Skip to content

Commit ba28ea5

Browse files
kevatsakevatsa
committed
Created using Colaboratory
1 parent 3e6648b commit ba28ea5

File tree

1 file changed

+207
-0
lines changed

1 file changed

+207
-0
lines changed

‎LLM_Question_Answering_Application.ipynb‎

Lines changed: 207 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,207 @@
1+
{
2+
"nbformat": 4,
3+
"nbformat_minor": 0,
4+
"metadata": {
5+
"colab": {
6+
"provenance": [],
7+
"mount_file_id": "10E5yOrYsLu1k3zhTg0ji9QG3o690Q_Wh",
8+
"authorship_tag": "ABX9TyNYWsOTZcD6wahQCgxO8jJj",
9+
"include_colab_link": true
10+
},
11+
"kernelspec": {
12+
"name": "python3",
13+
"display_name": "Python 3"
14+
},
15+
"language_info": {
16+
"name": "python"
17+
}
18+
},
19+
"cells": [
20+
{
21+
"cell_type": "markdown",
22+
"metadata": {
23+
"id": "view-in-github",
24+
"colab_type": "text"
25+
},
26+
"source": [
27+
"<a href=\"https://colab.research.google.com/github/kevatsa/Python-programming-exercises/blob/master/LLM_Question_Answering_Application.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
28+
]
29+
},
30+
{
31+
"cell_type": "code",
32+
"execution_count": null,
33+
"metadata": {
34+
"id": "CrgGY3vJ0nYV"
35+
},
36+
"outputs": [],
37+
"source": [
38+
"pip install -r /content/drive/MyDrive/Frontend/requirements.txt\n",
39+
"pip install chromadb\n",
40+
"pip install tiktoken\n",
41+
"pip install cohere\n"
42+
]
43+
},
44+
{
45+
"cell_type": "code",
46+
"source": [
47+
"\n",
48+
"import streamlit as st\n",
49+
"from langchain.embeddings.openai import OpenAIEmbeddings\n",
50+
"from langchain.vectorstores import Chroma\n",
51+
"import os\n",
52+
"\n",
53+
"\n",
54+
"\n",
55+
"def load_document(file):\n",
56+
" import os\n",
57+
" name, extension = os.path.splitext(file)\n",
58+
"\n",
59+
"\n",
60+
" if extension == '.pdf':\n",
61+
" from langchain.document_loaders import PyPDFLoader\n",
62+
" print(f'Loading {file}')\n",
63+
" loader = PyPDFLoader(file)\n",
64+
" elif extension == '.docx':\n",
65+
" from langchain.document_loaders import Docx2txtLoader\n",
66+
" print(f'Loading {file}')\n",
67+
" loader = Docx2txtLoader(file)\n",
68+
" elif extension == '.txt':\n",
69+
" from langchain.document_loaders import TextLoader\n",
70+
" print(f'Loading {file}')\n",
71+
" loader = TextLoader(file)\n",
72+
" else:\n",
73+
" print('Doc format is not supported')\n",
74+
" return None\n",
75+
"\n",
76+
" data = loader.load()\n",
77+
" return data\n",
78+
"\n",
79+
"\n",
80+
"\n",
81+
"\n",
82+
"def chunk_data(data, chunk_size=256, chunk_overlap=20):\n",
83+
" from langchain.text_splitter import RecursiveCharacterTextSplitter\n",
84+
" text_splitter = RecursiveCharacterTextSplitter(chunk_size=chunk_size, chunk_overlap=chunk_overlap)\n",
85+
" chunks = text_splitter.split_documents(data)\n",
86+
" return chunks\n",
87+
"\n",
88+
"\n",
89+
"\n",
90+
"def create_embeddings(chunks):\n",
91+
" embeddings = OpenAIEmbeddings()\n",
92+
" vector_store = Chroma.from_documents(chunks, embeddings)\n",
93+
" return vector_store\n",
94+
"\n",
95+
"\n",
96+
"\n",
97+
"def ask_and_get_answer(vector_store, q, k=3):\n",
98+
" from langchain.chains import RetrievalQA\n",
99+
" from langchain.chat_models import ChatOpenAI\n",
100+
"\n",
101+
" llm = ChatOpenAI(model='gpt-3.5-turbo', temperature=2)\n",
102+
"\n",
103+
" retriever = vector_store.as_retriever(search_type='similarity', search_kwargs={'k':k})\n",
104+
"\n",
105+
" chain = RetrievalQA.from_chain_type(llm=llm, chain_type='stuff', retriever=retriever)\n",
106+
"\n",
107+
" answer = chain.run(q)\n",
108+
" return answer\n",
109+
"\n",
