Spaces:
Runtime error
Runtime error
nishantgaurav23
commited on
Commit
β’
6f7b9d9
1
Parent(s):
bc26371
Update app.py
Browse files
app.py
CHANGED
@@ -7,24 +7,14 @@ import torch
|
|
7 |
import torch.nn.functional as F
|
8 |
import re
|
9 |
import requests
|
10 |
-
#from dotenv import load_dotenv
|
11 |
from embedding_processor import SentenceTransformerRetriever, process_data
|
12 |
import pickle
|
13 |
-
|
14 |
-
import os
|
15 |
-
import warnings
|
16 |
-
import json # Add this import
|
17 |
-
|
18 |
-
# Add at the top with other imports
|
19 |
-
from llama_cpp import Llama
|
20 |
-
import requests
|
21 |
-
from tqdm import tqdm
|
22 |
-
|
23 |
-
|
24 |
import logging
|
25 |
import sys
|
|
|
|
|
26 |
|
27 |
-
# Set page config
|
28 |
st.set_page_config(
|
29 |
page_title="The Sport Chatbot",
|
30 |
page_icon="π",
|
@@ -38,16 +28,21 @@ logging.basicConfig(
|
|
38 |
handlers=[logging.StreamHandler(sys.stdout)]
|
39 |
)
|
40 |
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
|
|
|
|
|
|
|
|
|
|
51 |
|
52 |
@st.cache_data
|
53 |
def load_from_drive(file_id: str):
|
@@ -72,93 +67,72 @@ def load_from_drive(file_id: str):
|
|
72 |
st.error(f"Error loading file from Drive: {str(e)}")
|
73 |
return None
|
74 |
|
75 |
-
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
80 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
81 |
|
82 |
class RAGPipeline:
|
83 |
-
|
84 |
def __init__(self, data_folder: str, k: int = 5):
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
# Model path with absolute path
|
93 |
-
current_dir = os.path.dirname(os.path.abspath(__file__))
|
94 |
-
self.model_path = os.path.join(current_dir, "models", "mistral-7b-v0.1.Q4_K_M.gguf")
|
95 |
-
|
96 |
-
# Initialize model
|
97 |
-
self.llm = self.get_model()
|
98 |
-
|
99 |
-
except Exception as e:
|
100 |
-
logging.error(f"Error in RAGPipeline initialization: {str(e)}")
|
101 |
-
raise
|
102 |
-
|
103 |
-
@st.cache_resource(show_spinner=False)
|
104 |
-
def get_model(_self):
|
105 |
-
"""Get or initialize the model with caching"""
|
106 |
-
try:
|
107 |
-
if not os.path.exists(_self.model_path):
|
108 |
-
os.makedirs(os.path.dirname(_self.model_path), exist_ok=True)
|
109 |
-
st.info("Downloading model... This may take a while.")
|
110 |
-
direct_url = "https://huggingface.co/TheBloke/Mistral-7B-v0.1-GGUF/resolve/main/mistral-7b-v0.1.Q4_K_M.gguf"
|
111 |
-
_self.download_file_with_progress(direct_url, _self.model_path)
|
112 |
-
|
113 |
-
# Verify file exists and has content
|
114 |
-
if not os.path.exists(_self.model_path):
|
115 |
-
raise FileNotFoundError(f"Model file {_self.model_path} not found after download attempts")
|
116 |
-
|
117 |
-
if os.path.getsize(_self.model_path) < 1000000: # Less than 1MB
|
118 |
-
os.remove(_self.model_path)
|
119 |
-
raise ValueError("Downloaded model file is too small, likely corrupted")
|
120 |
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
126 |
-
"n_gpu_layers": 0,
|
127 |
-
"verbose": False
|
128 |
-
}
|
129 |
-
|
130 |
-
model = Llama(**llm_config)
|
131 |
-
st.success("Model loaded successfully!")
