Update app.py
Browse files
app.py
CHANGED
@@ -6,6 +6,7 @@ from transformers import AutoTokenizer, AutoModel
|
|
6 |
import torch
|
7 |
from pdf2image import convert_from_path
|
8 |
import io
|
|
|
9 |
|
10 |
# CSS styles
|
11 |
css = """
|
@@ -64,14 +65,12 @@ def get_text_embeddings(text, model_name='bert-base-uncased'):
|
|
64 |
# Function to process PDF and generate a response
|
65 |
def process_pdf_and_generate_response(pdf_file):
|
66 |
try:
|
67 |
-
#
|
68 |
-
|
69 |
-
with open(tmp_pdf_path, 'wb') as tmp_pdf:
|
70 |
-
tmp_pdf.write(pdf_file.read())
|
71 |
|
72 |
# Convert PDF to images
|
73 |
img_dir = "pdf_images"
|
74 |
-
pdf_to_images(
|
75 |
|
76 |
# Generate embeddings for each image
|
77 |
image_embeddings = []
|
@@ -93,7 +92,7 @@ def process_pdf_and_generate_response(pdf_file):
|
|
93 |
|
94 |
iface = gr.Interface(
|
95 |
fn=process_pdf_and_generate_response,
|
96 |
-
inputs=gr.File(label="Upload PDF", type="
|
97 |
outputs=gr.Textbox("Placeholder for PDF insights", label="Insights", type="text"),
|
98 |
title="pdf-chatbot",
|
99 |
description="Upload a PDF and receive insights based on its content.",
|
|
|
6 |
import torch
|
7 |
from pdf2image import convert_from_path
|
8 |
import io
|
9 |
+
from io import BytesIO
|
10 |
|
11 |
# CSS styles
|
12 |
css = """
|
|
|
65 |
# Function to process PDF and generate a response
|
66 |
def process_pdf_and_generate_response(pdf_file):
|
67 |
try:
|
68 |
+
# Convert the binary stream to a file-like object
|
69 |
+
pdf_file_stream = BytesIO(pdf_file)
|
|
|
|
|
70 |
|
71 |
# Convert PDF to images
|
72 |
img_dir = "pdf_images"
|
73 |
+
pdf_to_images(pdf_file_stream, img_dir)
|
74 |
|
75 |
# Generate embeddings for each image
|
76 |
image_embeddings = []
|
|
|
92 |
|
93 |
iface = gr.Interface(
|
94 |
fn=process_pdf_and_generate_response,
|
95 |
+
inputs=gr.File(label="Upload PDF", type="binary"), # Changed 'file' to 'binary'
|
96 |
outputs=gr.Textbox("Placeholder for PDF insights", label="Insights", type="text"),
|
97 |
title="pdf-chatbot",
|
98 |
description="Upload a PDF and receive insights based on its content.",
|