Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,223 +1,100 @@
|
|
| 1 |
-
from smolagents import
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
import re
|
| 3 |
-
from
|
| 4 |
-
|
|
|
|
| 5 |
|
| 6 |
@tool
|
| 7 |
-
def
|
| 8 |
-
"""
|
| 9 |
-
Searches for information about a YouTuber using DuckDuckGo.
|
| 10 |
-
|
| 11 |
Args:
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
Returns:
|
| 15 |
-
Dictionary containing search results about the YouTuber
|
| 16 |
"""
|
| 17 |
search_tool = DuckDuckGoSearchTool()
|
| 18 |
|
| 19 |
-
# First search for the
|
| 20 |
-
|
| 21 |
|
| 22 |
-
# Then search for
|
| 23 |
-
|
| 24 |
|
| 25 |
-
#
|
| 26 |
-
|
| 27 |
|
| 28 |
-
#
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
}
|
| 37 |
-
|
| 38 |
-
@tool
|
| 39 |
-
def extract_youtube_info(search_results: Dict[str, Any]) -> Dict[str, Any]:
|
| 40 |
-
"""
|
| 41 |
-
Extracts relevant YouTuber information from search results.
|
| 42 |
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
|
| 50 |
-
channel_url = None
|
| 51 |
-
for result in search_results.get("channel_results", []):
|
| 52 |
-
if "youtube.com/channel/" in result.get("url", "") or "youtube.com/c/" in result.get("url", "") or "youtube.com/@" in result.get("url", ""):
|
| 53 |
-
channel_url = result.get("url")
|
| 54 |
break
|
| 55 |
|
| 56 |
-
#
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
for result in search_results.get("stats_results", []):
|
| 60 |
-
desc = result.get("desc", "")
|
| 61 |
-
# Look for subscriber patterns like "10M subscribers" or "10 million subscribers"
|
| 62 |
-
sub_match = re.search(r"(\d+(?:\.\d+)?[KMB]?)\s+subscribers", desc, re.IGNORECASE)
|
| 63 |
-
if sub_match:
|
| 64 |
-
subscriber_count = sub_match.group(1)
|
| 65 |
-
|
| 66 |
-
# Look for view patterns
|
| 67 |
-
view_match = re.search(r"(\d+(?:\.\d+)?[KMB]?)\s+views", desc, re.IGNORECASE)
|
| 68 |
-
if view_match:
|
| 69 |
-
view_count = view_match.group(1)
|
| 70 |
-
|
| 71 |
-
# Extract recent videos/content
|
| 72 |
-
recent_videos = []
|
| 73 |
-
for result in search_results.get("content_results", []):
|
| 74 |
-
if "video" in result.get("desc", "").lower() or "upload" in result.get("desc", "").lower():
|
| 75 |
-
recent_videos.append({
|
| 76 |
-
"title": result.get("title"),
|
| 77 |
-
"url": result.get("url"),
|
| 78 |
-
"description": result.get("desc")
|
| 79 |
-
})
|
| 80 |
-
if len(recent_videos) >= 3:
|
| 81 |
-
break
|
| 82 |
|
| 83 |
-
# Extract
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
"title": result.get("title"),
|
| 89 |
-
"url": result.get("url"),
|
| 90 |
-
"description": result.get("desc")
|
| 91 |
-
})
|
| 92 |
-
if len(background_info) >= 2:
|
| 93 |
-
break
|
| 94 |
|
| 95 |
-
return
|
| 96 |
-
"channel_url": channel_url,
|
| 97 |
-
"subscriber_count": subscriber_count,
|
| 98 |
-
"view_count": view_count,
|
| 99 |
-
"recent_videos": recent_videos,
|
| 100 |
-
"background_info": background_info,
|
| 101 |
-
"raw_results": {
|
| 102 |
-
"channel": search_results.get("channel_results", [])[:3],
|
| 103 |
-
"stats": search_results.get("stats_results", [])[:3],
|
| 104 |
-
"content": search_results.get("content_results", [])[:3],
|
| 105 |
-
"background": search_results.get("background_results", [])[:3]
|
| 106 |
-
}
|
| 107 |
-
}
|
| 108 |
|
| 109 |
@tool
|
| 110 |
-
def
|
| 111 |
-
"""
|
| 112 |
-
Formats YouTuber information with proper citations.