|
132 |
-
return model
|
133 |
-
|
134 |
-
except Exception as e:
|
135 |
-
st.error(f"Error initializing model: {str(e)}")
|
136 |
-
raise
|
137 |
|
138 |
-
def
|
139 |
-
"""
|
140 |
-
response =
|
141 |
-
|
142 |
-
|
143 |
-
|
144 |
-
desc=filename,
|
145 |
-
total=total_size,
|
146 |
-
unit='iB',
|
147 |
-
unit_scale=True,
|
148 |
-
unit_divisor=1024,
|
149 |
-
) as progress_bar:
|
150 |
-
for data in response.iter_content(chunk_size=1024):
|
151 |
-
size = file.write(data)
|
152 |
-
progress_bar.update(size)
|
153 |
|
154 |
-
# Alternative API call with streaming
|
155 |
def query_model(self, prompt: str) -> str:
|
156 |
-
"""Query the local Llama model
|
157 |
try:
|
158 |
if self.llm is None:
|
159 |
raise RuntimeError("Model not initialized")
|
160 |
-
|
161 |
-
# Generate response using Llama model
|
162 |
response = self.llm(
|
163 |
prompt,
|
164 |
max_tokens=512,
|
@@ -167,47 +141,41 @@ class RAGPipeline:
|
|
167 |
echo=False,
|
168 |
stop=["Question:", "\n\n"]
|
169 |
)
|
170 |
-
|
171 |
-
# Check and extract response
|
172 |
if response and 'choices' in response and len(response['choices']) > 0:
|
173 |
text = response['choices'][0].get('text', '').strip()
|
174 |
return text
|
175 |
else:
|
176 |
raise ValueError("No valid response generated")
|
177 |
-
|
178 |
except Exception as e:
|
179 |
logging.error(f"Error in query_model: {str(e)}")
|
180 |
raise
|
181 |
-
|
182 |
-
"""Clean and prepare the query"""
|
183 |
-
query = query.lower().strip()
|
184 |
-
query = re.sub(r'\s+', ' ', query)
|
185 |
-
return query
|
186 |
-
|
187 |
def process_query(self, query: str, placeholder) -> str:
|
188 |
try:
|
189 |
# Preprocess query
|
190 |
query = self.preprocess_query(query)
|
191 |
-
|
192 |
# Show retrieval status
|
193 |
status = placeholder.empty()
|
194 |
status.write("π Finding relevant information...")
|
195 |
-
|
196 |
# Get embeddings and search
|
197 |
query_embedding = self.retriever.encode([query])
|
198 |
similarities = F.cosine_similarity(query_embedding, self.retriever.doc_embeddings)
|
199 |
scores, indices = torch.topk(similarities, k=min(self.k, len(self.documents)))
|
200 |
-
|
201 |
relevant_docs = [self.documents[idx] for idx in indices.tolist()]
|
202 |
-
|
203 |
# Update status
|
204 |
status.write("π Generating response...")
|
205 |
-
|
206 |
# Prepare context and prompt
|
207 |
-
context = "\n".join(relevant_docs[:3])
|
208 |
prompt = f"""Context information is below:
|
209 |
{context}
|
210 |
-
|
211 |
Given the context above, please answer the following question:
|
212 |
{query}
|
213 |
|
@@ -217,12 +185,12 @@ class RAGPipeline:
|
|
217 |
- Only include sports-related information
|
218 |
- No dates or timestamps in the response
|
219 |
- Use clear, natural language
|
220 |
-
|
221 |
Answer:"""
|
222 |
-
|
223 |
# Generate response
|
224 |
response_placeholder = placeholder.empty()
|
225 |
-
|
226 |
try:
|
227 |
response_text = self.query_model(prompt)
|
228 |
if response_text:
|
@@ -233,174 +201,27 @@ class RAGPipeline:
|
|
233 |
message = "No relevant answer found. Please try rephrasing your question."
|
234 |
response_placeholder.warning(message)
|
235 |
return message
|
236 |
-
|
237 |
except Exception as e:
|
238 |
logging.error(f"Generation error: {str(e)}")
|
239 |
message = "Had some trouble generating the response. Please try again."
