Spaces:
Sleeping
Sleeping
oceansweep
commited on
Commit
•
5fab6ba
1
Parent(s):
a6ecdfa
Update App_Function_Libraries/RAG/RAG_Libary_2.py
Browse files
App_Function_Libraries/RAG/RAG_Libary_2.py
CHANGED
@@ -1,332 +1,153 @@
|
|
1 |
-
# RAG_Library_2.py
|
2 |
-
# Description: This script contains the main RAG pipeline function and related functions for the RAG pipeline.
|
3 |
-
#
|
4 |
-
# Import necessary modules and functions
|
5 |
-
import configparser
|
6 |
-
import logging
|
7 |
-
import os
|
8 |
-
from typing import Dict, Any, List, Optional
|
9 |
-
# Local Imports
|
10 |
-
from App_Function_Libraries.RAG.ChromaDB_Library import process_and_store_content, vector_search, chroma_client
|
11 |
-
from App_Function_Libraries.Article_Extractor_Lib import scrape_article
|
12 |
-
from App_Function_Libraries.DB.DB_Manager import add_media_to_database, search_db, get_unprocessed_media, \
|
13 |
-
fetch_keywords_for_media
|
14 |
-
from App_Function_Libraries.Utils.Utils import load_comprehensive_config
|
15 |
-
#
|
16 |
-
# 3rd-Party Imports
|
17 |
-
import openai
|
18 |
-
#
|
19 |
-
########################################################################################################################
|
20 |
-
#
|
21 |
-
# Functions:
|
22 |
-
|
23 |
-
# Initialize OpenAI client (adjust this based on your API key management)
|
24 |
-
openai.api_key = "your-openai-api-key"
|
25 |
-
|
26 |
-
# Get the directory of the current script
|
27 |
-
current_dir = os.path.dirname(os.path.abspath(__file__))
|
28 |
-
# Construct the path to the config file
|
29 |
-
config_path = os.path.join(current_dir, 'Config_Files', 'config.txt')
|
30 |
-
# Read the config file
|
31 |
-
config = configparser.ConfigParser()
|
32 |
-
# Read the configuration file
|
33 |
-
config.read('config.txt')
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
|
74 |
-
|
75 |
-
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
|
102 |
-
|
103 |
-
|
104 |
-
|
105 |
-
|
106 |
-
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
126 |
-
|
127 |
-
|
128 |
-
|
129 |
-
|
130 |
-
|
131 |
-
|
132 |
-
|
133 |
-
|
134 |
-
|
135 |
-
|
136 |
-
|
137 |
-
|
138 |
-
|
139 |
-
|
140 |
-
|
141 |
-
|
142 |
-
|
143 |
-
|
144 |
-
|
145 |
-
|
146 |
-
|
147 |
-
|
148 |
-
|
149 |
-
|
150 |
-
|
151 |
-
|
152 |
-
|
153 |
-
|
154 |
-
"context": ""
|
155 |
-
}
|
156 |
-
|
157 |
-
|
158 |
-
def generate_answer(api_choice: str, context: str, query: str) -> str:
|
159 |
-
logging.debug("Entering generate_answer function")
|
160 |
-
config = load_comprehensive_config()
|
161 |
-
logging.debug(f"Config sections: {config.sections()}")
|
162 |
-
prompt = f"Context: {context}\n\nQuestion: {query}"
|
163 |
-
if api_choice == "OpenAI":
|
164 |
-
from App_Function_Libraries.Summarization_General_Lib import summarize_with_openai
|
165 |
-
return summarize_with_openai(config['API']['openai_api_key'], prompt, "")
|
166 |
-
elif api_choice == "Anthropic":
|
167 |
-
from App_Function_Libraries.Summarization_General_Lib import summarize_with_anthropic
|
168 |
-
return summarize_with_anthropic(config['API']['anthropic_api_key'], prompt, "")
|
169 |
-
elif api_choice == "Cohere":
|
170 |
-
from App_Function_Libraries.Summarization_General_Lib import summarize_with_cohere
|
171 |
-
return summarize_with_cohere(config['API']['cohere_api_key'], prompt, "")
|
172 |
-
elif api_choice == "Groq":
|
173 |
-
from App_Function_Libraries.Summarization_General_Lib import summarize_with_groq
|
174 |
-
return summarize_with_groq(config['API']['groq_api_key'], prompt, "")
|
175 |
-
elif api_choice == "OpenRouter":
|
176 |
-
from App_Function_Libraries.Summarization_General_Lib import summarize_with_openrouter
|
177 |
-
return summarize_with_openrouter(config['API']['openrouter_api_key'], prompt, "")
|
178 |
-
elif api_choice == "HuggingFace":
|
179 |
-
from App_Function_Libraries.Summarization_General_Lib import summarize_with_huggingface
|
180 |
-
return summarize_with_huggingface(config['API']['huggingface_api_key'], prompt, "")
|
181 |
-
elif api_choice == "DeepSeek":
|
182 |
-
from App_Function_Libraries.Summarization_General_Lib import summarize_with_deepseek
|
183 |
-
return summarize_with_deepseek(config['API']['deepseek_api_key'], prompt, "")
|
184 |
-
elif api_choice == "Mistral":
|
185 |
-
