Spaces:
Paused
Paused
Update app.py
Browse files
app.py
CHANGED
@@ -3,7 +3,7 @@ import json
|
|
3 |
import gradio as gr
|
4 |
import pandas as pd
|
5 |
from tempfile import NamedTemporaryFile
|
6 |
-
|
7 |
from langchain_core.prompts import ChatPromptTemplate
|
8 |
from langchain_community.vectorstores import FAISS
|
9 |
from langchain_community.document_loaders import PyPDFLoader
|
@@ -11,119 +11,125 @@ from langchain_core.output_parsers import StrOutputParser
|
|
11 |
from langchain_community.embeddings import HuggingFaceEmbeddings
|
12 |
from langchain_community.llms import HuggingFaceHub
|
13 |
from langchain_core.runnables import RunnableParallel, RunnablePassthrough
|
14 |
-
|
|
|
15 |
huggingface_token = os.environ.get("HUGGINGFACE_TOKEN")
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
23 |
def get_embeddings():
|
24 |
-
|
25 |
-
|
26 |
def create_or_update_database(data, embeddings):
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
|
|
|
|
|
|
|
|
|
|
34 |
prompt = """
|
35 |
Answer the question based only on the following context:
|
36 |
{context}
|
37 |
Question: {question}
|
38 |
-
|
39 |
Provide a concise and direct answer to the question:
|
40 |
"""
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
|
|
|
|
|
|
49 |
def generate_chunked_response(model, prompt, max_tokens=500, max_chunks=5):
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
def response(database, model, question):
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
|
74 |
-
|
75 |
-
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
return "Please enter a question."
|
87 |
-
embed = get_embeddings()
|
88 |
-
database = FAISS.load_local("faiss_database", embed, allow_dangerous_deserialization=True)
|
89 |
-
model = get_model()
|
90 |
-
return response(database, model, question)
|
91 |
-
|
92 |
def extract_db_to_excel():
|
93 |
-
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
|
102 |
-
df.to_excel(excel_path, index=False)
|
103 |
-
|
104 |
-
return excel_path
|
105 |
-
|
106 |
# Gradio interface
|
107 |
with gr.Blocks() as demo:
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
126 |
-
|
127 |
-
|
|
|
128 |
if __name__ == "__main__":
|
129 |
-
|
|
|
3 |
import gradio as gr
|
4 |
import pandas as pd
|
5 |
from tempfile import NamedTemporaryFile
|
6 |
+
from typing import List
|
7 |
from langchain_core.prompts import ChatPromptTemplate
|
8 |
from langchain_community.vectorstores import FAISS
|
9 |
from langchain_community.document_loaders import PyPDFLoader
|
|
|
11 |
from langchain_community.embeddings import HuggingFaceEmbeddings
|
12 |
from langchain_community.llms import HuggingFaceHub
|
13 |
from langchain_core.runnables import RunnableParallel, RunnablePassthrough
|
14 |
+
from langchain_core.text_splitters import RecursiveCharacterTextSplitter
|
15 |
+
from langchain_core.document import Document
|
16 |
huggingface_token = os.environ.get("HUGGINGFACE_TOKEN")
|
17 |
+
def load_and_split_document_basic(file):
|
18 |
+
"""Loads and splits the document into pages."""
|
19 |
+
loader = PyPDFLoader(file.name)
|
20 |
+
data = loader.load_and_split()
|
21 |
+
return data
|
22 |
+
def load_and_split_document_recursive(file: NamedTemporaryFile) -> List[Document]:
|
23 |
+
"""Loads and splits the document into chunks."""
|
24 |
+
loader = PyPDFLoader(file.name)
|
25 |
+
pages = loader.load()
|
26 |
+
text_splitter = RecursiveCharacterTextSplitter(
|
27 |
+
chunk_size=1000,
|
28 |
+
chunk_overlap=200,
|
29 |
+
length_function=len,
|
30 |
+
)
|
31 |
+
chunks = text_splitter.split_documents(pages)
|
32 |
+
return chunks
|
33 |
def get_embeddings():
|
34 |
+
return HuggingFaceEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2")
|
|
|
35 |
def create_or_update_database(data, embeddings):
|
36 |
+
if os.path.exists("faiss_database"):
|
37 |
+
db = FAISS.load_local("faiss_database", embeddings, allow_dangerous_deserialization=True)
|
38 |
+
db.add_documents(data)
|
39 |
+
else:
|
40 |
+
db = FAISS.from_documents(data, embeddings)
|
41 |
+
db.save_local("faiss_database")
|
42 |
+
def clear_cache():
|
43 |
+
if os.path.exists("faiss_database"):
|
44 |
+
os.remove("faiss_database")
|
45 |
+
return "Cache cleared successfully."
|
46 |
+
else:
|
47 |
+
return "No cache to clear."