|
240 |
response_placeholder.warning(message)
|
241 |
return message
|
242 |
-
|
243 |
except Exception as e:
|
244 |
logging.error(f"Process error: {str(e)}")
|
245 |
message = "Something went wrong. Please try again with a different question."
|
246 |
placeholder.warning(message)
|
247 |
return message
|
248 |
|
249 |
-
def postprocess_response(self, response: str) -> str:
|
250 |
-
"""Clean up the generated response"""
|
251 |
-
response = response.strip()
|
252 |
-
response = re.sub(r'\s+', ' ', response)
|
253 |
-
response = re.sub(r'\d{4}-\d{2}-\d{2}\s\d{2}:\d{2}:\d{2}(?:\+\d{2}:?\d{2})?', '', response)
|
254 |
-
return response
|
255 |
-
|
256 |
-
|
257 |
-
# def process_query(self, query: str, placeholder) -> str:
|
258 |
-
# try:
|
259 |
-
# # Preprocess query
|
260 |
-
# query = self.preprocess_query(query)
|
261 |
-
|
262 |
-
# # Show retrieval status
|
263 |
-
# status = placeholder.empty()
|
264 |
-
# status.write("π Finding relevant information...")
|
265 |
-
|
266 |
-
# # Get embeddings and search using tensor operations
|
267 |
-
# query_embedding = self.retriever.encode([query])
|
268 |
-
# similarities = F.cosine_similarity(query_embedding, self.retriever.doc_embeddings)
|
269 |
-
# scores, indices = torch.topk(similarities, k=min(self.k, len(self.documents)))
|
270 |
-
|
271 |
-
# # Print search results for debugging
|
272 |
-
# print("\nSearch Results:")
|
273 |
-
# for idx, score in zip(indices.tolist(), scores.tolist()):
|
274 |
-
# print(f"Score: {score:.4f} | Document: {self.documents[idx][:100]}...")
|
275 |
-
|
276 |
-
# relevant_docs = [self.documents[idx] for idx in indices.tolist()]
|
277 |
-
|
278 |
-
# # Update status
|
279 |
-
# status.write("π Generating response...")
|
280 |
-
|
281 |
-
# # Prepare context and prompt
|
282 |
-
# context = "\n".join(relevant_docs[:3]) # Only use top 3 most relevant docs
|
283 |
-
# prompt = f"""Answer this question using the given context. Be specific and detailed.
|
284 |
-
|
285 |
-
# Context: {context}
|
286 |
-
|
287 |
-
# Question: {query}
|
288 |
-
|
289 |
-
# Answer (provide a complete, detailed response):"""
|
290 |
-
|
291 |
-
# # Generate response
|
292 |
-
# response_placeholder = placeholder.empty()
|
293 |
-
|
294 |
-
# try:
|
295 |
-
# response = requests.post(
|
296 |
-
# model_name,
|
297 |
-
# #headers=headers,
|
298 |
-
# json={
|
299 |
-
# "inputs": prompt,
|
300 |
-
# "parameters": {
|
301 |
-
# "max_new_tokens": 1024,
|
302 |
-
# "temperature": 0.5,
|
303 |
-
# "top_p": 0.9,
|
304 |
-
# "top_k": 50,
|
305 |
-
# "repetition_penalty": 1.03,
|
306 |
-
# "do_sample": True
|
307 |
-
# }
|
308 |
-
# },
|
309 |
-
# timeout=30
|
310 |
-
# ).json()
|
311 |
-
|
312 |
-
# if response and isinstance(response, list) and len(response) > 0:
|
313 |
-
# generated_text = response[0].get('generated_text', '').strip()
|
314 |
-
# if generated_text:
|
315 |
-
# # Find and extract only the answer part
|
316 |
-
# if "Answer:" in generated_text:
|
317 |
-
# answer_part = generated_text.split("Answer:")[-1].strip()
|
318 |
-