from App_Function_Libraries.Summarization_General_Lib import summarize_with_mistral
|
186 |
-
return summarize_with_mistral(config['API']['mistral_api_key'], prompt, "")
|
187 |
-
elif api_choice == "Local-LLM":
|
188 |
-
from App_Function_Libraries.Local_Summarization_Lib import summarize_with_local_llm
|
189 |
-
return summarize_with_local_llm(config['API']['local_llm_path'], prompt, "")
|
190 |
-
elif api_choice == "Llama.cpp":
|
191 |
-
from App_Function_Libraries.Local_Summarization_Lib import summarize_with_llama
|
192 |
-
return summarize_with_llama(config['API']['llama_api_key'], prompt, "")
|
193 |
-
elif api_choice == "Kobold":
|
194 |
-
from App_Function_Libraries.Local_Summarization_Lib import summarize_with_kobold
|
195 |
-
return summarize_with_kobold(config['API']['kobold_api_key'], prompt, "")
|
196 |
-
elif api_choice == "Ooba":
|
197 |
-
from App_Function_Libraries.Local_Summarization_Lib import summarize_with_oobabooga
|
198 |
-
return summarize_with_oobabooga(config['API']['ooba_api_key'], prompt, "")
|
199 |
-
elif api_choice == "TabbyAPI":
|
200 |
-
from App_Function_Libraries.Local_Summarization_Lib import summarize_with_tabbyapi
|
201 |
-
return summarize_with_tabbyapi(config['API']['tabby_api_key'], prompt, "")
|
202 |
-
elif api_choice == "vLLM":
|
203 |
-
from App_Function_Libraries.Local_Summarization_Lib import summarize_with_vllm
|
204 |
-
return summarize_with_vllm(config['API']['vllm_api_key'], prompt, "")
|
205 |
-
elif api_choice == "ollama":
|
206 |
-
from App_Function_Libraries.Local_Summarization_Lib import summarize_with_ollama
|
207 |
-
return summarize_with_ollama(config['API']['ollama_api_key'], prompt, "")
|
208 |
-
else:
|
209 |
-
raise ValueError(f"Unsupported API choice: {api_choice}")
|
210 |
-
|
211 |
-
# Function to preprocess and store all existing content in the database
|
212 |
-
def preprocess_all_content():
|
213 |
-
unprocessed_media = get_unprocessed_media()
|
214 |
-
for row in unprocessed_media:
|
215 |
-
media_id = row[0]
|
216 |
-
content = row[1]
|
217 |
-
media_type = row[2]
|
218 |
-
collection_name = f"{media_type}_{media_id}"
|
219 |
-
process_and_store_content(content, collection_name, media_id, "")
|
220 |
-
|
221 |
-
|
222 |
-
def perform_vector_search(query: str, relevant_media_ids: List[str] = None) -> List[Dict[str, Any]]:
|
223 |
-
all_collections = chroma_client.list_collections()
|
224 |
-
vector_results = []
|
225 |
-
for collection in all_collections:
|
226 |
-
collection_results = vector_search(collection.name, query, k=5)
|
227 |
-
filtered_results = [
|
228 |
-
result for result in collection_results
|
229 |
-
if relevant_media_ids is None or result['metadata'].get('media_id') in relevant_media_ids
|
230 |
-
]
|
231 |
-
vector_results.extend(filtered_results)
|
232 |
-
return vector_results
|
233 |
-
|
234 |
-
|
235 |
-
def perform_full_text_search(query: str, relevant_media_ids: List[str] = None) -> List[Dict[str, Any]]:
|
236 |
-
fts_results = search_db(query, ["content"], "", page=1, results_per_page=5)
|
237 |
-
filtered_fts_results = [
|
238 |
-
{
|
239 |
-
"content": result['content'],
|
240 |
-
"metadata": {"media_id": result['id']}
|
241 |
-
}
|
242 |
-
for result in fts_results
|
243 |
-
if relevant_media_ids is None or result['id'] in relevant_media_ids
|
244 |
-
]
|
245 |
-
return filtered_fts_results
|
246 |
-
|
247 |
-
|
248 |
-
def fetch_relevant_media_ids(keywords: List[str]) -> List[int]:
|
249 |
-
relevant_ids = set()
|
250 |
-
try:
|
251 |
-
for keyword in keywords:
|
252 |
-
media_ids = fetch_keywords_for_media(keyword)
|
253 |
-
relevant_ids.update(media_ids)
|
254 |
-
except Exception as e:
|
255 |
-
logging.error(f"Error fetching relevant media IDs: {str(e)}")
|
256 |
-
return list(relevant_ids)
|
257 |
-
|
258 |
-
|
259 |
-
def filter_results_by_keywords(results: List[Dict[str, Any]], keywords: List[str]) -> List[Dict[str, Any]]:
|
260 |
-
if not keywords:
|
261 |
-
return results
|
262 |
-
|
263 |
-
filtered_results = []
|
264 |
-
for result in results:
|
265 |
-
try:
|
266 |
-
metadata = result.get('metadata', {})
|
267 |
-
if metadata is None:
|
268 |
-
logging.warning(f"No metadata found for result: {result}")
|
269 |
-
continue
|
270 |
-
if not isinstance(metadata, dict):
|
271 |
-
logging.warning(f"Unexpected metadata type: {type(metadata)}. Expected dict.")
