
|
345 |
+
audio_path = text_to_speech(caption)
|
346 |
+
|
347 |
+
return {
|
348 |
+
"answer": caption,
|
349 |
+
"audio": f"/files/{os.path.basename(audio_path)}" if audio_path else None
|
350 |
+
}
|
351 |
+
|
352 |
+
except HTTPException:
|
353 |
+
raise
|
354 |
+
except Exception as e:
|
355 |
+
raise HTTPException(
|
356 |
+
status_code=500,
|
357 |
+
detail=str(e)
|
358 |
+
)
|
359 |
+
finally:
|
360 |
+
if 'temp_path' in locals() and os.path.exists(temp_path):
|
361 |
+
os.unlink(temp_path)
|
362 |
+
|
363 |
+
@app.get("/files/{filename}")
|
364 |
+
async def serve_file(filename: str):
|
365 |
+
file_path = os.path.join(tempfile.gettempdir(), filename)
|
366 |
+
if os.path.exists(file_path):
|
367 |
+
return FileResponse(file_path)
|
368 |
+
raise HTTPException(status_code=404, detail="File not found")
|
369 |
+
|
370 |
+
@app.get("/", response_class=HTMLResponse)
|
371 |
+
async def serve_home(request: Request):
|
372 |
+
return templates.TemplateResponse("HomeS.html", {"request": request})
|