# elif "Answer (provide a complete, detailed response):" in generated_text:
|
319 |
-
# answer_part = generated_text.split("Answer (provide a complete, detailed response):")[-1].strip()
|
320 |
-
# else:
|
321 |
-
# answer_part = generated_text.strip()
|
322 |
-
|
323 |
-
# # Clean up the answer
|
324 |
-
# answer_part = answer_part.replace("Context:", "").replace("Question:", "")
|
325 |
-
|
326 |
-
# final_response = self.postprocess_response(answer_part)
|
327 |
-
# response_placeholder.markdown(final_response)
|
328 |
-
# return final_response
|
329 |
-
|
330 |
-
# message = "No relevant answer found. Please try rephrasing your question."
|
331 |
-
# response_placeholder.warning(message)
|
332 |
-
# return message
|
333 |
-
|
334 |
-
# except Exception as e:
|
335 |
-
# print(f"Generation error: {str(e)}")
|
336 |
-
# message = "Had some trouble generating the response. Please try again."
|
337 |
-
# response_placeholder.warning(message)
|
338 |
-
# return message
|
339 |
-
|
340 |
-
# except Exception as e:
|
341 |
-
# print(f"Process error: {str(e)}")
|
342 |
-
# message = "Something went wrong. Please try again with a different question."
|
343 |
-
# placeholder.warning(message)
|
344 |
-
# return message
|
345 |
-
def check_environment():
|
346 |
-
"""Check if the environment is properly set up"""
|
347 |
-
# if not headers['Authorization']:
|
348 |
-
# st.error("HUGGINGFACE_API_KEY environment variable not set!")
|
349 |
-
# st.stop()
|
350 |
-
# return False
|
351 |
-
|
352 |
-
try:
|
353 |
-
import torch
|
354 |
-
import sentence_transformers
|
355 |
-
return True
|
356 |
-
except ImportError as e:
|
357 |
-
st.error(f"Missing required package: {str(e)}")
|
358 |
-
st.stop()
|
359 |
-
return False
|
360 |
-
|
361 |
-
# @st.cache_resource
|
362 |
-
# def initialize_rag_pipeline():
|
363 |
-
# """Initialize the RAG pipeline once"""
|
364 |
-
# data_folder = "ESPN_data"
|
365 |
-
# return RAGPipeline(data_folder)
|
366 |
-
def check_space_requirements():
|
367 |
-
"""Check if we're running on HF Space and have necessary resources"""
|
368 |
-
try:
|
369 |
-
# Check if we're on HF Space
|
370 |
-
is_space = os.environ.get('SPACE_ID') is not None
|
371 |
-
|
372 |
-
if is_space:
|
373 |
-
# Check disk space
|
374 |
-
disk_space = os.statvfs('/')
|
375 |
-
free_space_gb = (disk_space.f_frsize * disk_space.f_bavail) / (1024**3)
|
376 |
-
|
377 |
-
if free_space_gb < 10: # Need at least 10GB free
|
378 |
-
st.warning(f"Low disk space: {free_space_gb:.1f}GB free")
|
379 |
-
|
380 |
-
# Check if model exists
|
381 |
-
model_path = "mistral-7b-v0.1.Q4_K_M.gguf"
|
382 |
-
if not os.path.exists(model_path):
|
383 |
-
st.info("Model will be downloaded on first run")
|
384 |
-
|
385 |
-
# Check if embeddings exist
|
386 |
-
if not os.path.exists('embeddings_cache/embeddings.pkl'):
|
387 |
-
st.info("Embeddings will be loaded from Drive")
|
388 |
-
|
389 |
-
return True
|
390 |
-
|
391 |
-
except Exception as e:
|
392 |
-
logging.error(f"Space requirements check failed: {str(e)}")
|
393 |
-
return False
|
394 |
-
|
395 |
@st.cache_resource(show_spinner=False)