|
272 |
-
continue
|
273 |
-
|
274 |
-
media_id = metadata.get('media_id')
|
275 |
-
if media_id is None:
|
276 |
-
logging.warning(f"No media_id found in metadata: {metadata}")
|
277 |
-
continue
|
278 |
-
|
279 |
-
media_keywords = fetch_keywords_for_media(media_id)
|
280 |
-
if any(keyword.lower() in [mk.lower() for mk in media_keywords] for keyword in keywords):
|
281 |
-
filtered_results.append(result)
|
282 |
-
except Exception as e:
|
283 |
-
logging.error(f"Error processing result: {result}. Error: {str(e)}")
|
284 |
-
|
285 |
-
return filtered_results
|
286 |
-
|
287 |
-
# FIXME: to be implememted
|
288 |
-
def extract_media_id_from_result(result: str) -> Optional[int]:
|
289 |
-
# Implement this function based on how you store the media_id in your results
|
290 |
-
# For example, if it's stored at the beginning of each result:
|
291 |
-
try:
|
292 |
-
return int(result.split('_')[0])
|
293 |
-
except (IndexError, ValueError):
|
294 |
-
logging.error(f"Failed to extract media_id from result: {result}")
|
295 |
-
return None
|
296 |
-
|
297 |
-
|
298 |
-
|
299 |
-
|
300 |
-
# Example usage:
|
301 |
-
# 1. Initialize the system:
|
302 |
-
# create_tables(db) # Ensure FTS tables are set up
|
303 |
-
#
|
304 |
-
# 2. Create ChromaDB
|
305 |
-
# chroma_client = ChromaDBClient()
|
306 |
-
#
|
307 |
-
# 3. Create Embeddings
|
308 |
-
# Store embeddings in ChromaDB
|
309 |
-
# preprocess_all_content() or create_embeddings()
|
310 |
-
#
|
311 |
-
# 4. Perform RAG search across all content:
|
312 |
-
# result = rag_search("What are the key points about climate change?")
|
313 |
-
# print(result['answer'])
|
314 |
-
#
|
315 |
-
# (Extra)5. Perform RAG on a specific URL:
|
316 |
-
# result = rag_pipeline("https://example.com/article", "What is the main topic of this article?")
|
317 |
-
# print(result['answer'])
|
318 |
-
#
|
319 |
-
########################################################################################################################
|
320 |
-
|
321 |
-
|
322 |
-
############################################################################################################
|
323 |
-
#
|
324 |
-
# ElasticSearch Retriever
|
325 |
-
|
326 |
-
# https://github.com/langchain-ai/langchain/tree/44e3e2391c48bfd0a8e6a20adde0b6567f4f43c3/templates/rag-elasticsearch
|
327 |
-
#
|
328 |
-
# https://github.com/langchain-ai/langchain/tree/44e3e2391c48bfd0a8e6a20adde0b6567f4f43c3/templates/rag-self-query
|
329 |
-
|
330 |
-
#
|
331 |
-
# End of RAG_Library_2.py
|
332 |
-
############################################################################################################
