Spaces:
Sleeping
Sleeping
Update summarizer.py
Browse files- summarizer.py +204 -204
summarizer.py
CHANGED
@@ -1,205 +1,205 @@
|
|
1 |
-
import os
|
2 |
-
import re
|
3 |
-
from typing import Dict, Optional
|
4 |
-
import google.generativeai as genai
|
5 |
-
import logging
|
6 |
-
from dotenv import load_dotenv
|
7 |
-
from urllib.parse import urlparse
|
8 |
-
from cachetools import TTLCache
|
9 |
-
|
10 |
-
# Load environment variables
|
11 |
-
load_dotenv()
|
12 |
-
|
13 |
-
# Set up logging
|
14 |
-
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
|
15 |
-
|
16 |
-
# In-memory cache: 1000 items, 1-hour TTL
|
17 |
-
cache = TTLCache(maxsize=1000, ttl=3600)
|
18 |
-
|
19 |
-
async def summarize_text(text: str, url: str = "") -> Dict[str, str]:
|
20 |
-
"""Summarize text into a title and description using Gemini-1.5 Flash."""
|
21 |
-
try:
|
22 |
-
# Validate inputs
|
23 |
-
text = text.strip() if text else ""
|
24 |
-
if not url:
|
25 |
-
url = "https://example.com"
|
26 |
-
try:
|
27 |
-
parsed_url = urlparse(url)
|
28 |
-
domain = parsed_url.netloc or "example.com"
|
29 |
-
except Exception:
|
30 |
-
logging.warning(f"Invalid URL: {url}. Using default domain.")
|
31 |
-
domain = "example.com"
|
32 |
-
|
33 |
-
# Check cache
|
34 |
-
cache_key = f"summarize_{hash(text + url)}"
|
35 |
-
if cache_key in cache:
|
36 |
-
logging.info(f"Cache hit for {cache_key}")
|
37 |
-
return cache[cache_key]
|
38 |
-
|
39 |
-
# Get Gemini API key
|
40 |
-
api_key = os.getenv("GEMINI_API_KEY")
|
41 |
-
if not api_key:
|
42 |
-
logging.error("Gemini API key not found. Please set GEMINI_API_KEY in .env file.")
|
43 |
-
raise ValueError("Gemini API key is required for summarization.")
|
44 |
-
|
45 |
-
# Configure Gemini client
|
46 |
-
genai.configure(api_key=api_key)
|
47 |
-
model = genai.GenerativeModel('gemini-1.5-flash')
|
48 |
-
|
49 |
-
# Handle short or empty text
|
50 |
-
if len(text) < 20:
|
51 |
-
logging.warning(f"Text too short ({len(text)} chars): '{text}'. Using URL context.")
|
52 |
-
text = f"Content from {url} about news, products, or services."
|
53 |
-
|
54 |
-
# Split text into chunks to avoid quota issues (e.g., 1000 chars per chunk)
|
55 |
-
chunk_size = 1000
|
56 |
-
text_chunks = [text[i:i + chunk_size] for i in range(0, len(text), chunk_size)]
|
57 |
-
summaries = []
|
58 |
-
|
59 |
-
for chunk in text_chunks[:2]: # Limit to first 2000 chars for efficiency
|
60 |
-
prompt = (
|
61 |
-
f"Summarize the following text into a title (30-50 characters) and a description (80-100 characters) "
|
62 |
-
f"for RCS messaging. Ensure titles are catchy and descriptions are engaging, relevant to the content, "
|
63 |
-
f"and suitable for a news, product, or service context inferred from the URL ({url}). "
|
64 |
-
f"Output as JSON:\n{{\"title\": \"[title]\", \"description\": \"[description]\"}}\n\nText: {chunk}"
|
65 |
-
)
|
66 |
-
|
67 |
-
response = await model.generate_content_async(prompt)
|
68 |
-
raw_content = response.text.strip()
|
69 |
-
logging.info(f"Raw Gemini response: {raw_content}")
|
70 |
-
|
71 |
-
# Parse response with regex
|
72 |
-
try:
|
73 |
-
match = re.search(r'\{[\s\S]*"title":\s*"([^"]+)"[\s\S]*"description":\s*"([^"]+)"[\s\S]*\}', raw_content)
|
74 |
-
if match:
|
75 |
-
title = match.group(1)
|
76 |
-
description = match.group(2)
|
77 |
-
summaries.append({"title": title, "description": description})
|
78 |
-
else:
|
79 |
-
raise ValueError("Invalid JSON format in Gemini response")
|
80 |
-
except Exception as e:
|
81 |
-
logging.warning(f"Failed to parse Gemini response: {e}. Skipping chunk.")