|
396 |
def initialize_rag_pipeline():
|
397 |
"""Initialize the RAG pipeline once"""
|
398 |
try:
|
399 |
-
#
|
400 |
-
|
401 |
-
|
402 |
-
|
403 |
-
# Load embeddings from Drive first
|
404 |
drive_file_id = "1MuV63AE9o6zR9aBvdSDQOUextp71r2NN"
|
405 |
with st.spinner("Loading embeddings from Google Drive..."):
|
406 |
cache_data = load_from_drive(drive_file_id)
|
@@ -408,7 +229,7 @@ def initialize_rag_pipeline():
|
|
408 |
st.error("Failed to load embeddings from Google Drive")
|
409 |
st.stop()
|
410 |
|
411 |
-
#
|
412 |
data_folder = "ESPN_data"
|
413 |
rag = RAGPipeline(data_folder)
|
414 |
|
@@ -426,20 +247,9 @@ def initialize_rag_pipeline():
|
|
426 |
def main():
|
427 |
try:
|
428 |
# Environment check
|
429 |
-
if not check_environment()
|
430 |
return
|
431 |
|
432 |
-
# Session state for initialization status
|
433 |
-
if 'initialized' not in st.session_state:
|
434 |
-
st.session_state.initialized = False
|
435 |
-
|
436 |
-
# # Page config
|
437 |
-
# st.set_page_config(
|
438 |
-
# page_title="The Sport Chatbot",
|
439 |
-
# page_icon="π",
|
440 |
-
# layout="wide"
|
441 |
-
# )
|
442 |
-
|
443 |
# Improved CSS styling
|
444 |
st.markdown("""
|
445 |
<style>
|
@@ -510,7 +320,7 @@ def main():
|
|
510 |
</style>
|
511 |
""", unsafe_allow_html=True)
|
512 |
|
513 |
-
# Header section
|
514 |
st.markdown("<h1 class='main-title'>π The Sport Chatbot</h1>", unsafe_allow_html=True)
|
515 |
st.markdown("<h3 class='sub-title'>Using ESPN API</h3>", unsafe_allow_html=True)
|
516 |
st.markdown("""
|
@@ -523,40 +333,22 @@ def main():
|
|
523 |
</p>
|
524 |
""", unsafe_allow_html=True)
|
525 |
|
526 |
-
# Add some spacing
|
527 |
-
st.markdown("<br>", unsafe_allow_html=True)
|
528 |
-
|
529 |
# Initialize the pipeline
|
530 |
-
if not st.session_state
|
531 |
-
|
532 |
-
|
533 |
-
# Create necessary directories
|
534 |
-
for directory in ['models', 'ESPN_data', 'embeddings_cache']:
|
535 |
-
os.makedirs(directory, exist_ok=True)
|
536 |
-
|
537 |
-
# Initialize RAG pipeline
|
538 |
-
st.session_state.rag = initialize_rag_pipeline()
|
539 |
-
st.session_state.initialized = True
|
540 |
-
|
541 |
-
st.success("System initialized successfully!")
|
542 |
-
except Exception as e:
|
543 |
-
logging.error(f"Initialization error: {str(e)}")
|
544 |
-
st.error("Unable to initialize the system. Please check if all required files are present.")
|
545 |
-
st.stop()
|
546 |
|
547 |
-
# Create columns for layout
|
548 |
col1, col2, col3 = st.columns([1, 6, 1])
|
549 |
|
550 |
with col2:
|
551 |
-
# Query input
|
552 |
query = st.text_input("What would you like to know about sports?")
|
553 |
|
554 |
-
# Centered button
|
555 |
if st.button("Get Answer"):
|
556 |
if query:
|
557 |
response_placeholder = st.empty()
|
558 |
try:
|
559 |
-
# Get response from RAG pipeline
|
560 |
response = st.session_state.rag.process_query(query, response_placeholder)
|
561 |
logging.info(f"Generated response: {response}")
|
562 |
except Exception as e:
|
@@ -565,13 +357,12 @@ def main():
|
|
565 |
else:
|
566 |
st.warning("Please enter a question!")
