Spaces:
Build error
Build error
Commit
•
e2dccc5
1
Parent(s):
f7e5eb3
- pages/Tesla_Alerts.py +143 -0
- pages/tc1.ppy +129 -0
pages/Tesla_Alerts.py
ADDED
@@ -0,0 +1,143 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from langchain.prompts import PromptTemplate
|
2 |
+
from langchain.output_parsers import PydanticOutputParser
|
3 |
+
|
4 |
+
from llama_index import VectorStoreIndex, ServiceContext, StorageContext
|
5 |
+
from llama_index.vector_stores import FaissVectorStore
|
6 |
+
from llama_index.tools import QueryEngineTool, ToolMetadata
|
7 |
+
from llama_index.query_engine import SubQuestionQueryEngine
|
8 |
+
from llama_index.embeddings import OpenAIEmbedding
|
9 |
+
from llama_index.schema import Document
|
10 |
+
from llama_index.node_parser import UnstructuredElementNodeParser
|
11 |
+
|
12 |
+
from src.utils import get_model, process_pdf2
|
13 |
+
|
14 |
+
import streamlit as st
|
15 |
+
import os
|
16 |
+
import faiss
|
17 |
+
import time
|
18 |
+
from pypdf import PdfReader
|
19 |
+
|
20 |
+
|
21 |
+
st.set_page_config(page_title="Yield Case Analyzer", page_icon=":card_index_dividers:", initial_sidebar_state="expanded", layout="wide")
|
22 |
+
|
23 |
+
st.title(":card_index_dividers: Yield Case Analyzer")
|
24 |
+
st.info("""
|
25 |
+
Begin by uploading the case report in PDF format. Afterward, click on 'Process Document'. Once the document has been processed. You can enter question and click send, system will answer your question.
|
26 |
+
""")
|
27 |
+
|
28 |
+
def process_pdf(pdf):
|
29 |
+
file = PdfReader(pdf)
|
30 |
+
print("in process pdf")
|
31 |
+
document_list = []
|
32 |
+
for page in file.pages:
|
33 |
+
document_list.append(Document(text=str(page.extract_text())))
|
34 |
+
print("in process pdf 1")
|
35 |
+
|
36 |
+
node_paser = UnstructuredElementNodeParser()
|
37 |
+
print("in process pdf 1")
|
38 |
+
|
39 |
+
nodes = node_paser.get_nodes_from_documents(document_list, show_progress=True)
|
40 |
+
|
41 |
+
return nodes
|
42 |
+
|
43 |
+
|
44 |
+
def get_vector_index(nodes, vector_store):
|
45 |
+
print(nodes)
|
46 |
+
llm = get_model("openai")
|
47 |
+
if vector_store == "faiss":
|
48 |
+
d = 1536
|
49 |
+
faiss_index = faiss.IndexFlatL2(d)
|
50 |
+
vector_store = FaissVectorStore(faiss_index=faiss_index)
|
51 |
+
storage_context = StorageContext.from_defaults(vector_store=vector_store)
|
52 |
+
# embed_model = OpenAIEmbedding()
|
53 |
+
# service_context = ServiceContext.from_defaults(embed_model=embed_model)
|
54 |
+
service_context = ServiceContext.from_defaults(llm=llm)
|
55 |
+
index = VectorStoreIndex(nodes,
|
56 |
+
service_context=service_context,
|
57 |
+
storage_context=storage_context
|
58 |
+
)
|
59 |
+
elif vector_store == "simple":
|
60 |
+
index = VectorStoreIndex.from_documents(nodes)
|
61 |
+
|
62 |
+
|
63 |
+
return index
|
64 |
+
|
65 |
+
|
66 |
+
|
67 |
+
def generate_insight(engine, search_string):
|
68 |
+
|
69 |
+
with open("prompts/main.prompt", "r") as f:
|
70 |
+
template = f.read()
|
71 |
+
|
72 |
+
prompt_template = PromptTemplate(
|
73 |
+
template=template,
|
74 |
+
input_variables=['search_string']
|
75 |
+
)
|
76 |
+
|
77 |
+
formatted_input = prompt_template.format(search_string=search_string)
|
78 |
+
print(formatted_input)
|
79 |
+
response = engine.query(formatted_input)
|
80 |
+
return response.response
|
81 |
+
|
82 |
+
|
83 |
+
def get_query_engine(engine):
|
84 |
+
llm = get_model("openai")