|
82 |
-
continue
|
83 |
-
|
84 |
-
# Combine summaries (prioritize first valid summary)
|
85 |
-
if summaries:
|
86 |
-
result = summaries[0]
|
87 |
-
else:
|
88 |
-
logging.warning("No valid summaries generated. Using fallback.")
|
89 |
-
result = {
|
90 |
-
"title": "News Summary",
|
91 |
-
"description": f"Discover news and insights from {domain}."[:100]
|
92 |
-
}
|
93 |
-
|
94 |
-
# Ensure non-empty outputs
|
95 |
-
if not result["title"].strip():
|
96 |
-
result["title"] = "News Summary"
|
97 |
-
if not result["description"].strip():
|
98 |
-
result["description"] = f"Discover news and insights from {domain}."
|
99 |
-
|
100 |
-
cache[cache_key] = result
|
101 |
-
logging.info(f"Summary - Title: {result['title']}, Description: {result['description']}")
|
102 |
-
return result
|
103 |
-
|
104 |
-
except Exception as e:
|
105 |
-
logging.error(f"Error summarizing text: {e}")
|
106 |
-
domain = urlparse(url).netloc or "example.com"
|
107 |
-
result = {
|
108 |
-
"title": "News Summary",
|
109 |
-
"description": f"Discover news and insights from {domain}."[:100]
|
110 |
-
}
|
111 |
-
cache[cache_key] = result
|
112 |
-
return result
|
113 |
-
|
114 |
-
async def quick_summarize(text: str, url: str = "") -> Dict[str, str]:
|
115 |
-
"""Quickly summarize text with a lightweight prompt using Gemini-1.5 Flash."""
|
116 |
-
try:
|
117 |
-
# Validate inputs
|
118 |
-
text = text.strip() if text else ""
|
119 |
-
if not url:
|
120 |
-
url = "https://example.com"
|
121 |
-
try:
|
122 |
-
parsed_url = urlparse(url)
|
123 |
-
domain = parsed_url.netloc or "example.com"
|
124 |
-
except Exception:
|
125 |
-
logging.warning(f"Invalid URL: {url}. Using default domain.")
|
126 |
-
domain = "example.com"
|
127 |
-
|
128 |
-
# Check cache
|
129 |
-
cache_key = f"quick_summarize_{hash(text + url)}"
|
130 |
-
if cache_key in cache:
|
131 |
-
logging.info(f"Cache hit for {cache_key}")
|
132 |
-
return cache[cache_key]
|
133 |
-
|
134 |
-
# Get Gemini API key
|
135 |
-
api_key = os.getenv("GEMINI_API_KEY")
|
136 |
-
if not api_key:
|
137 |
-
logging.error("Gemini API key not found. Please set GEMINI_API_KEY in .env file.")
|
138 |
-
raise ValueError("Gemini API key is required for summarization.")
|
139 |
-
|
140 |
-
# Configure Gemini client
|
141 |
-
genai.configure(api_key=api_key)
|
142 |
-
model = genai.GenerativeModel('gemini-1.5-
|
143 |
-
|
144 |
-
# Handle short or empty text
|
145 |
-
if len(text) < 20:
|
146 |
-
logging.warning(f"Text too short ({len(text)} chars): '{text}'. Using URL context.")
|
147 |
-
text = f"Content from {url} about news, products, or services."