|
567 |
|
568 |
-
# Footer
|
569 |
st.markdown("<br><br>", unsafe_allow_html=True)
|
570 |
st.markdown("---")
|
571 |
st.markdown("""
|
572 |
<p style='text-align: center; color: #666666; padding: 1rem 0;'>
|
573 |
-
Powered by ESPN Data & Mistral AI
|
574 |
-
<small>Running on Hugging Face Spaces</small>
|
575 |
</p>
|
576 |
""", unsafe_allow_html=True)
|
577 |
|
@@ -580,8 +371,4 @@ def main():
|
|
580 |
st.error("An unexpected error occurred. Please check the logs and try again.")
|
581 |
|
582 |
if __name__ == "__main__":
|
583 |
-
|
584 |
-
main()
|
585 |
-
except Exception as e:
|
586 |
-
logging.error(f"Application error: {str(e)}")
|
587 |
-
st.error("An unexpected error occurred. Please check the logs and try again.")
|
|
|
7 |
import torch.nn.functional as F
|
8 |
import re
|
9 |
import requests
|
|
|
10 |
from embedding_processor import SentenceTransformerRetriever, process_data
|
11 |
import pickle
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
import logging
|
13 |
import sys
|
14 |
+
from llama_cpp import Llama
|
15 |
+
from tqdm import tqdm
|
16 |
|
17 |
+
# Set page config first
|
18 |
st.set_page_config(
|
19 |
page_title="The Sport Chatbot",
|
20 |
page_icon="π",
|
|
|
28 |
handlers=[logging.StreamHandler(sys.stdout)]
|
29 |
)
|
30 |
|
31 |
+
def download_file_with_progress(url: str, filename: str):
|
32 |
+
"""Download a file with progress bar using requests"""
|
33 |
+
response = requests.get(url, stream=True)
|
34 |
+
total_size = int(response.headers.get('content-length', 0))
|
35 |
+
|
36 |
+
with open(filename, 'wb') as file, tqdm(
|
37 |
+
desc=filename,
|
38 |
+
total=total_size,
|
39 |
+
unit='iB',
|
40 |
+
unit_scale=True,
|
41 |
+
unit_divisor=1024,
|
42 |
+
) as progress_bar:
|
43 |
+
for data in response.iter_content(chunk_size=1024):
|
44 |
+
size = file.write(data)
|
45 |
+
progress_bar.update(size)
|
46 |
|
47 |
@st.cache_data
|
48 |
def load_from_drive(file_id: str):
|
|
|
67 |
st.error(f"Error loading file from Drive: {str(e)}")
|
68 |
return None
|
69 |
|
70 |
+
@st.cache_resource(show_spinner=False)
|
71 |
+
def load_llama_model():
|
72 |
+
"""Load Llama model with caching"""
|
73 |
+
try:
|
74 |
+
model_path = "mistral-7b-v0.1.Q4_K_M.gguf"
|
75 |
+
|
76 |
+
if not os.path.exists(model_path):
|
77 |
+
st.info("Downloading model... This may take a while.")
|
78 |
+
direct_url = "https://huggingface.co/TheBloke/Mistral-7B-v0.1-GGUF/resolve/main/mistral-7b-v0.1.Q4_K_M.gguf"
|
79 |
+
download_file_with_progress(direct_url, model_path)
|
80 |
+
|
81 |
+
llm_config = {
|
82 |
+
"model_path": model_path,
|
83 |
+
"n_ctx": 2048,
|
84 |
+
"n_threads": 4,
|
85 |
+
"n_batch": 512,
|
86 |
+
"n_gpu_layers": 0,
|
87 |
+
"verbose": False
|
88 |
+
}
|
89 |
+
|
90 |
+
model = Llama(**llm_config)
|
91 |
+
st.success("Model loaded successfully!")
