Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,341 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import json
|
3 |
+
import requests
|
4 |
+
import gradio as gr
|
5 |
+
import threading
|
6 |
+
import time
|
7 |
+
import PyPDF2
|
8 |
+
import chromadb
|
9 |
+
import shutil
|
10 |
+
from pydantic import BaseModel, Field
|
11 |
+
from typing import Dict
|
12 |
+
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
13 |
+
from langchain_huggingface import HuggingFaceEmbeddings
|
14 |
+
|
15 |
+
|
16 |
+
API_KEY = os.getenv("mistral")
|
17 |
+
BASE_URL = "https://api.together.xyz"
|
18 |
+
|
19 |
+
# Store user inputs
|
20 |
+
user_inputs = {
|
21 |
+
"organization": "",
|
22 |
+
"rules_l1": "",
|
23 |
+
"rules_l2": "",
|
24 |
+
"rules_l3": "",
|
25 |
+
}
|
26 |
+
|
27 |
+
# Function to classify query
|
28 |
+
def classify_query(query: str) -> Dict:
|
29 |
+
if not all(user_inputs.values()):
|
30 |
+
raise ValueError("Please fill all input fields first.")
|
31 |
+
|
32 |
+
messages = [
|
33 |
+
{"role": "system", "content": f"""You are a Customer Query Classification Agent for {user_inputs["organization"]}.
|
34 |
+
What is considered Level 1 Query (Requires no account info just provided documents by the admin is enough to answer):
|
35 |
+
{user_inputs["rules_l1"]}
|
36 |
+
What is considered Level 2 Query (Requires account info and provided documents by the admin is enough to answer):
|
37 |
+
{user_inputs["rules_l2"]}
|
38 |
+
What is considered as Level 3 Query (Immediate Escalation to Human Customer Service Agents):
|
39 |
+
{user_inputs["rules_l3"]}
|
40 |
+
Classify the following customer query and provide the output in JSON format:
|
41 |
+
```json
|
42 |
+
{{
|
43 |
+
"title": "title of the query in under 10 words",
|
44 |
+
"level": "1 or 2 or 3"
|
45 |
+
}}
|
46 |
+
```"""},
|
47 |
+
|
48 |
+
{"role": "user", "content": query}
|
49 |
+
]
|
50 |
+
|
51 |
+
headers = {
|
52 |
+
"Content-Type": "application/json",
|
53 |
+
"Authorization": f"Bearer {API_KEY}"
|
54 |
+
}
|
55 |
+
|
56 |
+
data = {
|
57 |
+
"model": "mistralai/Mixtral-8x7B-Instruct-v0.1",
|
58 |
+
"messages": messages,
|
59 |
+
"temperature": 0.7,
|
60 |
+
"response_format": {
|
61 |
+
"type": "json_object",
|
62 |
+
"schema": {
|
63 |
+
"type": "object",
|
64 |
+
"properties": {
|
65 |
+
"title": {"type": "string"},
|
66 |
+
"level": {"type": "integer"}
|
67 |
+
},
|
68 |
+
"required": ["title", "level"]
|
69 |
+
}
|
70 |
+
}
|
71 |
+
}
|
72 |
+
|
73 |
+
response = requests.post(f"{BASE_URL}/chat/completions", headers=headers, json=data)
|
74 |
+
response.raise_for_status()
|
75 |
+
classification_result = response.json().get('choices')[0].get('message').get('content')
|
76 |
+
return classification_result
|
77 |
+
|
78 |
+
# Function to convert PDF to text
|
79 |
+
def pdf_to_text(file_path):
|
80 |
+
pdf_file = open(file_path, 'rb')
|
81 |
+
pdf_reader = PyPDF2.PdfReader(pdf_file)
|
82 |
+
text = ""
|
83 |
+
for page_num in range(len(pdf_reader.pages)):
|
84 |
+
text += pdf_reader.pages[page_num].extract_text()
|
85 |
+
pdf_file.close()
|
86 |
+
return text
|
87 |
+
|
88 |
+
# Function to handle file upload and save embeddings to ChromaDB
|
89 |
+
def handle_file_upload(files, collection_name):
|
90 |
+
if not collection_name:
|
91 |
+
return "Please provide a collection name."