|
148 |
-
|
149 |
-
# Lightweight prompt with chunking
|
150 |
-
chunk_size = 1000
|
151 |
-
text_chunks = [text[i:i + chunk_size] for i in range(0, len(text), chunk_size)]
|
152 |
-
summaries = []
|
153 |
-
|
154 |
-
for chunk in text_chunks[:1]: # Limit to first 1000 chars for quick summary
|
155 |
-
prompt = (
|
156 |
-
f"Create a title (30-50 chars) and description (80-100 chars) for RCS messaging from this text. "
|
157 |
-
f"Keep it engaging and relevant to {url}. Output as JSON:\n{{\"title\": \"[title]\", \"description\": \"[description]\"}}\n\nText: {chunk}"
|
158 |
-
)
|
159 |
-
|
160 |
-
response = await model.generate_content_async(prompt)
|
161 |
-
raw_content = response.text.strip()
|
162 |
-
logging.info(f"Raw Gemini response (quick): {raw_content}")
|
163 |
-
|
164 |
-
# Parse response with regex
|
165 |
-
try:
|
166 |
-
match = re.search(r'\{[\s\S]*"title":\s*"([^"]+)"[\s\S]*"description":\s*"([^"]+)"[\s\S]*\}', raw_content)
|
167 |
-
if match:
|
168 |
-
title = match.group(1)
|
169 |
-
description = match.group(2)
|
170 |
-
summaries.append({"title": title, "description": description})
|
171 |
-
else:
|
172 |
-
raise ValueError("Invalid JSON format in Gemini response")
|
173 |
-
except Exception as e:
|
174 |
-
logging.warning(f"Failed to parse Gemini response: {e}. Skipping chunk.")
|
175 |
-
continue
|
176 |
-
|
177 |
-
# Use first valid summary or fallback
|
178 |
-
if summaries:
|
179 |
-
result = summaries[0]
|
180 |
-
else:
|
181 |
-
logging.warning("No valid summaries generated. Using fallback.")
|
182 |
-
result = {
|
183 |
-
"title": "Quick Summary",
|
184 |
-
"description": f"Check out content from {domain}."
|
185 |
-
}
|
186 |
-
|
187 |
-
# Ensure non-empty outputs
|
188 |
-
if not result["title"].strip():
|
189 |
-
result["title"] = "Quick Summary"
|
190 |
-
if not result["description"].strip():
|
191 |
-
result["description"] = f"Check out content from {domain}."
|
192 |
-
|
193 |
-
cache[cache_key] = result
|
194 |
-
logging.info(f"Quick summary - Title: {result['title']}, Description: {result['description']}")
|
195 |
-
return result
|
196 |
-
|
197 |
-
except Exception as e:
|
198 |
-
logging.error(f"Error in quick summarize: {e}")
|
199 |
-
domain = urlparse(url).netloc or "example.com"
|
200 |
-
result = {
|
201 |
-
"title": "Quick Summary",
|
202 |
-
"description": f"Check out content from {domain}."
|
203 |
-
}
|
204 |
-
cache[cache_key] = result
|
205 |
return result
|
|
|
1 |
+
import os
|
2 |
+
import re
|
3 |
+
from typing import Dict, Optional
|
4 |
+
import google.generativeai as genai
|
5 |
+
import logging
|
6 |
+
from dotenv import load_dotenv
|
7 |
+
from urllib.parse import urlparse
|
8 |
+
from cachetools import TTLCache
|
9 |
+
|
10 |
+
# Load environment variables
|
11 |
+
load_dotenv()
|
12 |
+
|
13 |
+
# Set up logging
|
14 |
+
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
|
15 |
+
|
16 |
+
# In-memory cache: 1000 items, 1-hour TTL
|
17 |
+
cache = TTLCache(maxsize=1000, ttl=3600)
|
18 |
+
|
19 |
+
async def summarize_text(text: str, url: str = "") -> Dict[str, str]:
|
20 |
+
"""Summarize text into a title and description using Gemini-1.5 Flash."""
|
21 |
+
try:
|
22 |
+
# Validate inputs
|
23 |
+
text = text.strip() if text else ""
|
24 |
+
if not url:
|
25 |
+
url = "https://example.com"
|
26 |
+
try:
|
27 |
+
parsed_url = urlparse(url)
|
28 |
+
domain = parsed_url.netloc or "example.com"
|
29 |
+
except Exception:
|
30 |
+
logging.warning(f"Invalid URL: {url}. Using default domain.")
|
31 |
+
domain = "example.com"
|
32 |
+
|
33 |
+
# Check cache
|
34 |
+
cache_key = f"summarize_{hash(text + url)}"
|
35 |
+
if cache_key in cache:
|
36 |
+
logging.info(f"Cache hit for {cache_key}")
|
37 |
+
return cache[cache_key]
|
38 |
+
|
39 |
+
# Get Gemini API key
|
40 |
+
api_key = os.getenv("GEMINI_API_KEY")
|
41 |
+
if not api_key:
|
42 |
+
logging.error("Gemini API key not found. Please set GEMINI_API_KEY in .env file.")
|
43 |
+
raise ValueError("Gemini API key is required for summarization.")