|
92 |
+
return model
|
93 |
+
except Exception as e:
|
94 |
+
st.error(f"Error loading model: {str(e)}")
|
95 |
+
raise
|
96 |
|
97 |
+
def check_environment():
|
98 |
+
"""Check if the environment is properly set up"""
|
99 |
+
try:
|
100 |
+
import torch
|
101 |
+
import sentence_transformers
|
102 |
+
return True
|
103 |
+
except ImportError as e:
|
104 |
+
st.error(f"Missing required package: {str(e)}")
|
105 |
+
st.stop()
|
106 |
+
return False
|
107 |
|
108 |
class RAGPipeline:
|
|
|
109 |
def __init__(self, data_folder: str, k: int = 5):
|
110 |
+
self.data_folder = data_folder
|
111 |
+
self.k = k
|
112 |
+
self.retriever = SentenceTransformerRetriever()
|
113 |
+
self.documents = []
|
114 |
+
self.device = torch.device("cpu")
|
115 |
+
self.llm = load_llama_model()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
116 |
|
117 |
+
def preprocess_query(self, query: str) -> str:
|
118 |
+
"""Clean and prepare the query"""
|
119 |
+
query = query.lower().strip()
|
120 |
+
query = re.sub(r'\s+', ' ', query)
|
121 |
+
return query
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
122 |
|
123 |
+
def postprocess_response(self, response: str) -> str:
|
124 |
+
"""Clean up the generated response"""
|
125 |
+
response = response.strip()
|
126 |
+
response = re.sub(r'\s+', ' ', response)
|
127 |
+
response = re.sub(r'\d{4}-\d{2}-\d{2}\s\d{2}:\d{2}:\d{2}(?:\+\d{2}:?\d{2})?', '', response)
|
128 |
+
return response
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
129 |
|
|
|
130 |
def query_model(self, prompt: str) -> str:
|
131 |
+
"""Query the local Llama model"""
|
132 |
try:
|
133 |
if self.llm is None:
|
134 |
raise RuntimeError("Model not initialized")
|
135 |
+
|
|
|
136 |
response = self.llm(
|
137 |
prompt,
|
138 |
max_tokens=512,
|
|
|
141 |
echo=False,
|
142 |
stop=["Question:", "\n\n"]
|
143 |
)
|
144 |
+
|
|
|
145 |
if response and 'choices' in response and len(response['choices']) > 0:
|
146 |
text = response['choices'][0].get('text', '').strip()
|
147 |
return text
|
148 |
else:
|
149 |
raise ValueError("No valid response generated")
|
150 |
+
|
151 |
except Exception as e:
|
152 |
logging.error(f"Error in query_model: {str(e)}")
|
153 |
raise
|
154 |
+
|
|
|
|
|
|
|
|
|
|
|
155 |
def process_query(self, query: str, placeholder) -> str:
|
156 |
try:
|
157 |
# Preprocess query
|
158 |
query = self.preprocess_query(query)
|
159 |
+
|
160 |
# Show retrieval status
|
161 |
status = placeholder.empty()
|
162 |
status.write("π Finding relevant information...")
|
163 |
+
|
164 |
# Get embeddings and search
|
165 |
query_embedding = self.retriever.encode([query])
|
166 |
similarities = F.cosine_similarity(query_embedding, self.retriever.doc_embeddings)
|
167 |
scores, indices = torch.topk(similarities, k=min(self.k, len(self.documents)))
|
168 |
+
|
169 |
relevant_docs = [self.documents[idx] for idx in indices.tolist()]
|
170 |
+
|
171 |
# Update status
|
172 |
status.write("π Generating response...")
|
173 |
+
|
174 |
# Prepare context and prompt
|
175 |
+
context = "\n".join(relevant_docs[:3])
|
176 |
prompt = f"""Context information is below:
|
177 |
{context}
|
178 |
+
|
179 |
Given the context above, please answer the following question:
|
180 |
{query}
|
181 |
|
|
|
185 |
- Only include sports-related information
|
186 |
- No dates or timestamps in the response
|
187 |
- Use clear, natural language
|
188 |
+
|
189 |
Answer:"""
|
190 |
+
|
191 |
# Generate response
|
192 |
response_placeholder = placeholder.empty()
|
193 |
+
|
194 |
try:
|
195 |
response_text = self.query_model(prompt)
|
196 |
if response_text:
|
|
|
201 |
message = "No relevant answer found. Please try rephrasing your question."