|
85 |
+
service_context = ServiceContext.from_defaults(llm=llm)
|
86 |
+
|
87 |
+
query_engine_tools = [
|
88 |
+
QueryEngineTool(
|
89 |
+
query_engine=engine,
|
90 |
+
metadata=ToolMetadata(
|
91 |
+
name="Alert Report",
|
92 |
+
description=f"Provides information about the cases from its case report.",
|
93 |
+
),
|
94 |
+
),
|
95 |
+
]
|
96 |
+
|
97 |
+
|
98 |
+
s_engine = SubQuestionQueryEngine.from_defaults(
|
99 |
+
query_engine_tools=query_engine_tools,
|
100 |
+
service_context=service_context
|
101 |
+
)
|
102 |
+
return s_engine
|
103 |
+
|
104 |
+
|
105 |
+
if "process_doc" not in st.session_state:
|
106 |
+
st.session_state.process_doc = False
|
107 |
+
|
108 |
+
|
109 |
+
OPENAI_API_KEY = "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"
|
110 |
+
os.environ["OPENAI_API_KEY"] = OPENAI_API_KEY
|
111 |
+
|
112 |
+
|
113 |
+
if OPENAI_API_KEY:
|
114 |
+
pptx_files = st.sidebar.file_uploader("Upload the case report in PDF format", type="pptx")
|
115 |
+
st.sidebar.info("""
|
116 |
+
Example pdf reports you can upload here:
|
117 |
+
""")
|
118 |
+
|
119 |
+
if st.sidebar.button("Process Document"):
|
120 |
+
with st.spinner("Processing Document..."):
|
121 |
+
nodes = process_pptx(pptx_files)
|
122 |
+
st.session_state.index = get_vector_index(nodes, vector_store="faiss")
|
123 |
+
#st.session_state.index = get_vector_index(nodes, vector_store="simple")
|
124 |
+
st.session_state.process_doc = True
|
125 |
+
st.toast("Document Processsed!")
|
126 |
+
|
127 |
+
st.session_state.process_doc = True
|
128 |
+
|
129 |
+
if st.session_state.process_doc:
|
130 |
+
search_text = st.text_input("Enter your question")
|
131 |
+
if st.button("Submit"):
|
132 |
+
engine = get_query_engine(st.session_state.index.as_query_engine(similarity_top_k=3))
|
133 |
+
start_time = time.time()
|
134 |
+
|
135 |
+
st.write("Alert search result...")
|
136 |
+
response = generate_insight(engine, search_text)
|
137 |
+
st.session_state["end_time"] = "{:.2f}".format((time.time() - start_time))
|
138 |
+
|
139 |
+
st.toast("Report Analysis Complete!")
|
140 |
+
|
141 |
+
if st.session_state.end_time:
|
142 |
+
st.write("Report Analysis Time: ", st.session_state.end_time, "s")
|
143 |
+
|
pages/tc1.ppy
ADDED
@@ -0,0 +1,129 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
|
3 |
+
from llama_index import VectorStoreIndex, SimpleDirectoryReader, ServiceContext
|
4 |
+
from llama_index import download_loader, SimpleDirectoryReader, StorageContext, load_index_from_storage
|
5 |
+
from llama_index.llms import HuggingFaceLLM
|
6 |
+
from llama_index.embeddings import HuggingFaceEmbedding
|
7 |
+
|
8 |
+
from IPython.display import Markdown, display
|
9 |
+
|
10 |
+
import chromadb
|
11 |
+
|
12 |
+
import streamlit as st
|
13 |
+
import time
|
14 |
+
from pypdf import PdfReader
|
15 |
+
|
16 |
+
from pathlib import Path
|
17 |
+
|
18 |
+
import os
|
19 |
+
|
20 |
+
import torch
|
21 |
+
#torch.set_default_device('cuda')
|
22 |
+
|
23 |
+
|
24 |
+
st.set_page_config(page_title="Tesla Case Analyzer", page_icon=":card_index_dividers:", initial_sidebar_state="expanded", layout="wide")
|
25 |
+
|
26 |
+
st.title(":card_index_dividers: Tesla Case Analyzer")
|
27 |
+
st.info("""
|
28 |
+
Begin by uploading the case report in pptx format. Afterward, click on 'Process Document'. Once the document has been processed. You can enter question and click send, system will answer your question.