|
|
|
1 |
+
# RAG_Library_2.py
|
2 |
+
# Description: This script contains the main RAG pipeline function and related functions for the RAG pipeline.
|
3 |
+
#
|
4 |
+
# Import necessary modules and functions
|
5 |
+
import configparser
|
6 |
+
import logging
|
7 |
+
import os
|
8 |
+
from typing import Dict, Any, List, Optional
|
9 |
+
# Local Imports
|
10 |
+
#from App_Function_Libraries.RAG.ChromaDB_Library import process_and_store_content, vector_search, chroma_client
|
11 |
+
from App_Function_Libraries.Article_Extractor_Lib import scrape_article
|
12 |
+
from App_Function_Libraries.DB.DB_Manager import add_media_to_database, search_db, get_unprocessed_media, \
|
13 |
+
fetch_keywords_for_media
|
14 |
+
from App_Function_Libraries.Utils.Utils import load_comprehensive_config
|
15 |
+
#
|
16 |
+
# 3rd-Party Imports
|
17 |
+
import openai
|
18 |
+
#
|
19 |
+
########################################################################################################################
|
20 |
+
#
|
21 |
+
# Functions:
|
22 |
+
|
23 |
+
# Initialize OpenAI client (adjust this based on your API key management)
|
24 |
+
openai.api_key = "your-openai-api-key"
|
25 |
+
|
26 |
+
# Get the directory of the current script
|
27 |
+
current_dir = os.path.dirname(os.path.abspath(__file__))
|
28 |
+
# Construct the path to the config file
|
29 |
+
config_path = os.path.join(current_dir, 'Config_Files', 'config.txt')
|
30 |
+
# Read the config file
|
31 |
+
config = configparser.ConfigParser()
|
32 |
+
# Read the configuration file
|
33 |
+
config.read('config.txt')
|
34 |
+
|
35 |
+
|
36 |
+
|
37 |
+
|
38 |
+
|
39 |
+
|
40 |
+
|
41 |
+
def generate_answer(api_choice: str, context: str, query: str) -> str:
|
42 |
+
logging.debug("Entering generate_answer function")
|
43 |
+
config = load_comprehensive_config()
|
44 |
+
logging.debug(f"Config sections: {config.sections()}")
|
45 |
+
prompt = f"Context: {context}\n\nQuestion: {query}"
|
46 |
+
if api_choice == "OpenAI":
|
47 |
+
from App_Function_Libraries.Summarization_General_Lib import summarize_with_openai
|
48 |
+
return summarize_with_openai(config['API']['openai_api_key'], prompt, "")
|
49 |
+
elif api_choice == "Anthropic":
|
50 |
+
from App_Function_Libraries.Summarization_General_Lib import summarize_with_anthropic
|
51 |
+
return summarize_with_anthropic(config['API']['anthropic_api_key'], prompt, "")
|
52 |
+
elif api_choice == "Cohere":
|
53 |
+
from App_Function_Libraries.Summarization_General_Lib import summarize_with_cohere
|
54 |
+
return summarize_with_cohere(config['API']['cohere_api_key'], prompt, "")
|
55 |
+
elif api_choice == "Groq":
|
56 |
+
from App_Function_Libraries.Summarization_General_Lib import summarize_with_groq
|
57 |
+
return summarize_with_groq(config['API']['groq_api_key'], prompt, "")
|
58 |
+
elif api_choice == "OpenRouter":
|
59 |
+
from App_Function_Libraries.Summarization_General_Lib import summarize_with_openrouter
|
60 |
+
return summarize_with_openrouter(config['API']['openrouter_api_key'], prompt, "")
|
61 |
+
elif api_choice == "HuggingFace":
|
62 |
+
from App_Function_Libraries.Summarization_General_Lib import summarize_with_huggingface
|
63 |
+
return summarize_with_huggingface(config['API']['huggingface_api_key'], prompt, "")
|
64 |
+
elif api_choice == "DeepSeek":
|
65 |
+
from App_Function_Libraries.Summarization_General_Lib import summarize_with_deepseek
|
66 |
+
return summarize_with_deepseek(config['API']['deepseek_api_key'], prompt, "")
|
67 |
+
elif api_choice == "Mistral":
|
68 |
+
from App_Function_Libraries.Summarization_General_Lib import summarize_with_mistral
|
69 |
+
return summarize_with_mistral(config['API']['mistral_api_key'], prompt, "")
|
70 |
+
elif api_choice == "Local-LLM":
|
71 |
+
from App_Function_Libraries.Local_Summarization_Lib import summarize_with_local_llm
|
72 |
+
return summarize_with_local_llm(config['API']['local_llm_path'], prompt, "")
