Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -1,95 +1,98 @@
|
|
|
|
1 |
import streamlit as st
|
2 |
import pandas as pd
|
3 |
-
import os
|
4 |
-
import tempfile
|
5 |
-
from PyPDF2 import PdfReader
|
6 |
-
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
7 |
-
from sentence_transformers import SentenceTransformer
|
8 |
-
import faiss
|
9 |
import openai
|
|
|
|
|
|
|
|
|
|
|
10 |
|
11 |
-
#
|
12 |
-
|
13 |
-
|
14 |
-
|
15 |
-
# Initialize FAISS and embedding model
|
16 |
-
embedding_model = SentenceTransformer('all-MiniLM-L6-v2')
|
17 |
-
faiss_index = None
|
18 |
-
data_chunks = []
|
19 |
-
chunk_mapping = {}
|
20 |
-
|
21 |
-
# File Upload and Processing
|
22 |
-
def load_files(uploaded_files):
|
23 |
-
global data_chunks, chunk_mapping, faiss_index
|
24 |
-
data_chunks = []
|
25 |
-
chunk_mapping = {}
|
26 |
-
for uploaded_file in uploaded_files:
|
27 |
-
file_type = uploaded_file.name.split('.')[-1]
|
28 |
-
with tempfile.NamedTemporaryFile(delete=False) as tmp_file:
|
29 |
-
tmp_file.write(uploaded_file.read())
|
30 |
-
tmp_file_path = tmp_file.name
|
31 |
-
|
32 |
-
if file_type == "csv":
|
33 |
-
df = pd.read_csv(tmp_file_path)
|
34 |
-
content = "\n".join(df.astype(str).values.flatten())
|
35 |
-
elif file_type == "xlsx":
|
36 |
-
df = pd.read_excel(tmp_file_path)
|
37 |
-
content = "\n".join(df.astype(str).values.flatten())
|
38 |
-
elif file_type == "pdf":
|
39 |
-
reader = PdfReader(tmp_file_path)
|
40 |
-
content = "".join([page.extract_text() for page in reader.pages])
|
41 |
-
else:
|
42 |
-
st.error(f"Unsupported file type: {file_type}")
|
43 |
-
continue
|
44 |
-
|
45 |
-
# Split into chunks
|
46 |
-
splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50)
|
47 |
-
chunks = splitter.split_text(content)
|
48 |
-
data_chunks.extend(chunks)
|
49 |
-
chunk_mapping.update({i: (uploaded_file.name, chunk) for i, chunk in enumerate(chunks)})
|
50 |
-
|
51 |
-
# Create FAISS index
|
52 |
-
embeddings = embedding_model.encode(data_chunks)
|
53 |
-
faiss_index = faiss.IndexFlatL2(embeddings.shape[1])
|
54 |
-
faiss_index.add(embeddings)
|
55 |
|
56 |
-
|
57 |
-
def handle_query(query):
|
58 |
-
if not faiss_index:
|
59 |
-
return "No data available. Please upload files first."
|
60 |
|
61 |
-
|
62 |
-
|
63 |
-
distances, indices = faiss_index.search(query_embedding, k=5)
|
64 |
-
relevant_chunks = [chunk_mapping[idx][1] for idx in indices[0]]
|
65 |
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
|
|
|
74 |
|
75 |
-
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
-
|
|
|
|
|
|
|
80 |
|
81 |
-
|
82 |
-
|
83 |
-
|
|
|
|
|
|
|
|
|
|
|
84 |
|
85 |
-
|
86 |
-
|
87 |
-
if openai.api_key and query:
|
88 |
-
answer = handle_query(query)
|
89 |
-
st.subheader("Answer:")
|
90 |
-
st.write(answer)
|
91 |
-
else:
|
92 |
-
st.error("Please provide a valid API key and query.")