|
44 |
+
|
45 |
+
# Configure Gemini client
|
46 |
+
genai.configure(api_key=api_key)
|
47 |
+
model = genai.GenerativeModel('gemini-1.5-flash')
|
48 |
+
|
49 |
+
# Handle short or empty text
|
50 |
+
if len(text) < 20:
|
51 |
+
logging.warning(f"Text too short ({len(text)} chars): '{text}'. Using URL context.")
|
52 |
+
text = f"Content from {url} about news, products, or services."
|
53 |
+
|
54 |
+
# Split text into chunks to avoid quota issues (e.g., 1000 chars per chunk)
|
55 |
+
chunk_size = 1000
|
56 |
+
text_chunks = [text[i:i + chunk_size] for i in range(0, len(text), chunk_size)]
|
57 |
+
summaries = []
|
58 |
+
|
59 |
+
for chunk in text_chunks[:2]: # Limit to first 2000 chars for efficiency
|
60 |
+
prompt = (
|
61 |
+
f"Summarize the following text into a title (30-50 characters) and a description (80-100 characters) "
|
62 |
+
f"for RCS messaging. Ensure titles are catchy and descriptions are engaging, relevant to the content, "
|
63 |
+
f"and suitable for a news, product, or service context inferred from the URL ({url}). "
|
64 |
+
f"Output as JSON:\n{{\"title\": \"[title]\", \"description\": \"[description]\"}}\n\nText: {chunk}"
|
65 |
+
)
|
66 |
+
|
67 |
+
response = await model.generate_content_async(prompt)
|
68 |
+
raw_content = response.text.strip()
|
69 |
+
logging.info(f"Raw Gemini response: {raw_content}")
|
70 |
+
|
71 |
+
# Parse response with regex
|
72 |
+
try:
|
73 |
+
match = re.search(r'\{[\s\S]*"title":\s*"([^"]+)"[\s\S]*"description":\s*"([^"]+)"[\s\S]*\}', raw_content)
|
74 |
+
if match:
|
75 |
+
title = match.group(1)
|
76 |
+
description = match.group(2)
|
77 |
+
summaries.append({"title": title, "description": description})
|
78 |
+
else:
|
79 |
+
raise ValueError("Invalid JSON format in Gemini response")
|
80 |
+
except Exception as e:
|
81 |
+
logging.warning(f"Failed to parse Gemini response: {e}. Skipping chunk.")
|
82 |
+
continue
|
83 |
+
|
84 |
+
# Combine summaries (prioritize first valid summary)
|
85 |
+
if summaries:
|
86 |
+
result = summaries[0]
|
87 |
+
else:
|
88 |
+
logging.warning("No valid summaries generated. Using fallback.")
|
89 |
+
result = {
|
90 |
+
"title": "News Summary",
|
91 |
+
"description": f"Discover news and insights from {domain}."[:100]
|
92 |
+
}
|
93 |
+
|
94 |
+
# Ensure non-empty outputs
|
95 |
+
if not result["title"].strip():
|
96 |
+
result["title"] = "News Summary"
|
97 |
+
if not result["description"].strip():
|
98 |
+
result["description"] = f"Discover news and insights from {domain}."
|
99 |
+
|
100 |
+
cache[cache_key] = result
|
101 |
+
logging.info(f"Summary - Title: {result['title']}, Description: {result['description']}")
|
102 |
+
return result
|
103 |
+
|
104 |
+
except Exception as e:
|
105 |
+
logging.error(f"Error summarizing text: {e}")
|
106 |
+
domain = urlparse(url).netloc or "example.com"
|
107 |
+
result = {
|
108 |
+
"title": "News Summary",
|
109 |
+
"description": f"Discover news and insights from {domain}."[:100]
|
110 |
+
}
|
111 |
+
cache[cache_key] = result
|
112 |
+
return result
|
113 |
+
|
114 |
+
async def quick_summarize(text: str, url: str = "") -> Dict[str, str]:
|
115 |
+
"""Quickly summarize text with a lightweight prompt using Gemini-1.5 Flash."""
|
116 |
+
try:
|
117 |
+
# Validate inputs
|
118 |
+
text = text.strip() if text else ""
|
119 |
+
if not url:
|
120 |
+
url = "https://example.com"
|
121 |
+
try:
|
122 |
+
parsed_url = urlparse(url)
|
123 |
+
domain = parsed_url.netloc or "example.com"
|
124 |
+
except Exception:
|
125 |
+
logging.warning(f"Invalid URL: {url}. Using default domain.")