|
92 |
+
|
93 |
+
os.makedirs('chabot_pdfs', exist_ok=True)
|
94 |
+
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=100)
|
95 |
+
embeddings = HuggingFaceEmbeddings(model_name="thenlper/gte-small")
|
96 |
+
|
97 |
+
# Initialize Chroma DB client
|
98 |
+
client = chromadb.PersistentClient(path="./db")
|
99 |
+
try:
|
100 |
+
collection = client.create_collection(name=collection_name)
|
101 |
+
except ValueError as e:
|
102 |
+
return f"Error creating collection: {str(e)}. Please try a different collection name."
|
103 |
+
|
104 |
+
for file in files:
|
105 |
+
file_name = os.path.basename(file.name)
|
106 |
+
file_path = os.path.join('chabot_pdfs', file_name)
|
107 |
+
shutil.copy(file.name, file_path) # Copy the file instead of saving
|
108 |
+
text = pdf_to_text(file_path)
|
109 |
+
chunks = text_splitter.split_text(text)
|
110 |
+
|
111 |
+
documents_list = []
|
112 |
+
embeddings_list = []
|
113 |
+
ids_list = []
|
114 |
+
|
115 |
+
for i, chunk in enumerate(chunks):
|
116 |
+
vector = embeddings.embed_query(chunk)
|
117 |
+
documents_list.append(chunk)
|
118 |
+
embeddings_list.append(vector)
|
119 |
+
ids_list.append(f"{file_name}_{i}")
|
120 |
+
|
121 |
+
collection.add(
|
122 |
+
embeddings=embeddings_list,
|
123 |
+
documents=documents_list,
|
124 |
+
ids=ids_list
|
125 |
+
)
|
126 |
+
|
127 |
+
return "Files uploaded and processed successfully."
|
128 |
+
|
129 |
+
# Function to search vector database
|
130 |
+
def search_vector_database(query, collection_name):
|
131 |
+
if not collection_name:
|
132 |
+
return "Please provide a collection name."
|
133 |
+
|
134 |
+
embeddings = HuggingFaceEmbeddings(model_name="thenlper/gte-small")
|
135 |
+
client = chromadb.PersistentClient(path="./db")
|
136 |
+
try:
|
137 |
+
collection = client.get_collection(name=collection_name)
|
138 |
+
except ValueError as e:
|
139 |
+
return f"Error accessing collection: {str(e)}. Make sure the collection name is correct."
|
140 |
+
|
141 |
+
query_vector = embeddings.embed_query(query)
|
142 |
+
results = collection.query(query_embeddings=[query_vector], n_results=2, include=["documents"])
|
143 |
+
|
144 |
+
return "\n\n".join("\n".join(result) for result in results["documents"])
|
145 |
+
|
146 |
+
# New function to handle login
|
147 |
+
def handle_login(username, password):
|
148 |
+
# This is a simple example. In a real application, you'd want to use secure authentication methods.
|
149 |
+
if username == "admin" and password == "password":
|
150 |
+
return """
|
151 |
+
"NeoBank": {
|
152 |
+
"user_id": "NB782940",
|
153 |
+
"user_name": "john_doe123",
|
154 |
+
"full_name": "John Doe",
|
155 |
+
"email": "[email protected]",
|
156 |
+
"balance": 2875.43,
|
157 |
+
"transactions": [
|
158 |
+
{"date": "2024-06-20", "description": "Coffee Shop", "amount": -4.50},
|
159 |
+
{"date": "2024-06-19", "description": "Grocery Store", "amount": -85.22},
|
160 |
+
{"date": "2024-06-18", "description": "Salary Deposit", "amount": 2500.00}
|
161 |
+
]
|
162 |
+
},
|
163 |
+
"CryptoInvest": {
|
164 |
+
"user_id": "CI549217",
|
165 |
+
"user_name": "crypto_enthusiast",
|
166 |
+
"full_name": "Alice Johnson",
|
167 |
+
"email": "[email protected]",
|
168 |
+
"portfolio": {
|
169 |
+
"BTC": {"amount": 0.025, "value": 7500.00},
|
170 |
+
"ETH": {"amount": 1.2, "value": 2100.00},
|
171 |
+
"SOL": {"amount": 5.8, "value": 450.50}
|
172 |
+
},
|
173 |
+
"transactions": [
|
174 |
+
{"date": "2024-06-22", "description": "Bought ETH", "amount": -500.00},
|
175 |
+
{"date": "2024-06-20", "description": "Sold BTC", "amount": 1200.00}
|
176 |
+
]
|
177 |
+
},
|
178 |
+
"RoboAdvisor": {
|
179 |
+
"user_id": "RA385712",
|
180 |
+
"user_name": "jane_smith",
|
181 |
