YouTube Recommender Skill

When to Use This Skill

Use this skill when the user asks for:

• •YouTube video recommendations for any topic
• •Best videos for learning or entertainment
• •Videos of specific durations (short, medium, long)
• •Educational content, tutorials, or entertainment

Commands

ytrec <topic> <duration>

Arguments:

• •topic: Search term (e.g., "machine learning", "guitar tutorial", "cooking pasta")
• •duration: Video length preference
  • •tiny - Less than 5 minutes
  • •short - 5 to 20 minutes
  • •long - 20+ minutes

Example Usage:

cd ~/.nanobot/workspace/skills/youtube-recommender
python youtube_recommender.py "react hooks tutorial" short
python youtube_recommender.py "space exploration" long
python youtube_recommender.py "cooking tips" tiny

Python API

For programmatic use within nanobot or other Python code:

import sys
sys.path.insert(0, '/home/ubuntu/.nanobot/workspace/skills/youtube-recommender')
import youtube_recommender

# Get recommendations
videos = youtube_recommender.get_recommendations(
    topic="python tutorial",
    duration="short",
    backend="youtube",  # or "invidious"
    num_results=5
)

# Format a video for display
formatted = youtube_recommender.format_recommendation(videos[0])
print(formatted)

# Access raw data
for video in videos:
    print(f"{video['title']} by {video['author']}")
    print(f"Score: {video['score']:.2f}")
    print(f"URL: https://youtube.com/watch?v={video['videoId']}")

API Functions

get_recommendations(topic, duration='short', backend=None, num_results=5)

Returns a list of video dictionaries sorted by recommendation score.

Parameters:

• •topic (str): Search query
• •duration (str): 'tiny', 'short', or 'long' (default: 'short')
• •backend (str): 'youtube' or 'invidious' (default: auto-detect)
• •num_results (int): Number of videos to return (default: 5)

Returns:

• •list: List of video dicts with keys: videoId, title, author, lengthSeconds, viewCount, likeCount, published, score

Raises:

• •ValueError: Invalid duration parameter
• •APIKeyError: YouTube API key missing or invalid
• •BackendError: Backend failed to fetch results

format_recommendation(video, include_explanation=True)

Formats a video dictionary as a human-readable string.

Parameters:

• •video (dict): Video data dict
• •include_explanation (bool): Include "Why this video" section

Returns:

• •str: Formatted recommendation string with emojis

How to Execute

1. •

   Check API Key: The script requires YOUTUBE_API_KEY environment variable

   • •Auto-detects from shell environment
   • •Get YouTube API key from: https://console.cloud.google.com/
   • •Add to ~/.bashrc:
     bash

     export YOUTUBE_API_KEY="your-api-key"
     source ~/.bashrc
     

2. •

   Run the Script: Execute using Python:

   bash

   cd ~/.nanobot/workspace/skills/youtube-recommender
   python youtube_recommender.py "<topic>" "<duration>"
   

3. •

   Return Results: Present the full output to the user, including:

   • •Top Pick with title, channel, duration, views, and explanation
   • •"Also worth watching" section with additional recommendations
   • •Direct YouTube links

Output Format

The script will return structured output with emojis:

Top Pick:

• •🎬 Title
• •📺 Channel name
• •⏱️ Duration
• •👀 Views
• •📅 Posted date
• •📝 Why this video (explanation)
• •🔗 Watch link

Also worth watching:

• •3 additional recommended videos
• •Each with duration, view count, and link

Security Notes

• •The API key is loaded from the environment, never from files
• •The script uses YouTube Data API v3 or Invidious
• •No user input is directly executed as commands
• •The search query is safely passed as arguments

Troubleshooting

If the script fails:

1. •Check if YOUTUBE_API_KEY is set: echo $YOUTUBE_API_KEY
2. •If not set, instruct user to add to ~/.bashrc:
   bash

   echo 'export YOUTUBE_API_KEY="your-api-key"' >> ~/.bashrc
   source ~/.bashrc
   

3. •For API quota issues, suggest checking quota in Google Cloud Console
4. •Invidious backend is unreliable - public instances frequently fail with 403/404 errors. Use YouTube API for reliable results.

Duration Filters:

• •tiny = < 5 minutes (quick tips, overviews)
• •short = 5-20 minutes (tutorials, explainers)
• •long = 20+ minutes (deep dives, documentaries)

Algorithm Notes

The recommendation uses a weighted scoring system:

• •View count (logarithmic scaling - prevents viral-only results)
• •Duration bonus (+2 for medium 5-20min videos)
• •Recency bonus (slight boost for newer content)

This ensures a balance between popularity, content quality, and freshness.