Delete app.py
Browse files
app.py
DELETED
@@ -1,93 +0,0 @@
|
|
1 |
-
import PyPDF2
|
2 |
-
from pprint import pprint
|
3 |
-
from haystack import Pipeline
|
4 |
-
from haystack.schema import Document
|
5 |
-
from haystack.nodes import BM25Retriever
|
6 |
-
from haystack.document_stores import InMemoryDocumentStore
|
7 |
-
from haystack.nodes import PreProcessor, PromptTemplate, PromptNode
|
8 |
-
from pdf2image import convert_from_path
|
9 |
-
import pytesseract
|
10 |
-
from PIL import Image
|
11 |
-
import gradio as gr
|
12 |
-
import os
|
13 |
-
|
14 |
-
# Function to extract text from a PDF file using OCR
|
15 |
-
def extract_text_from_pdf(pdf_path):
|
16 |
-
text = ""
|
17 |
-
# Convert PDF pages to images
|
18 |
-
images = convert_from_path(pdf_path)
|
19 |
-
for image in images:
|
20 |
-
# Perform OCR on the image
|
21 |
-
text += pytesseract.image_to_string(image)
|
22 |
-
return text
|
23 |
-
|
24 |
-
# Process and retrieve answers
|
25 |
-
def process_invoice(pdf, hf_token, questions):
|
26 |
-
# Extract text from the PDF
|
27 |
-
extracted_text = extract_text_from_pdf(pdf.name)
|
28 |
-
document = Document(content=extracted_text)
|
29 |
-
docs = [document]
|
30 |
-
|
31 |
-
# Initializing the processor
|
32 |
-
processor = PreProcessor(
|
33 |
-
clean_empty_lines=True,
|
34 |
-
clean_whitespace=True,
|
35 |
-
clean_header_footer=True,
|
36 |
-
split_by="word",
|
37 |
-
split_length=500,
|
38 |
-
split_respect_sentence_boundary=True,
|
39 |
-
split_overlap=0,
|
40 |
-
)
|
41 |
-
|
42 |
-
preprocessed_docs = processor.process(docs)
|
43 |
-
document_store = InMemoryDocumentStore(use_bm25=True)
|
44 |
-
document_store.write_documents(preprocessed_docs)
|
45 |
-
retriever = BM25Retriever(document_store, top_k=2)
|
46 |
-
|
47 |
-
qa_template = PromptTemplate(prompt=
|
48 |
-
""" Using exclusively the information contained in the context, answer only the question asked without adding
|
49 |
-
suggestions for possible questions, and respond exclusively in English. If the answer cannot be deduced from the
|
50 |
-
context, Don't add anything from the references if it is not asked explicitly. Do not repeat the same information twice
|
51 |
-
respond: "Not sure because not relevant to the context.
|
52 |
-
Context: {join(documents)};
|
53 |
-
Question: {query}
|
54 |
-
""")
|
55 |
-
|
56 |
-
prompt_node = PromptNode(
|
57 |
-
model_name_or_path='mistralai/Mixtral-8x7B-Instruct-v0.1',
|
58 |
-
api_key=hf_token,
|
59 |
-
default_prompt_template=qa_template,
|
60 |
-
max_length=500,
|
61 |
-
model_kwargs={"model_max_length": 5000}
|
62 |
-
)
|
63 |
-
|
64 |
-
rag_pipeline = Pipeline()
|
65 |
-
rag_pipeline.add_node(component=retriever, name="retriever", inputs=["Query"])
|
66 |
-
rag_pipeline.add_node(component=prompt_node, name="prompt_node", inputs=["retriever"])
|
67 |
-
|
68 |
-
answers = {}
|
69 |
-
for question in questions.split(','):
|
70 |
-
result = rag_pipeline.run(query=question.strip())
|
71 |
-
answers[question] = result["results"][0].strip()
|
72 |
-
|
73 |
-
return answers
|
74 |
-
|
75 |
-
# Gradio interface
|
76 |
-
def gradio_interface(pdf, hf_token, questions):
|
77 |
-
answers = process_invoice(pdf, hf_token, questions)
|
78 |
-
return answers
|
79 |
-
|
80 |
-
interface = gr.Interface(
|
81 |
-
fn=gradio_interface,
|
82 |
-
inputs=[
|
83 |
-
gr.inputs.File(file_count="single", type="file", label="Upload Invoice (PDF)"),
|
84 |
-
gr.inputs.Textbox(type="password", label="Enter your Hugging Face Token"),
|
85 |
-
gr.inputs.Textbox(lines=5, placeholder="Enter your questions separated by commas")
|
86 |
-
],
|
87 |
-
outputs="json",
|
88 |
-
title="Invoice Data Extraction",
|
89 |
-
description="Upload an invoice PDF, provide your Hugging Face token, and get the extracted data based on your questions."
|
90 |
-
)
|
91 |
-
|
92 |
-
if __name__ == "__main__":
|
93 |
-
interface.launch()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|