Spaces:
Sleeping
Sleeping
Upload 5 files
Browse files- main.py +338 -0
- requirements.txt +0 -0
- rich_card_builder.py +71 -0
- scraper.py +73 -0
- summarizer.py +205 -0
main.py
ADDED
@@ -0,0 +1,338 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from fastapi import FastAPI, HTTPException, Request
|
2 |
+
from fastapi.responses import HTMLResponse
|
3 |
+
from fastapi.templating import Jinja2Templates
|
4 |
+
from fastapi.staticfiles import StaticFiles
|
5 |
+
from pydantic import HttpUrl
|
6 |
+
from scraper import scrape_page
|
7 |
+
from summarizer import quick_summarize
|
8 |
+
from rich_card_builder import build_rich_card
|
9 |
+
import asyncio
|
10 |
+
from urllib.parse import urlparse
|
11 |
+
import logging
|
12 |
+
import http.client
|
13 |
+
import json
|
14 |
+
from dotenv import load_dotenv
|
15 |
+
import os
|
16 |
+
import google.generativeai as genai
|
17 |
+
from typing import Optional, List, Dict
|
18 |
+
|
19 |
+
# Load environment variables
|
20 |
+
load_dotenv()
|
21 |
+
|
22 |
+
# Set up logging
|
23 |
+
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
|
24 |
+
|
25 |
+
app = FastAPI(title="Website Scraper API (Enhanced for RCS)")
|
26 |
+
|
27 |
+
# Mount static files
|
28 |
+
app.mount("/static", StaticFiles(directory="static"), name="static")
|
29 |
+
|
30 |
+
# Set up Jinja2 templates
|
31 |
+
templates = Jinja2Templates(directory="templates")
|
32 |
+
|
33 |
+
# Sample rich cards for testing
|
34 |
+
SAMPLE_RICH_CARDS = [
|
35 |
+
{
|
36 |
+
"title": "Summer Collection 2025",
|
37 |
+
"text": "Discover vibrant summer styles at Pantaloons.",
|
38 |
+
"media": "https://example.com/summer.jpg",
|
39 |
+
"url": "https://example.com/summer",
|
40 |
+
"buttons": [
|
41 |
+
{"type": "weburl", "title": "View Now", "payload": "https://example.com/summer"},
|
42 |
+
{"type": "postback", "title": "Learn More", "payload": "learn_more_1"}
|
43 |
+
],
|
44 |
+
"quickReplies": [
|
45 |
+
{"type": "postback", "title": "Show Similar", "payload": "similar_content_1"},
|
46 |
+
{"type": "call", "title": "Contact Support", "payload": "+12345678901"}
|
47 |
+
]
|
48 |
+
},
|
49 |
+
{
|
50 |
+
"title": "Yu Collection",
|
51 |
+
"text": "Explore trendy youth fashion with Yu Collection.",
|
52 |
+
"media": "https://example.com/yu.jpg",
|
53 |
+
"url": "https://example.com/yu_collection",
|
54 |
+
"buttons": [
|
55 |
+
{"type": "weburl", "title": "View Now", "payload": "https://example.com/yu_collection"},
|
56 |
+
{"type": "postback", "title": "Learn More", "payload": "learn_more_2"}
|
57 |
+
],
|
58 |
+
"quickReplies": [
|
59 |
+
{"type": "postback", "title": "Show Similar", "payload": "similar_content_2"},
|
60 |
+
{"type": "call", "title": "Contact Support", "payload": "+12345678901"}
|
61 |
+
]
|
62 |
+
}
|
63 |
+
]
|
64 |
+
|
65 |
+
async def generate_dynamic_buttons(title: str, description: str, url: str, next_interaction: str = None) -> List[Dict]:
|
66 |
+
"""Generate dynamic quick reply buttons for the next interaction using Gemini-1.5 Flash."""
|
67 |
+
try:
|
68 |
+
# Validate inputs
|
69 |
+
title = title.strip() if title and title.strip() else "News Summary"
|
70 |
+
description = description.strip() if description and description.strip() else "Explore news and insights."
|
71 |
+
url = url.strip() if url and url.strip() else "https://example.com"
|
72 |
+
|
73 |
+
logging.info(f"Generating buttons for: title={title}, description={description[:30]}..., url={url}")
|
74 |
+
|
75 |
+
# Get Gemini API key
|
76 |
+
api_key = os.getenv("GEMINI_API_KEY")
|
77 |
+
if not api_key:
|
78 |
+
logging.error("Gemini API key not found. Please set GEMINI_API_KEY in .env file.")