|
73 |
+
elif api_choice == "Llama.cpp":
|
74 |
+
from App_Function_Libraries.Local_Summarization_Lib import summarize_with_llama
|
75 |
+
return summarize_with_llama(config['API']['llama_api_key'], prompt, "")
|
76 |
+
elif api_choice == "Kobold":
|
77 |
+
from App_Function_Libraries.Local_Summarization_Lib import summarize_with_kobold
|
78 |
+
return summarize_with_kobold(config['API']['kobold_api_key'], prompt, "")
|
79 |
+
elif api_choice == "Ooba":
|
80 |
+
from App_Function_Libraries.Local_Summarization_Lib import summarize_with_oobabooga
|
81 |
+
return summarize_with_oobabooga(config['API']['ooba_api_key'], prompt, "")
|
82 |
+
elif api_choice == "TabbyAPI":
|
83 |
+
from App_Function_Libraries.Local_Summarization_Lib import summarize_with_tabbyapi
|
84 |
+
return summarize_with_tabbyapi(config['API']['tabby_api_key'], prompt, "")
|
85 |
+
elif api_choice == "vLLM":
|
86 |
+
from App_Function_Libraries.Local_Summarization_Lib import summarize_with_vllm
|
87 |
+
return summarize_with_vllm(config['API']['vllm_api_key'], prompt, "")
|
88 |
+
elif api_choice == "ollama":
|
89 |
+
from App_Function_Libraries.Local_Summarization_Lib import summarize_with_ollama
|
90 |
+
return summarize_with_ollama(config['API']['ollama_api_key'], prompt, "")
|
91 |
+
else:
|
92 |
+
raise ValueError(f"Unsupported API choice: {api_choice}")
|
93 |
+
|
94 |
+
|
95 |
+
def perform_full_text_search(query: str, relevant_media_ids: List[str] = None) -> List[Dict[str, Any]]:
|
96 |
+
fts_results = search_db(query, ["content"], "", page=1, results_per_page=5)
|
97 |
+
filtered_fts_results = [
|
98 |
+
{
|
99 |
+
"content": result['content'],
|
100 |
+
"metadata": {"media_id": result['id']}
|
101 |
+
}
|
102 |
+
for result in fts_results
|
103 |
+
if relevant_media_ids is None or result['id'] in relevant_media_ids
|
104 |
+
]
|
105 |
+
return filtered_fts_results
|
106 |
+
|
107 |
+
|
108 |
+
def fetch_relevant_media_ids(keywords: List[str]) -> List[int]:
|
109 |
+
relevant_ids = set()
|
110 |
+
try:
|
111 |
+
for keyword in keywords:
|
112 |
+
media_ids = fetch_keywords_for_media(keyword)
|
113 |
+
relevant_ids.update(media_ids)
|
114 |
+
except Exception as e:
|
115 |
+
logging.error(f"Error fetching relevant media IDs: {str(e)}")
|
116 |
+
return list(relevant_ids)
|
117 |
+
|
118 |
+
|
119 |
+
|
120 |
+
|
121 |
+
# Example usage:
|
122 |
+
# 1. Initialize the system:
|
123 |
+
# create_tables(db) # Ensure FTS tables are set up
|
124 |
+
#
|
125 |
+
# 2. Create ChromaDB
|
126 |
+
# chroma_client = ChromaDBClient()
|
127 |
+
#
|
128 |
+
# 3. Create Embeddings
|
129 |
+
# Store embeddings in ChromaDB
|
130 |
+
# preprocess_all_content() or create_embeddings()
|
131 |
+
#
|
132 |
+
# 4. Perform RAG search across all content:
|
133 |
+
# result = rag_search("What are the key points about climate change?")
|
134 |
+
# print(result['answer'])
|
135 |
+
#
|
136 |
+
# (Extra)5. Perform RAG on a specific URL:
|
137 |
+
# result = rag_pipeline("https://example.com/article", "What is the main topic of this article?")
|
138 |
+
# print(result['answer'])
|
139 |
+
#
|
140 |
+
########################################################################################################################
|
141 |
+
|
142 |
+
|
143 |
+
############################################################################################################
|
144 |
+
#
|
145 |
+
# ElasticSearch Retriever
|
146 |
+
|
147 |
+
# https://github.com/langchain-ai/langchain/tree/44e3e2391c48bfd0a8e6a20adde0b6567f4f43c3/templates/rag-elasticsearch
|
148 |
+
#
|
149 |
+
# https://github.com/langchain-ai/langchain/tree/44e3e2391c48bfd0a8e6a20adde0b6567f4f43c3/templates/rag-self-query
|
150 |
+
|
151 |
+
#
|
152 |
+
# End of RAG_Library_2.py
|
153 |
+
############################################################################################################
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|