|
126 |
+
domain = "example.com"
|
127 |
+
|
128 |
+
# Check cache
|
129 |
+
cache_key = f"quick_summarize_{hash(text + url)}"
|
130 |
+
if cache_key in cache:
|
131 |
+
logging.info(f"Cache hit for {cache_key}")
|
132 |
+
return cache[cache_key]
|
133 |
+
|
134 |
+
# Get Gemini API key
|
135 |
+
api_key = os.getenv("GEMINI_API_KEY")
|
136 |
+
if not api_key:
|
137 |
+
logging.error("Gemini API key not found. Please set GEMINI_API_KEY in .env file.")
|
138 |
+
raise ValueError("Gemini API key is required for summarization.")
|
139 |
+
|
140 |
+
# Configure Gemini client
|
141 |
+
genai.configure(api_key=api_key)
|
142 |
+
model = genai.GenerativeModel('gemini-1.5-pro')
|
143 |
+
|
144 |
+
# Handle short or empty text
|
145 |
+
if len(text) < 20:
|
146 |
+
logging.warning(f"Text too short ({len(text)} chars): '{text}'. Using URL context.")
|
147 |
+
text = f"Content from {url} about news, products, or services."
|
148 |
+
|
149 |
+
# Lightweight prompt with chunking
|
150 |
+
chunk_size = 1000
|
151 |
+
text_chunks = [text[i:i + chunk_size] for i in range(0, len(text), chunk_size)]
|
152 |
+
summaries = []
|
153 |
+
|
154 |
+
for chunk in text_chunks[:1]: # Limit to first 1000 chars for quick summary
|
155 |
+
prompt = (
|
156 |
+
f"Create a title (30-50 chars) and description (80-100 chars) for RCS messaging from this text. "
|
157 |
+
f"Keep it engaging and relevant to {url}. Output as JSON:\n{{\"title\": \"[title]\", \"description\": \"[description]\"}}\n\nText: {chunk}"
|
158 |
+
)
|
159 |
+
|
160 |
+
response = await model.generate_content_async(prompt)
|
161 |
+
raw_content = response.text.strip()
|
162 |
+
logging.info(f"Raw Gemini response (quick): {raw_content}")
|
163 |
+
|
164 |
+
# Parse response with regex
|
165 |
+
try:
|
166 |
+
match = re.search(r'\{[\s\S]*"title":\s*"([^"]+)"[\s\S]*"description":\s*"([^"]+)"[\s\S]*\}', raw_content)
|
167 |
+
if match:
|
168 |
+
title = match.group(1)
|
169 |
+
description = match.group(2)
|
170 |
+
summaries.append({"title": title, "description": description})
|
171 |
+
else:
|
172 |
+
raise ValueError("Invalid JSON format in Gemini response")
|
173 |
+
except Exception as e:
|
174 |
+
logging.warning(f"Failed to parse Gemini response: {e}. Skipping chunk.")
|
175 |
+
continue
|
176 |
+
|
177 |
+
# Use first valid summary or fallback
|
178 |
+
if summaries:
|
179 |
+
result = summaries[0]
|
180 |
+
else:
|
181 |
+
logging.warning("No valid summaries generated. Using fallback.")
|
182 |
+
result = {
|
183 |
+
"title": "Quick Summary",
|
184 |
+
"description": f"Check out content from {domain}."
|
185 |
+
}
|
186 |
+
|
187 |
+
# Ensure non-empty outputs
|
188 |
+
if not result["title"].strip():
|
189 |
+
result["title"] = "Quick Summary"
|
190 |
+
if not result["description"].strip():
|
191 |
+
result["description"] = f"Check out content from {domain}."
|
192 |
+
|
193 |
+
cache[cache_key] = result
|
194 |
+
logging.info(f"Quick summary - Title: {result['title']}, Description: {result['description']}")
|
195 |
+
return result
|
196 |
+
|
197 |
+
except Exception as e:
|
198 |
+
logging.error(f"Error in quick summarize: {e}")
|
199 |
+
domain = urlparse(url).netloc or "example.com"
|
200 |
+
result = {
|
201 |
+
"title": "Quick Summary",
|
202 |
+
"description": f"Check out content from {domain}."
|
203 |
+
}
|
204 |
+
cache[cache_key] = result
|
205 |
return result
|