|
48 |
prompt = """
|
49 |
Answer the question based only on the following context:
|
50 |
{context}
|
51 |
Question: {question}
|
|
|
52 |
Provide a concise and direct answer to the question:
|
53 |
"""
|
54 |
+
def get_model(temperature, top_p, repetition_penalty):
|
55 |
+
return HuggingFaceHub(
|
56 |
+
repo_id="mistralai/Mistral-7B-Instruct-v0.3",
|
57 |
+
model_kwargs={
|
58 |
+
"temperature": temperature,
|
59 |
+
"top_p": top_p,
|
60 |
+
"repetition_penalty": repetition_penalty,
|
61 |
+
"max_length": 512
|
62 |
+
},
|
63 |
+
huggingfacehub_api_token=huggingface_token
|
64 |
+
)
|
65 |
def generate_chunked_response(model, prompt, max_tokens=500, max_chunks=5):
|
66 |
+
full_response = ""
|
67 |
+
for i in range(max_chunks):
|
68 |
+
chunk = model(prompt + full_response, max_new_tokens=max_tokens)
|
69 |
+
full_response += chunk
|
70 |
+
if chunk.strip().endswith((".", "!", "?")):
|
71 |
+
break
|
72 |
+
return full_response.strip()
|
|
|
73 |
def response(database, model, question):
|
74 |
+
prompt_val = ChatPromptTemplate.from_template(prompt)
|
75 |
+
retriever = database.as_retriever()
|
76 |
+
context = retriever.get_relevant_documents(question)
|
77 |
+
context_str = "\n".join([doc.page_content for doc in context])
|
78 |
+
formatted_prompt = prompt_val.format(context=context_str, question=question)
|
79 |
+
ans = generate_chunked_response(model, formatted_prompt)
|
80 |
+
return ans
|
81 |
+
def update_vectors(files, use_recursive_splitter):
|
82 |
+
if not files:
|
83 |
+
return "Please upload at least one PDF file."
|
84 |
+
embed = get_embeddings()
|
85 |
+
total_chunks = 0
|
86 |
+
for file in files:
|
87 |
+
if use_recursive_splitter:
|
88 |
+
data = load_and_split_document_recursive(file)
|
89 |
+
else:
|
90 |
+
data = load_and_split_document_basic(file)
|
91 |
+
create_or_update_database(data, embed)
|
92 |
+
total_chunks += len(data)
|
93 |
+
return f"Vector store updated successfully. Processed {total_chunks} chunks from {len(files)} files."
|
94 |
+
def ask_question(question, temperature, top_p, repetition_penalty):
|
95 |
+
if not question:
|
96 |
+
return "Please enter a question."
|
97 |
+
embed = get_embeddings()
|
98 |
+
database = FAISS.load_local("faiss_database", embed, allow_dangerous_deserialization=True)
|
99 |
+
model = get_model(temperature, top_p, repetition_penalty)
|
100 |
+
return response(database, model, question)
|
|
|
|
|
|
|
|
|
|
|
|
|
101 |
def extract_db_to_excel():
|
102 |
+
embed = get_embeddings()
|
103 |
+
database = FAISS.load_local("faiss_database", embed, allow_dangerous_deserialization=True)
|
104 |
+
documents = database.docstore._dict.values()
|
105 |
+
data = [{"page_content": doc.page_content, "metadata": json.dumps(doc.metadata)} for doc in documents]
|
106 |
+
df = pd.DataFrame(data)
|
107 |
+
with NamedTemporaryFile(delete=False, suffix='.xlsx') as tmp:
|
108 |
+
excel_path = tmp.name
|
109 |
+
df.to_excel(excel_path, index=False)
|
110 |
+
return excel_path
|
|
|
|
|
|
|
|
|
111 |
# Gradio interface
|
112 |
with gr.Blocks() as demo:
|
113 |
+
gr.Markdown("# Chat with your PDF documents")
|
114 |
+
with gr.Row():
|
115 |
+
file_input = gr.Files(label="Upload your PDF documents", file_types=[".pdf"])
|
116 |
+
update_button = gr.Button("Update Vector Store")
|
117 |
+
use_recursive_splitter = gr.Checkbox(label="Use Recursive Text Splitter", value=False)
|
118 |
+
update_output = gr.Textbox(label="Update Status")
|
119 |
+
update_button.click(update_vectors, inputs=[file_input, use_recursive_splitter], outputs=update_output)
|
120 |
+
with gr.Row():
|
121 |
+
question_input = gr.Textbox(label="Ask a question about your documents")
|
122 |
+
temperature_slider = gr.Slider(label="Temperature", minimum=0.0, maximum=1.0, value=0.5, step=0.1)
|
123 |
+
top_p_slider = gr.Slider(label="Top P", minimum=0.0, maximum=1.0, value=0.9, step=0.1)
|
124 |
+
repetition_penalty_slider = gr.Slider(label="Repetition Penalty", minimum=1.0, maximum=2.0, value=1.0, step=0.1)
|
125 |
+
submit_button = gr.Button("Submit")
|
126 |
+
answer_output = gr.Textbox(label="Answer")
|
127 |
+
submit_button.click(ask_question, inputs=[question_input, temperature_slider, top_p_slider, repetition_penalty_slider], outputs=answer_output)
|
128 |
+
extract_button = gr.Button("Extract Database to Excel")
|
129 |
+
excel_output = gr.File(label="Download Excel File")
|
130 |
+
extract_button.click(extract_db_to_excel, inputs=[], outputs=excel_output)
|
131 |
+
clear_button = gr.Button("Clear Cache")
|
132 |
+
clear_output = gr.Textbox(label="Cache Status")
|
133 |
+
clear_button.click(clear_cache, inputs=[], outputs=clear_output)
|
134 |
if __name__ == "__main__":
|
135 |
+
demo.launch()
|