|
202 |
response_placeholder.warning(message)
|
203 |
return message
|
204 |
+
|
205 |
except Exception as e:
|
206 |
logging.error(f"Generation error: {str(e)}")
|
207 |
message = "Had some trouble generating the response. Please try again."
|
208 |
response_placeholder.warning(message)
|
209 |
return message
|
210 |
+
|
211 |
except Exception as e:
|
212 |
logging.error(f"Process error: {str(e)}")
|
213 |
message = "Something went wrong. Please try again with a different question."
|
214 |
placeholder.warning(message)
|
215 |
return message
|
216 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
217 |
@st.cache_resource(show_spinner=False)
|
218 |
def initialize_rag_pipeline():
|
219 |
"""Initialize the RAG pipeline once"""
|
220 |
try:
|
221 |
+
# Create necessary directories
|
222 |
+
os.makedirs("ESPN_data", exist_ok=True)
|
223 |
+
|
224 |
+
# Load embeddings from Drive
|
|
|
225 |
drive_file_id = "1MuV63AE9o6zR9aBvdSDQOUextp71r2NN"
|
226 |
with st.spinner("Loading embeddings from Google Drive..."):
|
227 |
cache_data = load_from_drive(drive_file_id)
|
|
|
229 |
st.error("Failed to load embeddings from Google Drive")
|
230 |
st.stop()
|
231 |
|
232 |
+
# Initialize pipeline
|
233 |
data_folder = "ESPN_data"
|
234 |
rag = RAGPipeline(data_folder)
|
235 |
|
|
|
247 |
def main():
|
248 |
try:
|
249 |
# Environment check
|
250 |
+
if not check_environment():
|
251 |
return
|
252 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
253 |
# Improved CSS styling
|
254 |
st.markdown("""
|
255 |
<style>
|
|
|
320 |
</style>
|
321 |
""", unsafe_allow_html=True)
|
322 |
|
323 |
+
# Header section
|
324 |
st.markdown("<h1 class='main-title'>π The Sport Chatbot</h1>", unsafe_allow_html=True)
|
325 |
st.markdown("<h3 class='sub-title'>Using ESPN API</h3>", unsafe_allow_html=True)
|
326 |
st.markdown("""
|
|
|
333 |
</p>
|
334 |
""", unsafe_allow_html=True)
|
335 |
|
|
|
|
|
|
|
336 |
# Initialize the pipeline
|
337 |
+
if 'rag' not in st.session_state:
|
338 |
+
with st.spinner("Loading resources..."):
|
339 |
+
st.session_state.rag = initialize_rag_pipeline()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
340 |
|
341 |
+
# Create columns for layout
|
342 |
col1, col2, col3 = st.columns([1, 6, 1])
|
343 |
|
344 |
with col2:
|
345 |
+
# Query input
|
346 |
query = st.text_input("What would you like to know about sports?")
|
347 |
|
|
|
348 |
if st.button("Get Answer"):
|
349 |
if query:
|
350 |
response_placeholder = st.empty()
|
351 |
try:
|
|
|
352 |
response = st.session_state.rag.process_query(query, response_placeholder)
|
353 |
logging.info(f"Generated response: {response}")
|
354 |
except Exception as e:
|
|
|
357 |
else:
|
358 |
st.warning("Please enter a question!")
|
359 |
|
360 |
+
# Footer
|
361 |
st.markdown("<br><br>", unsafe_allow_html=True)
|
362 |
st.markdown("---")
|
363 |
st.markdown("""
|
364 |
<p style='text-align: center; color: #666666; padding: 1rem 0;'>
|
365 |
+
Powered by ESPN Data & Mistral AI π
|
|
|
366 |
</p>
|
367 |
""", unsafe_allow_html=True)
|
368 |
|
|
|
371 |
st.error("An unexpected error occurred. Please check the logs and try again.")
|
372 |
|
373 |
if __name__ == "__main__":
|
374 |
+
main()
|
|
|
|
|
|
|
|