🎬 Otaku Concierge Pro

A fully local, multi-agent entertainment concierge system that recommends movies, web series, and anime, powered by Google ADK and SQLite-based long-term memory.

1. Objective

Most streaming platforms recommend generic content and do not remember personal tastes deeply.
Otaku Concierge Pro solves that by:

• Learning YOUR taste
• Remembering it using SQLite
• Ranking suggestions by IMDB
• Providing anime/movie/series suggestions
• Logging your watch history
• Improving over time

This makes it a standalone “AI entertainment assistant”.

2. Architecture (Local Multi-Agent System)

All agents run inside a single Python process, coordinated by the root agent.

2.1 TasteProfilingAgent

Learns & stores:

• Favourite genres
• Disliked genres
• Preferred runtime (short / medium / long)
  Stored directly into SQLite.

2.2 TasteSummaryAgent

Builds an automatic summary from:

• Preferences
• Watch history
  This summary is injected into the recommendation agent.

2.3 RecommendationAgent

Uses:

• google_search
• IMDB-style ranking
• Genre matching
• Runtime matching
• Taste alignment
  Returns a ranked list of titles + reasons.
  Also logs top recommendations.

2.4 FeedbackAgent

Understands:

• What the user watched
• What rating they gave
• Whether they liked it
  Stores all of this in watch_history and updates preferences accordingly.

2.5 RootAgent (Coordinator)

Determines user intention:

• “I like sci-fi” → TasteProfilingAgent
• “Recommend me something” → Summary → RecommendationAgent
• “I watched Naruto and loved it” → FeedbackAgent
• “What do you remember about my taste?” → TasteSummaryAgent

Install dependencies

pip install -r requirements.txt

Set your Gemini API key

Linux/macOS: export GOOGLE_API_KEY="your_key_here"

Windows PowerShell: $env:GOOGLE_API_KEY="your_key_here"

start

python main.py