|
79 |
+
return [{"type": "postback", "title": "Explore More", "payload": f"goto_{next_interaction}", "execute": next_interaction}] if next_interaction else []
|
80 |
+
|
81 |
+
# Configure Gemini client
|
82 |
+
genai.configure(api_key=api_key)
|
83 |
+
model = genai.GenerativeModel('gemini-1.5-flash')
|
84 |
+
|
85 |
+
# Combine inputs (no chunking, as input is small)
|
86 |
+
input_text = f"Title: {title}\nDescription: {description}\nURL: {url}"
|
87 |
+
input_text = input_text[:500] # Truncate to 500 chars to stay within limits
|
88 |
+
|
89 |
+
# Optimized prompt
|
90 |
+
prompt = (
|
91 |
+
f"Based on the following content for the next interaction, suggest up to two concise (3-8 words) quick reply button titles that are action-oriented and invite the user to explore this content. The buttons should be engaging, relevant to the content, and avoid generic terms like 'Show Next'. Return the titles as a JSON array of strings.\n\n"
|
92 |
+
f"{input_text}\n\n"
|
93 |
+
f"Example output: [\"Discover Yu Collection\", \"Shop Youth Styles\"]\n"
|
94 |
+
f"Return only the JSON array, no markdown or extra text."
|
95 |
+
)
|
96 |
+
|
97 |
+
# Retry mechanism for API calls
|
98 |
+
max_retries = 3
|
99 |
+
for attempt in range(max_retries):
|
100 |
+
try:
|
101 |
+
response = await model.generate_content_async(prompt)
|
102 |
+
raw_content = response.text.strip()
|
103 |
+
logging.info(f"Gemini response: {raw_content}")
|
104 |
+
|
105 |
+
# Remove markdown code block markers if present
|
106 |
+
raw_content = raw_content.strip('```json').strip('```').strip()
|
107 |
+
|
108 |
+
# Parse response
|
109 |
+
button_titles = json.loads(raw_content)
|
110 |
+
if not isinstance(button_titles, list) or not all(isinstance(t, str) for t in button_titles):
|
111 |
+
logging.warning(f"Invalid Gemini response format: {raw_content}")
|
112 |
+
raise ValueError("Response is not a list of strings")
|
113 |
+
|
114 |
+
# Filter valid button titles
|
115 |
+
valid_buttons = [t.strip() for t in button_titles if t.strip() and 3 <= len(t.strip().split()) <= 8]
|
116 |
+
if not valid_buttons:
|
117 |
+
logging.warning("No valid button titles in response")
|
118 |
+
raise ValueError("No valid button titles")
|
119 |
+
|
120 |
+
# Create quick replies
|
121 |
+
quick_replies = [
|
122 |
+
{
|
123 |
+
"type": "postback",
|
124 |
+
"title": title[:20], # Ensure concise title
|
125 |
+
"payload": f"goto_{next_interaction}_{i}",
|
126 |
+
"execute": next_interaction
|
127 |
+
}
|
128 |
+
for i, title in enumerate(valid_buttons[:2])
|
129 |
+
]
|
130 |
+
logging.info(f"Generated quick replies: {quick_replies}")
|
131 |
+
return quick_replies
|
132 |
+
|
133 |
+
except Exception as e:
|
134 |
+
logging.warning(f"Attempt {attempt + 1} failed: {str(e)}")
|
135 |
+
if attempt < max_retries - 1:
|
136 |
+
await asyncio.sleep(1) # Wait before retrying
|
137 |
+
continue
|
138 |
+
|
139 |
+
# Fallback if all retries fail
|
140 |
+
logging.error("All retries failed for button generation")
|
141 |
+
return [{"type": "postback", "title": "Explore More", "payload": f"goto_{next_interaction}", "execute": next_interaction}] if next_interaction else []
|
142 |
+
|
143 |
+
except Exception as e:
|
144 |
+
logging.error(f"Error generating dynamic buttons: {str(e)}")
|
145 |
+
return [{"type": "postback", "title": "Explore More", "payload": f"goto_{next_interaction}", "execute": next_interaction}] if next_interaction else []
|
146 |
+
|
147 |
+
async def create_nativemsg_bot(rich_cards: List[Dict], url: str, bot_name: str, api_token: str) -> Dict:
|
148 |
+
"""Create a bot on NativeMSG with connected interactions based on rich cards."""
