Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,18 +1,33 @@
|
|
1 |
import os
|
2 |
-
|
3 |
-
import fitz # PyMuPDF
|
4 |
from docx import Document
|
5 |
from sentence_transformers import SentenceTransformer
|
6 |
import faiss
|
7 |
import numpy as np
|
8 |
import pickle
|
9 |
-
from langchain_community.llms import HuggingFaceEndpoint
|
10 |
from langchain_community.vectorstores import FAISS
|
11 |
from langchain_community.embeddings import HuggingFaceEmbeddings
|
12 |
import gradio as gr
|
|
|
13 |
|
14 |
-
#
|
15 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
|
17 |
# Initialize the embedding model
|
18 |
embedding_model = SentenceTransformer('all-MiniLM-L6-v2')
|
@@ -21,15 +36,16 @@ embedding_model = SentenceTransformer('all-MiniLM-L6-v2')
|
|
21 |
api_token = os.getenv('HUGGINGFACEHUB_API_TOKEN')
|
22 |
if not api_token:
|
23 |
raise ValueError("HUGGINGFACEHUB_API_TOKEN environment variable is not set")
|
|
|
24 |
print(f"API Token: {api_token[:5]}...")
|
25 |
|
26 |
-
# Initialize the HuggingFace LLM
|
27 |
llm = HuggingFaceEndpoint(
|
28 |
endpoint_url="https://api-inference.huggingface.co/models/gpt2",
|
29 |
model_kwargs={"api_key": api_token}
|
30 |
)
|
31 |
|
32 |
-
# Initialize the HuggingFace
|
33 |
embedding = HuggingFaceEmbeddings()
|
34 |
|
35 |
# Load or create FAISS index
|
@@ -43,56 +59,67 @@ else:
|
|
43 |
with open(index_path, "wb") as f:
|
44 |
pickle.dump(index, f)
|
45 |
|
46 |
-
|
47 |
-
|
48 |
-
def extract_text_from_pdf(pdf_path):
|
49 |
-
text = ""
|
50 |
-
doc = fitz.open(pdf_path)
|
51 |
-
for page_num in range(len(doc)):
|
52 |
-
page = doc.load_page(page_num)
|
53 |
-
text += page.get_text()
|
54 |
-
return text
|
55 |
-
|
56 |
-
|
57 |
-
# Function to extract text from a Word document
|
58 |
-
def extract_text_from_docx(docx_path):
|
59 |
-
doc = Document(docx_path)
|
60 |
-
text = "\n".join([para.text for para in doc.paragraphs])
|
61 |
-
return text
|
62 |
-
|
63 |
-
def process_and_query(state, text, file=None):
|
64 |
-
# Initialize state on first run
|
65 |
-
if state is None:
|
66 |
-
state = {"processed_text": None, "conversation": []}
|
67 |
-
|
68 |
-
# Check if a file is uploaded
|
69 |
-
if file:
|
70 |
-
# Get the uploaded file content
|
71 |
content = file.read()
|
72 |
-
if file.
|
73 |
with open("temp.pdf", "wb") as f:
|
74 |
f.write(content)
|
75 |
-
|
76 |
-
elif file.
|
77 |
with open("temp.docx", "wb") as f:
|
78 |
f.write(content)
|
79 |
-
|
80 |
else:
|
81 |
return {"error": "Unsupported file format"}
|
82 |
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
|
|
|
|
89 |
|
90 |
-
|
91 |
-
|
92 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
93 |
|
94 |
-
# Return the conversation history and potentially answer
|
95 |
-
return {"conversation": state["conversation"]}
|
96 |
|
97 |
|
98 |
|
|
|
1 |
import os
|
2 |
+
import fitz
|
|
|
3 |
from docx import Document
|
4 |
from sentence_transformers import SentenceTransformer
|
5 |
import faiss
|
6 |
import numpy as np
|
7 |
import pickle
|
8 |
+
from langchain_community.llms import HuggingFaceEndpoint
|
9 |
from langchain_community.vectorstores import FAISS
|
10 |
from langchain_community.embeddings import HuggingFaceEmbeddings
|
11 |
import gradio as gr
|
12 |
+
from fastapi import FastAPI
|
13 |
|
14 |
+
# Initialize FastAPI
|
15 |
+
app = FastAPI()
|
16 |
+
|
17 |
+
# Function to extract text from a PDF file
|
18 |
+
def extract_text_from_pdf(pdf_path):
|
19 |
+
text = ""
|
20 |
+
doc = fitz.open(pdf_path)
|
21 |
+
for page_num in range(len(doc)):
|
22 |
+
page = doc.load_page(page_num)
|
23 |
+
text += page.get_text()
|
24 |
+
return text
|
25 |
+
|
26 |
+
# Function to extract text from a Word document
|
27 |
+
def extract_text_from_docx(docx_path):
|
28 |
+
doc = Document(docx_path)
|
29 |
+
text = "\n".join([para.text for para in doc.paragraphs])
|
30 |
+
return text
|
31 |
|
32 |
# Initialize the embedding model
|
33 |
embedding_model = SentenceTransformer('all-MiniLM-L6-v2')
|
|
|
36 |
api_token = os.getenv('HUGGINGFACEHUB_API_TOKEN')
|
37 |
if not api_token:
|
38 |
raise ValueError("HUGGINGFACEHUB_API_TOKEN environment variable is not set")
|
39 |
+
|
40 |
print(f"API Token: {api_token[:5]}...")