|
93 |
|
94 |
-
if
|
95 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
import streamlit as st
|
3 |
import pandas as pd
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
import openai
|
5 |
+
import torch
|
6 |
+
import matplotlib.pyplot as plt
|
7 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
|
8 |
+
from dotenv import load_dotenv
|
9 |
+
import anthropic
|
10 |
|
11 |
+
# Load environment variables
|
12 |
+
load_dotenv()
|
13 |
+
os.environ["OPENAI_API_KEY"] = os.getenv("OPENAI_API_KEY")
|
14 |
+
os.environ["ANTHROPIC_API_KEY"] = os.getenv("ANTHROPIC_API_KEY")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
15 |
|
16 |
+
st.title("Excel Q&A Chatbot 📊")
|
|
|
|
|
|
|
17 |
|
18 |
+
# Model Selection
|
19 |
+
model_choice = st.selectbox("Select LLM Model", ["OpenAI GPT-3.5", "Claude 3 Haiku", "Mistral-7B"])
|
|
|
|
|
20 |
|
21 |
+
# Load appropriate model based on selection
|
22 |
+
if model_choice == "Mistral-7B":
|
23 |
+
model_name = "mistralai/Mistral-7B-Instruct"
|
24 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
25 |
+
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float16)
|
26 |
+
def ask_mistral(query):
|
27 |
+
inputs = tokenizer(query, return_tensors="pt").to("cuda")
|
28 |
+
output = model.generate(**inputs)
|
29 |
+
return tokenizer.decode(output[0])
|
30 |
|
31 |
+
elif model_choice == "Claude 3 Haiku":
|
32 |
+
client = anthropic.Anthropic(api_key=os.environ["ANTHROPIC_API_KEY"])
|
33 |
+
def ask_claude(query):
|
34 |
+
response = client.messages.create(
|
35 |
+
model="claude-3-haiku",
|
36 |
+
messages=[{"role": "user", "content": query}]
|
37 |
+
)
|
38 |
+
return response.content
|
39 |
|
40 |
+
else:
|
41 |
+
client = openai.OpenAI()
|
42 |
+
def ask_gpt(query):
|
43 |
+
response = client.chat.completions.create(
|
44 |
+
model="gpt-3.5-turbo",
|
45 |
+
messages=[{"role": "user", "content": query}]
|
46 |
+
)
|
47 |
+
return response.choices[0].message.content
|
48 |
|
49 |
+
# File Upload
|
50 |
+
uploaded_file = st.file_uploader("Upload an Excel file", type=["csv", "xlsx"])
|
|
|
|
|
|
|
|
|
|
|
|
|
51 |
|
52 |
+
if uploaded_file is not None:
|
53 |
+
file_extension = uploaded_file.name.split(".")[-1].lower()
|
54 |
+
df = pd.read_csv(uploaded_file) if file_extension == "csv" else pd.read_excel(uploaded_file)
|
55 |
+
st.write("### Preview of Data:")
|
56 |
+
st.write(df.head())
|
57 |
+
|
58 |
+
# Extract metadata
|
59 |
+
column_names = df.columns.tolist()
|
60 |
+
data_types = df.dtypes.apply(lambda x: x.name).to_dict()
|
61 |
+
missing_values = df.isnull().sum().to_dict()
|
62 |
+
|
63 |
+
# Display metadata
|
64 |
+
st.write("### Column Details:")
|
65 |
+
st.write(pd.DataFrame({"Column": column_names, "Type": data_types.values(), "Missing Values": missing_values.values()}))
|
66 |
+
|
67 |
+
# User Query
|
68 |
+
query = st.text_input("Ask a question about this data:")
|
69 |
+
|
70 |
+
if st.button("Submit Query"):
|
71 |
+
if query:
|
72 |
+
# Interpret the query using selected LLM
|
73 |
+
if model_choice == "Mistral-7B":
|
74 |
+
parsed_query = ask_mistral(f"Convert this question into a Pandas operation: {query}")
|
75 |
+
elif model_choice == "Claude 3 Haiku":
|
76 |
+
parsed_query = ask_claude(f"Convert this question into a Pandas operation: {query}")
|
77 |
+
else:
|
78 |
+
parsed_query = ask_gpt(f"Convert this question into a Pandas operation: {query}")
|
79 |
+
|
80 |
+
# Execute the query
|
81 |
+
try:
|
82 |
+
result = eval(f"df.{parsed_query}")
|
83 |
+
st.write("### Result:")
|
84 |
+
st.write(result if isinstance(result, pd.DataFrame) else str(result))
|
85 |
+
|
86 |
+
# If numerical data, show a visualization
|
87 |
+
if isinstance(result, pd.Series) and result.dtype in ["int64", "float64"]:
|
88 |
+
fig, ax = plt.subplots()
|
89 |
+
result.plot(kind="bar", ax=ax)
|
90 |
+
st.pyplot(fig)
|
91 |
+
|
92 |
+
except Exception as e:
|
93 |
+
st.error(f"Error executing query: {str(e)}")
|
94 |
+
|
95 |
+
# Memory for context retention
|
96 |
+
if "query_history" not in st.session_state:
|
97 |
+
st.session_state.query_history = []
|
98 |
+
st.session_state.query_history.append(query)
|