|
149 |
+
try:
|
150 |
+
# Validate API token
|
151 |
+
if not api_token:
|
152 |
+
logging.error("NativeMSG API token not provided and not found in .env file.")
|
153 |
+
raise ValueError("NativeMSG API token is required.")
|
154 |
+
|
155 |
+
# Use provided bot name or default to dynamic name
|
156 |
+
final_bot_name = bot_name or f"Bot for {urlparse(url).netloc}"
|
157 |
+
|
158 |
+
# Prepare bot payload
|
159 |
+
interactions = []
|
160 |
+
|
161 |
+
for idx, card in enumerate(rich_cards, 1):
|
162 |
+
# Determine the next interaction and next card (if any)
|
163 |
+
next_interaction = f"Interaction #{idx + 1}" if idx < len(rich_cards) else None
|
164 |
+
next_card = rich_cards[idx] if idx < len(rich_cards) else None
|
165 |
+
|
166 |
+
# Get buttons from current card
|
167 |
+
buttons = card.get("buttons", [])
|
168 |
+
|
169 |
+
# Generate dynamic quick replies based on *next* card
|
170 |
+
dynamic_quick_replies = []
|
171 |
+
if next_card:
|
172 |
+
dynamic_quick_replies = await generate_dynamic_buttons(
|
173 |
+
title=next_card.get("title", "News Summary"),
|
174 |
+
description=next_card.get("text", "Explore news and insights."),
|
175 |
+
url=next_card.get("url", ""),
|
176 |
+
next_interaction=next_interaction
|
177 |
+
)
|
178 |
+
|
179 |
+
# Define quick replies
|
180 |
+
quick_replies = dynamic_quick_replies + [
|
181 |
+
{
|
182 |
+
"type": "call",
|
183 |
+
"title": "Contact Support",
|
184 |
+
"payload": "+12345678901"
|
185 |
+
}
|
186 |
+
]
|
187 |
+
|
188 |
+
# Build message for current card
|
189 |
+
message = {
|
190 |
+
"text": f"{card['title']}\n\n{card['text']}",
|
191 |
+
"mediaType": "image",
|
192 |
+
"media": card.get("media", "") or "https://example.com/placeholder.jpg",
|
193 |
+
"richCard": {
|
194 |
+
"cardOrientation": "VERTICAL",
|
195 |
+
"mediaHeight": "MEDIUM"
|
196 |
+
},
|
197 |
+
"buttons": buttons,
|
198 |
+
"quickReplies": quick_replies
|
199 |
+
}
|
200 |
+
|
201 |
+
# Build interaction
|
202 |
+
interaction = {
|
203 |
+
"name": f"Interaction #{idx}",
|
204 |
+
"intents": ["show_content", f"content_{idx}"],
|
205 |
+
"actions": [
|
206 |
+
{
|
207 |
+
"send": {
|
208 |
+
"message": message
|
209 |
+
},
|
210 |
+
"type": "RichCard",
|
211 |
+
"name": f"Send Rich Card #{idx}"
|
212 |
+
}
|
213 |
+
]
|
214 |
+
}
|
215 |
+
interactions.append(interaction)
|
216 |
+
|
217 |
+
# Add welcome interaction
|
218 |
+
welcome_message = {
|
219 |
+
"text": f"Welcome to the {urlparse(url).netloc} RCS Bot! Explore the latest content.",
|
220 |
+
"richCard": {
|
221 |
+
"cardOrientation": "VERTICAL",
|
222 |
+
"mediaHeight": "MEDIUM"
|
223 |
+
},
|
224 |
+
"quickReplies": [
|
225 |
+
{
|
226 |
+
"type": "postback",
|
227 |
+
"title": "Start Exploring",
|
228 |
+
"payload": "start_exploring",
|
229 |
+
"execute": "Interaction #1"
|
230 |
+
}
|
231 |
+
]
|
232 |
+
}
|
233 |
+
welcome_interaction = {
|
234 |
+
"name": "Welcome Interaction",
|
235 |
+
"intents": ["start", "welcome"],
|
236 |
+
"actions": [
|
237 |
+
{
|
238 |
+
"send": {
|
239 |
+
"message": welcome_message
|
240 |
+
},
|
241 |
+
"type": "RichCard",
|