|
| 113 |
-
|
| 114 |
Args:
|
| 115 |
-
|
| 116 |
-
info: Dictionary containing YouTuber information
|
| 117 |
-
|
| 118 |
-
Returns:
|
| 119 |
-
Formatted information with citations
|
| 120 |
"""
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
| 128 |
-
|
| 129 |
-
|
| 130 |
-
if info.get("subscriber_count"):
|
| 131 |
-
result += f"- **Subscribers:** {info.get('subscriber_count')} \n"
|
| 132 |
-
|
| 133 |
-
if info.get("view_count"):
|
| 134 |
-
result += f"- **Views:** {info.get('view_count')} \n"
|
| 135 |
-
|
| 136 |
-
# Add channel source citation
|
| 137 |
-
if raw_results.get("channel") and len(raw_results.get("channel")) > 0:
|
| 138 |
-
source = raw_results.get("channel")[0]
|
| 139 |
-
result += f"\n: [{source.get('title')}]({source.get('url')})\n"
|
| 140 |
-
|
| 141 |
-
# Add stats source citation
|
| 142 |
-
if raw_results.get("stats") and len(raw_results.get("stats")) > 0:
|
| 143 |
-
source = raw_results.get("stats")[0]
|
| 144 |
-
result += f": [{source.get('title')}]({source.get('url')})\n"
|
| 145 |
-
|
| 146 |
-
# Recent content
|
| 147 |
-
if info.get("recent_videos") and len(info.get("recent_videos")) > 0:
|
| 148 |
-
result += "\n## Recent Content\n"
|
| 149 |
-
for i, video in enumerate(info.get("recent_videos")):
|
| 150 |
-
result += f"- **{video.get('title')}** $CITE_{3+i}\n"
|
| 151 |
-
result += f" {video.get('description')}\n"
|
| 152 |
-
result += f" [Link]({video.get('url')})\n\n"
|
| 153 |
-
|
| 154 |
-
# Content citations
|
| 155 |
-
for i, source in enumerate(raw_results.get("content", [])[:len(info.get("recent_videos", []))]):
|
| 156 |
-
result += f"$CITE_{3+i}: [{source.get('title')}]({source.get('url')})\n"
|
| 157 |
-
|
| 158 |
-
# Background information
|
| 159 |
-
if info.get("background_info") and len(info.get("background_info")) > 0:
|
| 160 |
-
result += "\n## Background & History\n"
|
| 161 |
-
cite_start = 3 + len(info.get("recent_videos", []))
|
| 162 |
-
for i, bg in enumerate(info.get("background_info")):
|
| 163 |
-
result += f"- **{bg.get('title')}** $CITE_{cite_start+i}\n"
|
| 164 |
-
result += f" {bg.get('description')}\n\n"
|
| 165 |
-
|
| 166 |
-
# Background citations
|
| 167 |
-
cite_start = 3 + len(info.get("recent_videos", []))
|
| 168 |
-
for i, source in enumerate(raw_results.get("background", [])[:len(info.get("background_info", []))]):
|
| 169 |
-
result += f"$CITE_{cite_start+i}: [{source.get('title')}]({source.get('url')})\n"
|
| 170 |
-
|
| 171 |
-
return result
|
| 172 |
-
|
| 173 |
-
# Create the specialized YouTuber information agent
|
| 174 |
-
search_tool = DuckDuckGoSearchTool()
|
| 175 |
|
| 176 |
-
|
| 177 |
-
youtuber_info_agent = Agent(
|
| 178 |
-
system_prompt="""You are a specialized YouTuber information agent. Your task is to find and provide detailed information about YouTube creators when given a YouTuber's name.
|
| 179 |
|
| 180 |
-
|
| 181 |
-
|
| 182 |
-
|
| 183 |
-
|
| 184 |
-
|
| 185 |
-
3. Format the information with proper citations using the format_youtuber_profile tool
|
| 186 |
-
4. Present the information to the user in a well-organized format
|
| 187 |
-
|
| 188 |
-
Always verify the information before presenting it and indicate if certain information couldn't be retrieved.
|
| 189 |
-
""",
|
| 190 |
-
tools=[
|
| 191 |
-
search_tool,
|
| 192 |
-
search_youtuber,
|
| 193 |
-
extract_youtube_info,
|
| 194 |
-
format_youtuber_profile
|
| 195 |
-
],
|
| 196 |
-
max_steps=8,
|
| 197 |
-
verbose=True
|
| 198 |
)
|
| 199 |
|
| 200 |
-
#
|
| 201 |
-
|
| 202 |
|
| 203 |
-
|
| 204 |
-
|
| 205 |
-
|
| 206 |
-
|
| 207 |
-
|
| 208 |
-
#
|
| 209 |
-
|
| 210 |
-
|
| 211 |
-
|
| 212 |
-
|
| 213 |
-
|
| 214 |
-
description=
|
| 215 |
-
|
| 216 |
-
["MrBeast"],
|
| 217 |
-
["PewDiePie"],
|
| 218 |
-
["Marques Brownlee"]
|
| 219 |
-
]
|
| 220 |
)
|
| 221 |
|
| 222 |
-
|
| 223 |
-
interface.launch(share=True)
|
|
|
|
| 1 |
+
from smolagents import CodeAgent, DuckDuckGoSearchTool, HfApiModel, load_tool, tool
|
| 2 |
+
import datetime
|
| 3 |
+
import requests
|
| 4 |
+
import pytz
|
| 5 |
+
import yaml
|
| 6 |
import re
|
| 7 |
+
from tools.final_answer import FinalAnswerTool
|
| 8 |
+
|
| 9 |
+
from Gradio_UI import GradioUI
|
| 10 |
|
| 11 |
@tool
|
| 12 |
+
def get_youtube_channel_info(channel_name: str) -> str:
|
| 13 |
+
"""A tool that fetches information about a YouTube channel including subscriber count.