|
41 |
|
42 |
+
# Initialize the HuggingFace LLM
|
43 |
llm = HuggingFaceEndpoint(
|
44 |
endpoint_url="https://api-inference.huggingface.co/models/gpt2",
|
45 |
model_kwargs={"api_key": api_token}
|
46 |
)
|
47 |
|
48 |
+
# Initialize the HuggingFace embeddings
|
49 |
embedding = HuggingFaceEmbeddings()
|
50 |
|
51 |
# Load or create FAISS index
|
|
|
59 |
with open(index_path, "wb") as f:
|
60 |
pickle.dump(index, f)
|
61 |
|
62 |
+
def upload_files(files):
|
63 |
+
for file in files:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
64 |
content = file.read()
|
65 |
+
if file.name.endswith('.pdf'):
|
66 |
with open("temp.pdf", "wb") as f:
|
67 |
f.write(content)
|
68 |
+
text = extract_text_from_pdf("temp.pdf")
|
69 |
+
elif file.name.endswith('.docx'):
|
70 |
with open("temp.docx", "wb") as f:
|
71 |
f.write(content)
|
72 |
+
text = extract_text_from_docx("temp.docx")
|
73 |
else:
|
74 |
return {"error": "Unsupported file format"}
|
75 |
|
76 |
+
# Process the text and update FAISS index
|
77 |
+
sentences = text.split("\n")
|
78 |
+
embeddings = embedding_model.encode(sentences)
|
79 |
+
index.add(np.array(embeddings))
|
80 |
+
|
81 |
+
# Save the updated index
|
82 |
+
with open(index_path, "wb") as f:
|
83 |
+
pickle.dump(index, f)
|
84 |
|
85 |
+
return "Files processed successfully"
|
86 |
+
|
87 |
+
def query_text(text):
|
88 |
+
# Encode the query text
|
89 |
+
query_embedding = embedding_model.encode([text])
|
90 |
+
|
91 |
+
# Search the FAISS index
|
92 |
+
D, I = index.search(np.array(query_embedding), k=5)
|
93 |
+
|
94 |
+
top_documents = []
|
95 |
+
for idx in I[0]:
|
96 |
+
if idx != -1: # Ensure that a valid index is found
|
97 |
+
top_documents.append(f"Document {idx}")
|
98 |
+
|
99 |
+
return top_documents
|
100 |
+
|
101 |
+
# Create Gradio interface
|
102 |
+
with gr.Blocks() as demo:
|
103 |
+
gr.Markdown("## Document Upload and Query System")
|
104 |
+
|
105 |
+
with gr.Tab("Upload Files"):
|
106 |
+
upload = gr.File(file_count="multiple", label="Upload PDF or DOCX files")
|
107 |
+
upload_button = gr.Button("Upload")
|
108 |
+
upload_output = gr.Textbox()
|
109 |
+
upload_button.click(fn=upload_files, inputs=upload, outputs=upload_output)
|
110 |
+
|
111 |
+
with gr.Tab("Query"):
|
112 |
+
query = gr.Textbox(label="Enter your query")
|
113 |
+
query_button = gr.Button("Search")
|
114 |
+
query_output = gr.Textbox()
|
115 |
+
query_button.click(fn=query_text, inputs=query, outputs=query_output)
|
116 |
+
|
117 |
+
demo.launch()
|
118 |
+
|
119 |
+
if __name__ == "__main__":
|
120 |
+
import uvicorn
|
121 |
+
uvicorn.run(app, host="0.0.0.0", port=8001)
|
122 |
|
|
|
|
|
123 |
|
124 |
|
125 |
|