242 |
+
"name": "Send Welcome Message"
|
243 |
+
}
|
244 |
+
]
|
245 |
+
}
|
246 |
+
interactions.insert(0, welcome_interaction)
|
247 |
+
|
248 |
+
payload = {
|
249 |
+
"name": final_bot_name,
|
250 |
+
"interactions": interactions
|
251 |
+
}
|
252 |
+
|
253 |
+
# Log the payload for debugging
|
254 |
+
logging.info(f"NativeMSG bot payload: {json.dumps(payload, indent=2)}")
|
255 |
+
|
256 |
+
# Send request to NativeMSG API
|
257 |
+
connection = http.client.HTTPSConnection("api.nativemsg.com")
|
258 |
+
headers = {
|
259 |
+
"Authorization": f"Bearer {api_token}",
|
260 |
+
"Content-Type": "application/json"
|
261 |
+
}
|
262 |
+
connection.request("POST", "/v1/bots", json.dumps(payload), headers)
|
263 |
+
response = connection.getresponse()
|
264 |
+
response_data = response.read().decode('utf-8')
|
265 |
+
|
266 |
+
logging.info(f"NativeMSG bot creation response: Status {response.status}, Data: {response_data}")
|
267 |
+
|
268 |
+
if response.status != 200:
|
269 |
+
logging.error(f"Failed to create bot: {response_data}")
|
270 |
+
raise HTTPException(status_code=500, detail=f"Failed to create bot: {response_data}")
|
271 |
+
|
272 |
+
return json.loads(response_data)
|
273 |
+
|
274 |
+
except Exception as e:
|
275 |
+
logging.error(f"Error creating NativeMSG bot: {str(e)}")
|
276 |
+
raise HTTPException(status_code=500, detail=f"Failed to create bot: {str(e)}")
|
277 |
+
|
278 |
+
@app.get("/scrape")
|
279 |
+
async def crawl_website(
|
280 |
+
url: HttpUrl,
|
281 |
+
use_sample: bool = False,
|
282 |
+
bot_name: Optional[str] = None,
|
283 |
+
nativemsg_token: Optional[str] = None
|
284 |
+
):
|
285 |
+
"""Crawl a website or use sample data, generate rich cards, and create a NativeMSG bot."""
|
286 |
+
try:
|
287 |
+
# Determine API token
|
288 |
+
api_token = nativemsg_token or os.getenv("NATIVEMSG_API_TOKEN")
|
289 |
+
|
290 |
+
if use_sample:
|
291 |
+
results = SAMPLE_RICH_CARDS
|
292 |
+
else:
|
293 |
+
visited = set()
|
294 |
+
to_visit = {str(url)}
|
295 |
+
base_domain = urlparse(str(url)).netloc
|
296 |
+
results = []
|
297 |
+
|
298 |
+
while to_visit and len(visited) < 20: # Limited to 10 for demo
|
299 |
+
current_url = to_visit.pop()
|
300 |
+
if current_url in visited:
|
301 |
+
continue
|
302 |
+
visited.add(current_url)
|
303 |
+
|
304 |
+
logging.info(f"Scraping page: {current_url}")
|
305 |
+
page_data, new_links = await scrape_page(current_url, visited, base_domain)
|
306 |
+
if page_data:
|
307 |
+
logging.info(f"Scraped data: {page_data}")
|
308 |
+
summary = await quick_summarize(page_data["text"], page_data["url"])
|
309 |
+
rich_card = build_rich_card(page_data, summary)
|
310 |
+
rich_card["title"] = summary.get("title", "News Summary")
|
311 |
+
rich_card["url"] = page_data.get("url", str(url))
|
312 |
+
results.append(rich_card)
|
313 |
+
|
314 |
+
to_visit.update(new_links)
|
315 |
+
await asyncio.sleep(0.5)
|
316 |
+
|
317 |
+
if not results:
|
318 |
+
logging.error("No rich cards generated from scraping.")
|
319 |
+
raise HTTPException(status_code=400, detail="No content scraped from the provided URL.")
