Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,121 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import PyPDF2
|
2 |
+
from openpyxl import load_workbook
|
3 |
+
from pptx import Presentation
|
4 |
+
import gradio as gr
|
5 |
+
import io
|
6 |
+
import docx2python
|
7 |
+
from huggingface_hub import InferenceClient
|
8 |
+
|
9 |
+
# Initialize the Mistral chat model
|
10 |
+
client = InferenceClient("mistralai/Mistral-Nemo-Instruct-2407")
|
11 |
+
|
12 |
+
def read_document(file):
|
13 |
+
file_path = file.name # Get the file path from NamedString
|
14 |
+
file_extension = file_path.split('.')[-1].lower()
|
15 |
+
|
16 |
+
with open(file_path, "rb") as f: # Open the file in binary read mode
|
17 |
+
file_content = f.read()
|
18 |
+
|
19 |
+
if file_extension == 'pdf':
|
20 |
+
try:
|
21 |
+
pdf_reader = PyPDF2.PdfReader(io.BytesIO(file_content))
|
22 |
+
content = ''
|
23 |
+
for page in range(len(pdf_reader.pages)):
|
24 |
+
content += pdf_reader.pages[page].extract_text()
|
25 |
+
return content
|
26 |
+
except Exception as e:
|
27 |
+
return f"Error reading PDF: {e}"
|
28 |
+
|
29 |
+
elif file_extension == 'xlsx':
|
30 |
+
try:
|
31 |
+
wb = load_workbook(io.BytesIO(file_content))
|
32 |
+
content = ''
|
33 |
+
for sheet in wb.worksheets:
|
34 |
+
for row in sheet.rows:
|
35 |
+
for cell in row:
|
36 |
+
content += str(cell.value) + ' '
|
37 |
+
return content
|
38 |
+
except Exception as e:
|
39 |
+
return f"Error reading XLSX: {e}"
|
40 |
+
|
41 |
+
elif file_extension == 'pptx':
|
42 |
+
try:
|
43 |
+
presentation = Presentation(io.BytesIO(file_content))
|
44 |
+
content = ''
|
45 |
+
for slide in presentation.slides:
|
46 |
+
for shape in slide.shapes:
|
47 |
+
if hasattr(shape, "text"):
|
48 |
+
content += shape.text + ' '
|
49 |
+
return content
|
50 |
+
except Exception as e:
|
51 |
+
return f"Error reading PPTX: {e}"
|
52 |
+
|
53 |
+
elif file_extension == 'doc' or file_extension == 'docx':
|
54 |
+
try:
|
55 |
+
doc_result = docx2python.convert(io.BytesIO(file_content))
|
56 |
+
content = ''
|
57 |
+
for page in doc_result:
|
58 |
+
for paragraph in page:
|
59 |
+
if isinstance(paragraph, str):
|
60 |
+
content += paragraph + ' '
|
61 |
+
elif isinstance(paragraph, list):
|
62 |
+
for sub_paragraph in paragraph:
|
63 |
+
if isinstance(sub_paragraph, str):
|
64 |
+
content += sub_paragraph + ' '
|
65 |
+
return content
|
66 |
+
except Exception as e:
|
67 |
+
return f"Error reading DOC/DOCX: {e}"
|
68 |
+
|
69 |
+
else:
|
70 |
+
try:
|
71 |
+
content = file_content.decode('utf-8')
|
72 |
+
return content
|
73 |
+
except Exception as e:
|
74 |
+
return f"Error reading file: {e}"
|
75 |
+
|
76 |
+
def chat_document(file, question):
|
77 |
+
content = str(read_document(file))
|
78 |
+
if len(content) > 128000:
|
79 |
+
content = content[:128000]
|
80 |
+
|
81 |
+
|
82 |
+
# Define system prompt for the chat API
|
83 |
+
system_prompt = """
|
84 |
+
You are a helpful and informative assistant that can answer questions based on the content of documents.
|
85 |
+
You will receive the content of a document and a question about it.
|
86 |
+
Your task is to provide a concise and accurate answer to the question based solely on the provided document content.
|
87 |
+
If the document does not contain enough information to answer the question, simply state that you cannot answer the question based on the provided information.
|
88 |
+
"""
|
89 |
+
|
90 |
+
message = f"""[INST] [SYSTEM] {system_prompt}
|
91 |
+
Document Content: {content}
|
92 |
+
Question: {question}
|
93 |
+
Answer:"""
|
94 |
+
|
95 |
+
stream = client.text_generation(message, max_new_tokens=512, stream=True, details=True, return_full_text=False)
|
96 |
+
output = ""
|
97 |
+
for response in stream:
|
98 |
+
output += response.token.text
|
99 |
+
return output
|
100 |
+
|
101 |
+
|
102 |
+
with gr.Blocks() as demo:
|
103 |
+
with gr.Tabs():
|
104 |
+
with gr.TabItem("Document Reader"):
|
105 |
+
iface1 = gr.Interface(
|
106 |
+
fn=read_document,
|
107 |
+
inputs=gr.File(label="Upload a Document"),
|
108 |
+
outputs=gr.Textbox(label="Document Content"),
|
109 |
+
title="Document Reader",
|
110 |
+
description="Upload a document (PDF, XLSX, PPTX, TXT, CSV, DOC, DOCX and Code or text file) to read its content."
|
111 |
+
)
|
112 |
+
with gr.TabItem("Document Chat"):
|
113 |
+
iface2 = gr.Interface(
|
114 |
+
fn=chat_document,
|
115 |
+
inputs=[gr.File(label="Upload a Document"), gr.Textbox(label="Question")],
|
116 |
+
outputs=gr.Textbox(label="Answer"),
|
117 |
+
title="Document Chat",
|
118 |
+
description="Upload a document and ask questions about its content."
|
119 |
+
)
|
120 |
+
|
121 |
+
demo.launch()
|