/About

Version 1.2.1 (Current)

PenineMate is a non-commercial, academic AI-powered movie assistant developed for educational and portfolio purposes. The application leverages multiple data sources, including the TMDb API for movie metadata, cast information, and popularity metrics, as well as Box Office and MovieLens datasets to enhance movie question answering and semantic-based recommendation features.

PenineMate is built using PostgreSQL for structured data storage and FAISS for efficient vector similarity search. The system's language understanding and reasoning capabilities are powered by OpenAI's GPT-5-nano model for natural language processing and intent classification, while vector embeddings are generated using OpenAI's text-embedding-3-small model (1536-dimensional) for semantic search and similarity matching.

Language Model

Model: OpenAI GPT-5-nano

Used for intent classification, query understanding, and natural language response generation

Embedding Model

Model: OpenAI text-embedding-3-small (1536 dimensions)

Used for semantic vector representation and similarity search in FAISS index

Updates:

  • •Language Model: OpenAI GPT-4o-mini → OpenAI GPT-5-nano
  • •Embedding Model: Sentence-Transformers (384d) → OpenAI text-embedding-3-small (1536d)
  • •Enhanced semantic search accuracy with 4x larger embedding dimensions
  • •Improved intent classification and natural language understanding
  • •Added franchise detection for 'How many X films?' type queries
  • •Better multilingual support for English and Indonesian conversations
  • •Optimized response generation with cloud-based LLM infrastructure

The project is already usable but still needs to be developed further.

Kaggle LogoTMDb LogoPEN Logo