|
320 |
+
|
321 |
+
# Create NativeMSG bot with the rich cards
|
322 |
+
bot_response = await create_nativemsg_bot(results, str(url), bot_name, api_token)
|
323 |
+
|
324 |
+
logging.info(f"Final response: {results}, Bot: {bot_response}")
|
325 |
+
return {"rich_cards": results, "bot_response": bot_response}
|
326 |
+
|
327 |
+
except Exception as e:
|
328 |
+
logging.error(f"Scraping or bot creation failed: {str(e)}")
|
329 |
+
raise HTTPException(status_code=500, detail=f"Scraping or bot creation failed: {str(e)}")
|
330 |
+
|
331 |
+
@app.get("/", response_class=HTMLResponse)
|
332 |
+
async def serve_home(request: Request):
|
333 |
+
"""Serve the frontend HTML page."""
|
334 |
+
return templates.TemplateResponse("index.html", {"request": request})
|
335 |
+
|
336 |
+
if __name__ == "__main__":
|
337 |
+
import uvicorn
|
338 |
+
uvicorn.run(app, host="0.0.0.0", port=8001)
|
requirements.txt
ADDED
Binary file (194 Bytes). View file
|
|
rich_card_builder.py
ADDED
@@ -0,0 +1,71 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import logging
|
2 |
+
|
3 |
+
# Set up logging
|
4 |
+
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
|
5 |
+
|
6 |
+
def select_relevant_image(images: list, text: str) -> str:
|
7 |
+
"""Select a contextually relevant image from the list based on text."""
|
8 |
+
if not images:
|
9 |
+
logging.info("No images available, using empty media URL.")
|
10 |
+
return ""
|
11 |
+
|
12 |
+
# Prioritize images that are likely content-related (avoid logos, icons)
|
13 |
+
for img in images:
|
14 |
+
if not any(keyword in img.lower() for keyword in ["logo", "icon", "banner", "ad"]):
|
15 |
+
logging.info(f"Selected image: {img}")
|
16 |
+
return img
|
17 |
+
|
18 |
+
# Fallback to first image if no clear content image
|
19 |
+
logging.info(f"No content image found, using first image: {images[0]}")
|
20 |
+
return images[0]
|
21 |
+
|
22 |
+
def build_rich_card(scraped_data: dict, summary: dict) -> dict:
|
23 |
+
"""Build the rich card JSON using only scraped data and summary."""
|
24 |
+
logging.info(f"Building rich card with scraped_data: {scraped_data}, summary: {summary}")
|
25 |
+
|
26 |
+
# Select relevant image
|
27 |
+
media_url = select_relevant_image(scraped_data.get("images", []), scraped_data.get("text", ""))
|
28 |
+
|
29 |
+
# Use scraped URL
|
30 |
+
page_url = scraped_data.get("url", "")
|
31 |
+
|
32 |
+
# Use summary description
|
33 |
+
description = summary.get("description", "Explore news and insights.")
|
34 |
+
|
35 |
+
rich_card = {
|
36 |
+
"targets": [{"ids": [1368], "targetType": "humans"}],
|
37 |
+
"text": description,
|
38 |
+
"mediaType": "image",
|
39 |
+
"media": media_url,
|
40 |
+
"buttons": [
|
41 |
+
{
|
42 |
+
"type": "weburl",
|
43 |
+
"title": "View Now",
|
44 |
+
"payload": page_url
|
45 |
+
},
|
46 |
+
{
|
47 |
+
"type": "postback",
|
48 |
+
"title": "Learn More",
|
49 |
+
"payload": "learn_more",
|
50 |
+
"execute": None
|
51 |
+
}
|
52 |
+
],
|
53 |
+
"quickReplies": [
|
54 |
+
{
|
55 |
+
"type": "postback",
|
56 |
+
"title": "Show Similar",
|
57 |
+
"payload": "similar_content"
|
58 |
+
},
|
59 |
+
{
|
60 |
+
"type": "call",
|
61 |
+
"title": "Contact Support",
|
62 |
+
"payload": "+12345678901"
|
63 |
+
}
|
64 |
+
],
|
65 |
+
"richCard": {
|
66 |
+
"cardOrientation": "VERTICAL",
|
67 |
+
"mediaHeight": "MEDIUM"
|
68 |
+
}
|
69 |
+
}
|
70 |
+
logging.info(f"Generated rich card: {rich_card}")
|
71 |
+
return rich_card
|
scraper.py
ADDED
@@ -0,0 +1,73 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from playwright.async_api import async_playwright
|
2 |
+
from urllib.parse import urljoin, urlparse
|
3 |
+
import logging
|
4 |
+
|
5 |
+
# Set up logging
|
6 |
+
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
|
7 |
+
|
8 |
+
async def scrape_page(url: str, visited: set, base_domain: str) -> tuple[dict, set]:
|
9 |
+
"""Scrape a single page for text, images, and links using Playwright."""
