1+ from fastapi import FastAPI , UploadFile , File
2+ from fastapi .responses import JSONResponse
3+ import uvicorn
4+ import os
5+ from groq import Groq
6+ import PyPDF2
7+ from pptx import Presentation
8+ import io
9+ from PIL import Image
10+ import tempfile
11+ import base64
12+ from mangum import Mangum
13+
14+ app = FastAPI ()
15+ handler = Mangum (app )
16+ # Initialize Groq client
17+ client = Groq (
18+ api_key = "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" ,
19+ )
20+
21+ @app .get ("/" )
22+ async def root ():
23+ return {"message" : "Welcome" }
24+
25+ @app .get ("/greeting" )
26+ async def greeting ():
27+ return {"message" : "Hello" , "details" : "You are using LAamAScholar Bot" }
28+
29+ @app .post ("/summarize" )
30+ async def summarize_file (file : UploadFile = File (...)):
31+ content = await file .read ()
32+
33+ if file .filename .endswith ('.pdf' ):
34+ text = extract_text_from_pdf (content )
35+ elif file .filename .endswith ('.pptx' ):
36+ text = extract_text_from_pptx (content )
37+ else :
38+ return JSONResponse (status_code = 400 , content = {"error" : "Unsupported file format. Please upload a PDF or PPTX file." })
39+
40+ summary = summarize_text (text )
41+ return {"summary" : summary }
42+
43+ @app .post ("/generate_questions" )
44+ async def generate_questions (file : UploadFile = File (...)):
45+ content = await file .read ()
46+
47+ if file .filename .endswith ('.pdf' ):
48+ text = extract_text_from_pdf (content )
49+ elif file .filename .endswith ('.pptx' ):
50+ text = extract_text_from_pptx (content )
51+ else :
52+ return JSONResponse (status_code = 400 , content = {"error" : "Unsupported file format. Please upload a PDF or PPTX file." })
53+
54+ questions = generate_important_questions (text )
55+ return {"questions" : questions }
56+
57+ @app .post ("/voice-notes" )
58+ async def transcribe_voice_notes (file : UploadFile = File (...)):
59+ if not file .filename .endswith ('.mp3' ):
60+ return JSONResponse (status_code = 400 , content = {"error" : "Unsupported file format. Please upload an MP3 file." })
61+
62+ # Create a temporary file to store the uploaded content
63+ with tempfile .NamedTemporaryFile (delete = False , suffix = ".mp3" ) as temp_file :
64+ content = await file .read ()
65+ temp_file .write (content )
66+ temp_file_path = temp_file .name
67+
68+ try :
69+ with open (temp_file_path , "rb" ) as audio_file :
70+ transcription = client .audio .transcriptions .create (
71+ file = (file .filename , audio_file ),
72+ model = "whisper-large-v3" ,
73+ response_format = "verbose_json" ,
74+ )
75+
76+ return {"transcription" : transcription .text }
77+ finally :
78+ # Clean up the temporary file
79+ os .unlink (temp_file_path )
80+ def encode_image (image_file ):
81+ return base64 .b64encode (image_file .read ()).decode ('utf-8' )
82+
83+ @app .post ("/ocr/" )
84+ async def perform_ocr (file : UploadFile = File (...)):
85+ # Read and encode the image
86+ contents = await file .read ()
87+ base64_image = encode_image (io .BytesIO (contents ))
88+
89+ # Prepare the message for Groq
90+ messages = [
91+ {
92+ "role" : "user" ,
93+ "content" : [
94+ {
95+ "type" : "image_url" ,
96+ "image_url" : {
97+ "url" : f"data:image/jpeg;base64,{ base64_image }
98+ }
99+ },
100+ {
101+ "type" : "text" ,
102+ "text" : "Explain what is written in the image from an educational Point of View."
103+ }
104+ ]
105+ }
106+ ]
107+
108+ # Make the API call to Groq
109+ completion = client .chat .completions .create (
110+ model = "llama-3.2-11b-vision-preview" ,
111+ messages = messages ,
112+ temperature = 0 , # Set to 0 for more deterministic results
113+ max_tokens = 1024 ,
114+ top_p = 1 ,
115+ stream = False ,
116+ stop = None ,
117+ )
118+
119+ # Extract the OCR result from the response
120+ ocr_result = completion .choices [0 ].message .content
121+
122+ return {"ocr_result" : ocr_result }
123+
124+ def extract_text_from_pdf (content ):
125+ pdf_reader = PyPDF2 .PdfReader (io .BytesIO (content ))
126+ text = ""
127+ for page in pdf_reader .pages :
128+ text += page .extract_text ()
129+ return text
130+
131+ def extract_text_from_pptx (content ):
132+ prs = Presentation (io .BytesIO (content ))
133+ text = ""
134+ for slide in prs .slides :
135+ for shape in slide .shapes :
136+ if hasattr (shape , 'text' ):
137+ text += shape .text + "\n "
138+ return text
139+
140+ def summarize_text (text ):
141+ prompt = f"Summarize the following text:\n \n { text [:4000 ]} # Truncate to 4000 characters to fit within token limit
142+
143+ chat_completion = client .chat .completions .create (
144+ messages = [
145+ {
146+ "role" : "user" ,
147+ "content" : prompt ,
148+ }
149+ ],
150+ model = "llama3-8b-8192" ,
151+ )
152+
153+ return chat_completion .choices [0 ].message .content
154+ def generate_important_questions (text ):
155+ prompt = f"Generate 5 important questions based on the following text:\n \n { text [:4000 ]} # Truncate to 4000 characters to fit within token limit
156+
157+ chat_completion = client .chat .completions .create (
158+ messages = [
159+ {
160+ "role" : "user" ,
161+ "content" : prompt ,
162+ }
163+ ],
164+ model = "llama3-8b-8192" ,
165+ )
166+
167+ return chat_completion .choices [0 ].message .content
168+
169+ if __name__ == "__main__" :
170+ uvicorn .run (app , host = "0.0.0.0" , port = 8000 )
0 commit comments