|
|
|
|
|
|
|
| 14 |
Args:
|
| 15 |
+
channel_name: The name of the YouTube channel to search for.
|
|
|
|
|
|
|
|
|
|
| 16 |
"""
|
| 17 |
search_tool = DuckDuckGoSearchTool()
|
| 18 |
|
| 19 |
+
# First search for the channel's basic information
|
| 20 |
+
basic_search_results = search_tool(f"{channel_name} youtube channel info")
|
| 21 |
|
| 22 |
+
# Then specifically search for subscriber count
|
| 23 |
+
sub_search_results = search_tool(f"{channel_name} youtube channel subscribers count")
|
| 24 |
|
| 25 |
+
# Extract relevant information from search results
|
| 26 |
+
channel_info = f"Information about YouTube channel '{channel_name}':\n\n"
|
| 27 |
|
| 28 |
+
# Try to extract subscriber count using regex patterns
|
| 29 |
+
subscriber_count = "Could not determine subscriber count"
|
| 30 |
+
sub_patterns = [
|
| 31 |
+
r'(\d+(?:\.\d+)?)\s*(?:M|million)\s+subscribers',
|
| 32 |
+
r'(\d+(?:\.\d+)?)\s*(?:K|thousand)\s+subscribers',
|
| 33 |
+
r'(\d+)\s+subscribers',
|
| 34 |
+
r'subscribers:\s*(\d+(?:\.\d+)?(?:\s*[KM])?)',
|
| 35 |
+
]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
|
| 37 |
+
for result in sub_search_results + basic_search_results:
|
| 38 |
+
for pattern in sub_patterns:
|
| 39 |
+
matches = re.search(pattern, result, re.IGNORECASE)
|
| 40 |
+
if matches:
|
| 41 |
+
subscriber_count = matches.group(1)
|
| 42 |
+
break
|
| 43 |
+
if subscriber_count != "Could not determine subscriber count":
|
|
|
|
|
|
|
|
|
|
|
|
|
| 44 |
break
|
| 45 |
|
| 46 |
+
# Compile all information
|
| 47 |
+
channel_info += f"Channel Name: {channel_name}\n"
|
| 48 |
+
channel_info += f"Subscriber Count: {subscriber_count}\n"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 49 |
|
| 50 |
+
# Extract other relevant information
|
| 51 |
+
for result in basic_search_results:
|
| 52 |
+
if "about" in result.lower() and channel_name.lower() in result.lower():
|
| 53 |
+
channel_info += f"Description: {result}\n"
|
| 54 |
+
break
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 55 |
|
| 56 |
+
return channel_info
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 57 |
|
| 58 |
@tool
|
| 59 |
+
def get_current_time_in_timezone(timezone: str) -> str:
|
| 60 |
+
"""A tool that fetches the current local time in a specified timezone.
|
|
|
|
|
|
|
| 61 |
Args:
|
| 62 |
+
timezone: A string representing a valid timezone (e.g., 'America/New_York').
|
|
|
|
|
|
|
|
|
|
|
|
|
| 63 |
"""
|
| 64 |
+
try:
|
| 65 |
+
# Create timezone object
|
| 66 |
+
tz = pytz.timezone(timezone)
|
| 67 |
+
# Get current time in that timezone
|
| 68 |
+
local_time = datetime.datetime.now(tz).strftime("%Y-%m-%d %H:%M:%S")
|
| 69 |
+
return f"The current local time in {timezone} is: {local_time}"
|
| 70 |
+
except Exception as e:
|
| 71 |
+
return f"Error fetching time for timezone '{timezone}': {str(e)}"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 72 |
|
| 73 |
+
final_answer = FinalAnswerTool()
|
|
|
|
|
|
|
| 74 |
|
| 75 |
+
model = HfApiModel(
|
| 76 |
+
max_tokens=2096,
|
| 77 |
+
temperature=0.5,
|
| 78 |
+
model_id='Qwen/Qwen2.5-Coder-32B-Instruct',
|
| 79 |
+
custom_role_conversions=None,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 80 |
)
|
| 81 |
|
| 82 |
+
# Import tool from Hub
|
| 83 |
+
image_generation_tool = load_tool("agents-course/text-to-image", trust_remote_code=True)
|
| 84 |
|
| 85 |
+
with open("prompts.yaml", 'r') as stream:
|
| 86 |
+
prompt_templates = yaml.safe_load(stream)
|
| 87 |
+
|
| 88 |
+
agent = CodeAgent(
|
| 89 |
+
model=model,
|
| 90 |
+
tools=[get_youtube_channel_info, get_current_time_in_timezone, DuckDuckGoSearchTool(), final_answer], # Added your custom tools
|
| 91 |
+
max_steps=6,
|
| 92 |
+
verbosity_level=1,
|
| 93 |
+
grammar=None,
|
| 94 |
+
planning_interval=None,
|
| 95 |
+
name=None,
|
| 96 |
+
description=None,
|
| 97 |
+
prompt_templates=prompt_templates
|
|
|
|
|
|
|
|
|
|
|
|
|
| 98 |
)
|
| 99 |
|
| 100 |
+
GradioUI(agent).launch()
|
|
|