|
10 |
+
try:
|
11 |
+
async with async_playwright() as p:
|
12 |
+
browser = await p.chromium.launch(headless=True)
|
13 |
+
context = await browser.new_context(
|
14 |
+
user_agent="Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/118.0.0.0 Safari/537.36",
|
15 |
+
viewport={"width": 1280, "height": 720}
|
16 |
+
)
|
17 |
+
page = await context.new_page()
|
18 |
+
await page.goto(url, wait_until="networkidle", timeout=30000)
|
19 |
+
await page.evaluate("window.scrollTo(0, document.body.scrollHeight)")
|
20 |
+
await page.wait_for_timeout(2000)
|
21 |
+
|
22 |
+
# Extract text content
|
23 |
+
text_content = await page.evaluate("document.body.innerText")
|
24 |
+
text_content = ' '.join(text_content.split()) if text_content else ""
|
25 |
+
|
26 |
+
# Extract images (only JPEG, PNG, WebP, exclude data URLs and SVGs)
|
27 |
+
images = await page.evaluate(
|
28 |
+
"""() => {
|
29 |
+
const validExtensions = ['.jpg', '.jpeg', '.png', '.webp'];
|
30 |
+
const imgElements = document.querySelectorAll('img');
|
31 |
+
const imgUrls = new Set();
|
32 |
+
imgElements.forEach(img => {
|
33 |
+
const src = img.src || '';
|
34 |
+
const dataSrc = img.dataset.src || '';
|
35 |
+
const srcset = img.srcset || '';
|
36 |
+
// Check src
|
37 |
+
if (src && !src.startsWith('data:') && validExtensions.some(ext => src.toLowerCase().endsWith(ext))) {
|
38 |
+
imgUrls.add(src);
|
39 |
+
}
|
40 |
+
// Check data-src
|
41 |
+
if (dataSrc && !dataSrc.startsWith('data:') && validExtensions.some(ext => dataSrc.toLowerCase().endsWith(ext))) {
|
42 |
+
imgUrls.add(dataSrc);
|
43 |
+
}
|
44 |
+
// Check srcset
|
45 |
+
if (srcset) {
|
46 |
+
srcset.split(',').forEach(src => {
|
47 |
+
const url = src.trim().split(' ')[0];
|
48 |
+
if (url && !url.startsWith('data:') && validExtensions.some(ext => url.toLowerCase().endsWith(ext))) {
|
49 |
+
imgUrls.add(url);
|
50 |
+
}
|
51 |
+
});
|
52 |
+
}
|
53 |
+
});
|
54 |
+
return Array.from(imgUrls);
|
55 |
+
}"""
|
56 |
+
)
|
57 |
+
images = [urljoin(url, img) for img in images if img]
|
58 |
+
|
59 |
+
# Extract links
|
60 |
+
links = await page.evaluate("Array.from(document.querySelectorAll('a')).map(a => a.href)")
|
61 |
+
links = set(urljoin(url, link) for link in links
|
62 |
+
if urlparse(urljoin(url, link)).netloc == base_domain
|
63 |
+
and urljoin(url, link) not in visited)
|
64 |
+
|
65 |
+
await browser.close()
|
66 |
+
|
67 |
+
page_data = {"url": url, "text": text_content, "images": images}
|
68 |
+
logging.info(f"Scraped data: url={url}, text_length={len(text_content)}, images={images}")
|
69 |
+
return page_data, links
|
70 |
+
|
71 |
+
except Exception as e:
|
72 |
+
logging.error(f"Error scraping {url}: {e}")
|
73 |
+
return {}, set()
|
summarizer.py
ADDED
@@ -0,0 +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}."[:100]
|
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-flash')
|
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)[:50]
|
169 |
+
description = match.group(2)[:100]
|
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}."[:100]
|
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}."[:100]
|
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}."[:100]
|
203 |
+
}
|
204 |
+
cache[cache_key] = result
|
205 |
+
return result
|