|
29 |
+
""")
|
30 |
+
|
31 |
+
if "process_doc" not in st.session_state:
|
32 |
+
st.session_state.process_doc = False
|
33 |
+
|
34 |
+
|
35 |
+
|
36 |
+
llm = HuggingFaceLLM(
|
37 |
+
context_window=8000,
|
38 |
+
max_new_tokens=256,
|
39 |
+
generate_kwargs={"temperature": 0.1, "do_sample": True},
|
40 |
+
system_prompt=system_prompt,
|
41 |
+
query_wrapper_prompt=query_wrapper_prompt,
|
42 |
+
tokenizer_name="mistralai/Mistral-7B-Instruct-v0.1",
|
43 |
+
model_name="mistralai/Mistral-7B-Instruct-v0.1",
|
44 |
+
device_map="auto",
|
45 |
+
tokenizer_kwargs={"max_length": 8000},
|
46 |
+
model_kwargs={"torch_dtype": torch.float16}
|
47 |
+
)
|
48 |
+
|
49 |
+
embed_model = HuggingFaceEmbedding(model_name="thenlper/gte-base")
|
50 |
+
|
51 |
+
service_context = ServiceContext.from_defaults(
|
52 |
+
chunk_size=1024,
|
53 |
+
llm=llm,
|
54 |
+
embed_model=embed_model
|
55 |
+
)
|
56 |
+
|
57 |
+
|
58 |
+
files_uploaded = st.sidebar.file_uploader("Upload the case report in pptx format", type="pptx",accept_multiple_files=True)
|
59 |
+
st.sidebar.info("""
|
60 |
+
Example pdf reports you can upload here:
|
61 |
+
""")
|
62 |
+
|
63 |
+
if st.sidebar.button("Process Document"):
|
64 |
+
with st.spinner("Processing Document..."):
|
65 |
+
|
66 |
+
data_dir = './data'
|
67 |
+
if not os.path.exists(data_dir):
|
68 |
+
os.makedirs(data_dir)
|
69 |
+
|
70 |
+
for pdf in files_uploaded:
|
71 |
+
print(f'file named {pdf.name}')
|
72 |
+
fname=f'{data_dir}/{pdf.name}'
|
73 |
+
with open(fname, 'wb') as f:
|
74 |
+
f.write(pdf.read())
|
75 |
+
|
76 |
+
|
77 |
+
def fmetadata(dummy: str): return {"file_path": ""}
|
78 |
+
|
79 |
+
PptxReader = download_loader("PptxReader")
|
80 |
+
loader = SimpleDirectoryReader(input_dir=data_dir, file_extractor={".pptx": PptxReader(),}, file_metadata=fmetadata)
|
81 |
+
|
82 |
+
documents = loader.load_data()
|
83 |
+
for doc in documents:
|
84 |
+
doc.metadata["file_path"]=""
|
85 |
+
|
86 |
+
print('stored')
|
87 |
+
|
88 |
+
st.session_state.process_doc = True
|
89 |
+
|
90 |
+
st.toast("Document Processsed!")
|
91 |
+
|
92 |
+
#st.session_state.process_doc = True
|
93 |
+
|
94 |
+
OPENAI_API_KEY = "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"
|
95 |
+
os.environ["OPENAI_API_KEY"] = OPENAI_API_KEY
|
96 |
+
|
97 |
+
|
98 |
+
if OPENAI_API_KEY:
|
99 |
+
pdfs = st.sidebar.file_uploader("Upload the case report in PDF format", type="pdf")
|
100 |
+
st.sidebar.info("""
|
101 |
+
Example pdf reports you can upload here:
|
102 |
+
""")
|
103 |
+
|
104 |
+
if st.sidebar.button("Process Document"):
|
105 |
+
with st.spinner("Processing Document..."):
|
106 |
+
nodes = process_pdf(pdfs)
|
107 |
+
#st.session_state.index = get_vector_index(nodes, vector_store="faiss")
|
108 |
+
st.session_state.index = get_vector_index(nodes, vector_store="simple")
|
109 |
+
st.session_state.process_doc = True
|
110 |
+
st.toast("Document Processsed!")
|
111 |
+
|
112 |
+
#st.session_state.process_doc = True
|
113 |
+
|
114 |
+
if st.session_state.process_doc:
|
115 |
+
search_text = st.text_input("Enter your question")
|
116 |
+
if st.button("Submit"):
|
117 |
+
engine = get_query_engine(st.session_state.index.as_query_engine(similarity_top_k=3))
|
118 |
+
start_time = time.time()
|
119 |
+
|
120 |
+
with st.status("**Analyzing Report...**"):
|
121 |
+
st.write("Case search result...")
|
122 |
+
response = generate_insight(engine, search_text)
|
123 |
+
st.session_state["end_time"] = "{:.2f}".format((time.time() - start_time))
|
124 |
+
|
125 |
+
st.toast("Report Analysis Complete!")
|
126 |
+
|
127 |
+
if st.session_state.end_time:
|
128 |
+
st.write("Report Analysis Time: ", st.session_state.end_time, "s")
|
129 |
+
|