+
"full_name": "Jane Smith",
|
182 |
+
"email": "[email protected]",
|
183 |
+
"risk_tolerance": "moderate",
|
184 |
+
"portfolio_value": 15800.75,
|
185 |
+
"allocations": {
|
186 |
+
"stocks": 0.60,
|
187 |
+
"bonds": 0.30,
|
188 |
+
"real_estate": 0.10
|
189 |
+
},
|
190 |
+
"recent_activity": [
|
191 |
+
{"date": "2024-06-21", "description": "Dividends received", "amount": 32.50},
|
192 |
+
{"date": "2024-06-15", "description": "Portfolio rebalanced" }
|
193 |
+
]
|
194 |
+
},
|
195 |
+
"PeerLend": {
|
196 |
+
"user_id": "PL916350",
|
197 |
+
"user_name": "bob_williams",
|
198 |
+
"full_name": "Bob Williams",
|
199 |
+
"email": "[email protected]",
|
200 |
+
"account_type": "borrower",
|
201 |
+
"loan_amount": 5000.00,
|
202 |
+
"interest_rate": 7.8,
|
203 |
+
"monthly_payment": 150.30,
|
204 |
+
"payment_history": [
|
205 |
+
{"date": "2024-06-22", "status": "paid"},
|
206 |
+
{"date": "2024-05-22", "status": "paid"},
|
207 |
+
{"date": "2024-04-22", "status": "paid"}
|
208 |
+
]
|
209 |
+
},
|
210 |
+
"InsureTech": {
|
211 |
+
"user_id": "IT264805",
|
212 |
+
"user_name": "eva_brown4",
|
213 |
+
"full_name": "Eva Brown",
|
214 |
+
"email": "[email protected]",
|
215 |
+
"policy_type": "auto",
|
216 |
+
"coverage_details": {
|
217 |
+
"liability": "50/100/50",
|
218 |
+
"collision": "500 deductible",
|
219 |
+
"comprehensive": "100 deductible"
|
220 |
+
},
|
221 |
+
"premium": 85.50,
|
222 |
+
"next_payment": "2024-07-10",
|
223 |
+
"claims": []
|
224 |
+
}
|
225 |
+
"""
|
226 |
+
else:
|
227 |
+
return "Invalid username or password"
|
228 |
+
|
229 |
+
# Gradio interface
|
230 |
+
def gradio_interface():
|
231 |
+
with gr.Blocks(theme='gl198976/The-Rounded') as interface:
|
232 |
+
gr.Markdown("# Admin Dashboard🧖🏻♀️")
|
233 |
+
|
234 |
+
with gr.Tab("Query Classifier Agent"):
|
235 |
+
with gr.Row():
|
236 |
+
with gr.Column():
|
237 |
+
organization_input = gr.Textbox(label="Organization Name")
|
238 |
+
rules_l1_input = gr.Textbox(label="Rules for Level 1 Query", lines=5)
|
239 |
+
rules_l2_input = gr.Textbox(label="Rules for Level 2 Query", lines=5)
|
240 |
+
rules_l3_input = gr.Textbox(label="Rules for Level 3 Query", lines=5)
|
241 |
+
submit_btn = gr.Button("Submit Rules")
|
242 |
+
with gr.Column():
|
243 |
+
query_input = gr.Textbox(label="Customer Query")
|
244 |
+
classification_output = gr.Textbox(label="Classification Result")
|
245 |
+
classify_btn = gr.Button("Classify Query")
|
246 |
+
api_details = gr.Markdown("""
|
247 |
+
### API Endpoint Details
|
248 |
+
- **URL:** `http://0.0.0.0:7860/classify`
|
249 |
+
- **Method:** POST
|
250 |
+
- **Request Body:** JSON with a single key `query`
|
251 |
+
- **Example Usage:**
|
252 |
+
```python
|
253 |
+
from gradio_client import Client
|
254 |
+
|
255 |
+
client = Client("http://0.0.0.0:7860/")
|
256 |
+
result = client.predict(
|
257 |
+
"Hello!!", # str in 'Customer Query' Textbox component
|
258 |
+
api_name="/classify_and_display"
|
259 |
+
)
|
260 |
+
print(result)
|
261 |
+
```
|
262 |
+
""")
|
263 |
+
|
264 |
+
submit_btn.click(lambda org, r1, r2, r3: (
|
265 |
+
setattr(user_inputs, "organization", org),
|
266 |
+
setattr(user_inputs, "rules_l1", r1),
|
267 |
+
setattr(user_inputs, "rules_l2", r2),
|
268 |
+
setattr(user_inputs, "rules_l3", r3)
|
269 |
+
), inputs=[organization_input, rules_l1_input, rules_l2_input, rules_l3_input])
|
270 |
+
|
271 |
+
classify_btn.click(classify_query, inputs=[query_input], outputs=[classification_output])
|
272 |
+
|
273 |
+
with gr.Tab("Organization Documentation Agent"):
|
274 |
+
gr.Markdown("""
|
275 |
+
### Warning
|
276 |
+
If you encounter an error when uploading files, try changing the collection name and upload again.