110+
"\n",
111+
"\n",
112+
"def calculate_embedding_cost(texts):\n",
113+
" import tiktoken\n",
114+
" enc = tiktoken.encoding_for_model('text-embedding-ada-002')\n",
115+
" total_tokens = sum([len(enc.encode(page.page_content)) for page in texts])\n",
116+
" #print(f'Total tokens: {total_tokens}')\n",
117+
" #print(f'Embedding cost in USD: {total_tokens / 1000 * 0.0004:.6f}')\n",
118+
" return total_tokens, total_tokens / 1000 * 0.0004\n",
119+
"\n",
120+
"\n",
121+
"def clear_history():\n",
122+
" if 'history' in st.session_state:\n",
123+
" del st.session_state['history']\n",
124+
"\n",
125+
"\n",
126+
"\n",
127+
"if __name__ == \"__main__\":\n",
128+
" import os\n",
129+
" from dotenv import load_dotenv, find_dotenv\n",
130+
" load_dotenv(find_dotenv(), override=True)\n",
131+
"\n",
132+
"\n",
133+
" st.image('/content/drive/MyDrive/Frontend/img.jpeg')\n",
134+
" st.subheader('LLM Question-Answering Application')\n",
135+
" with st.sidebar:\n",
136+
" api_key = st.text_input('OpenAI API Key: ', type='password')\n",
137+
" if api_key:\n",
138+
" os.environ['OPENAI_API_KEY'] = api_key\n",
139+
"\n",
140+
" uploaded_file = st.file_uploader('Upload a file:', type=['pdf', 'docx', 'txt'])\n",
141+
" chunk_size = st.number_input('Chunk size:', min_value=100, max_value=2048, value=512, on_change=clear_history)\n",
142+
" k = st.number_input('k', min_value=1, max_value=20,value=3, on_change=clear_history)\n",
143+
" add_data = st.button('Add Data', on_click=clear_history)\n",
144+
"\n",
145+
" if uploaded_file and add_data:\n",
146+
" with st.spinner('Reading, chunking, embedding file...'):\n",
147+
" bytes_data = uploaded_file.read()\n",
148+
" file_name = os.path.join('./', uploaded_file.name)\n",
149+
" with open(file_name, 'wb') as f:\n",
150+
" f.write(bytes_data)\n",
151+
"\n",
152+
" data = load_document(file_name)\n",
153+
" chunks = chunk_data(data, chunk_size=chunk_size)\n",
154+
" st.write(f'Chunk size: {chunk_size}, chunks: {len(chunks)}')\n",
155+
"\n",
156+
"\n",
157+
" tokens, embedding_cost = calculate_embedding_cost(chunks)\n",
158+
" st.write(f'Embedding cost: ${embedding_cost:.4f}')\n",
159+
"\n",
160+
" vector_store = create_embeddings(chunks)\n",
161+
"\n",
162+
" st.session_state.vs = vector_store\n",
163+
"\n",
164+
" st.success('File uploaded, Chunked and Embedded successfully.')\n",
165+
"\n",
166+
"\n",
167+
" q = st.text_input('Ask a question about the content of your file:')\n",
168+
" if q:\n",
169+
" if 'vs' in st.session_state:\n",
170+
" vector_store = st.session_state.vs\n",
171+
" #st.write(f'k: {k}')\n",
172+
" answer = ask_and_get_answer(vector_store, q, k)\n",
173+
" st.text_area('LLM Answer:', value=answer)\n",
174+
"\n",
175+
"\n",
176+
" st.divider()\n",
177+
" if 'history' not in st.session_state:\n",
178+
" st.session_state.history = ' '\n",
179+
"\n",
180+
" value = f'Q: {q} \\nA: {answer} '\n",
181+
" st.session_state.history = f'{value} \\n {\"-\" * 100} \\n {st.session_state.history}'\n",
182+
" h = st.session_state.history\n",
183+
" st.text_area(label='Chat History', value=h, key='history', height=400)\n",
184+
"\n",
185+
"\n",
186+
"\n",
187+
"\n"
188+
],
189+
"metadata": {
190+
"id": "Cw2UIcc22jjQ"
191+
},
192+
"execution_count": null,
193+
"outputs": []
194+
},
195+
{
196+
"cell_type": "code",
197+
"source": [
198+
"!streamlit run /usr/local/lib/python3.10/dist-packages/colab_kernel_launcher.py [ARGUMENTS]\n"
199+
],
200+
"metadata": {
201+
"id": "q_02qlJzh2Jd"
202+
},
203+
"execution_count": null,
204+
"outputs": []
205+
}
206+
]
207+
}

0 commit comments

Comments
 (0)