|
277 |
+
Each collection name must be unique.
|
278 |
+
""")
|
279 |
+
with gr.Row():
|
280 |
+
with gr.Column():
|
281 |
+
collection_name_input = gr.Textbox(label="Collection Name", placeholder="Enter a unique name for this collection")
|
282 |
+
file_upload = gr.Files(file_types=[".pdf"], label="Upload PDFs")
|
283 |
+
upload_btn = gr.Button("Upload and Process Files")
|
284 |
+
upload_status = gr.Textbox(label="Upload Status", interactive=False)
|
285 |
+
with gr.Column():
|
286 |
+
search_query_input = gr.Textbox(label="Search Query")
|
287 |
+
search_output = gr.Textbox(label="Search Results", lines=10)
|
288 |
+
search_btn = gr.Button("Search")
|
289 |
+
api_details = gr.Markdown("""
|
290 |
+
### API Endpoint Details
|
291 |
+
- **URL:** `http://0.0.0.0:7860/search_vector_database`
|
292 |
+
- **Method:** POST
|
293 |
+
- **Example Usage:**
|
294 |
+
```python
|
295 |
+
from gradio_client import Client
|
296 |
+
|
297 |
+
client = Client("http://0.0.0.0:7860/")
|
298 |
+
result = client.predict(
|
299 |
+
"search query", # str in 'Search Query' Textbox component
|
300 |
+
"name of collection given in ui", # str in 'Collection Name' Textbox component
|
301 |
+
api_name="/search_vector_database"
|
302 |
+
)
|
303 |
+
print(result)
|
304 |
+
```
|
305 |
+
""")
|
306 |
+
|
307 |
+
upload_btn.click(handle_file_upload, inputs=[file_upload, collection_name_input], outputs=[upload_status])
|
308 |
+
search_btn.click(search_vector_database, inputs=[search_query_input, collection_name_input], outputs=[search_output])
|
309 |
+
|
310 |
+
with gr.Tab("Account Information"):
|
311 |
+
with gr.Row():
|
312 |
+
with gr.Column():
|
313 |
+
username_input = gr.Textbox(label="Username")
|
314 |
+
password_input = gr.Textbox(label="Password", type="password")
|
315 |
+
login_btn = gr.Button("Login")
|
316 |
+
with gr.Column():
|
317 |
+
account_info_output = gr.Textbox(label="Account Info", lines=20)
|
318 |
+
api_details = gr.Markdown("""
|
319 |
+
### API Endpoint Details
|
320 |
+
- **URL:** `http://0.0.0.0:7860/handle_login`
|
321 |
+
- **Method:** POST
|
322 |
+
- **Example Usage:**
|
323 |
+
```python
|
324 |
+
from gradio_client import Client
|
325 |
+
|
326 |
+
client = Client("http://0.0.0.0:7860/")
|
327 |
+
result = client.predict(
|
328 |
+
"admin", # str in 'Username' Textbox component
|
329 |
+
"password", # str in 'Password' Textbox component
|
330 |
+
api_name="/handle_login"
|
331 |
+
)
|
332 |
+
print(result)
|
333 |
+
```
|
334 |
+
""")
|
335 |
+
|
336 |
+
login_btn.click(handle_login, inputs=[username_input, password_input], outputs=[account_info_output])
|
337 |
+
|
338 |
+
interface.launch(server_name="0.0.0.0", server_port=7860)
|
339 |
+
|
340 |
+
if __name__ == "__main__":
|
341 |
+
gradio_interface()
|