LLM-Powered Smart Movie Recommendation System Based on LangChain and Streamlit

This article introduces the open-source project langchain-movie-recommendation, which uses the LangChain framework, OpenRouter API, and Streamlit to build an interactive interface. It can intelligently recommend high-rated movies based on users' preferred genres, languages, and eras. The project does not require historical user data, relies on LLM pre-trained knowledge, and features strong interpretability and quick deployment, making it an entry-level case for LLM application development.

Project Background

Movie recommendation has always been one of the core functions of content platforms. Traditional recommendation systems usually rely on collaborative filtering or content-based algorithms, which require a large amount of user behavior data as support. This project takes a different approach, using the reasoning ability of large language models to generate personalized recommendations based on users' explicit preferences (such as genre, language, and era).

Core Objectives

The core objectives of the project are: to allow users to select preferences through a simple interface, and the system calls LLM to analyze and return high-quality movie recommendations, including movie name, year, rating, synopsis, and recommendation reasons.

1. LangChain Framework

LangChain is responsible for managing prompt templates (PromptTemplate) and interactions with models. It converts user preferences into model-understandable instructions through structured prompt engineering.

2. OpenRouter API

Provides a unified interface to access multiple mainstream models (such as GPT, Claude, etc.), with good model replaceability.

3. Streamlit Frontend

Builds an interactive web interface using pure Python code, providing controls like drop-down menus and radio buttons to facilitate users' preference input.

4. Environment Variable Management

Uses python-dotenv to manage sensitive information such as API keys, enhancing security.

Step 1: User Input Users select movie genre, language preference, and era preference (old films/new films/both).

Step 2: Prompt Construction LangChain's PromptTemplate dynamically generates structured prompts based on user input, specifying recommendations for high IMDB-rated movies and output format.

Step 3: Model Reasoning The prompt is sent to LLM via OpenRouter API, and the model analyzes preferences and recommends movies based on pre-trained knowledge.

Step 4: Result Parsing and Display Parses the model's returned results and displays the movie name, year, language, IMDB rating, brief review, and recommendation reasons.

1. Zero Data Dependency

No need for historical user data or movie feature databases; relies on LLM pre-trained knowledge, suitable for cold start scenarios.

2. Strong Interpretability

Provides recommendation reasons, allowing users to understand the recommendation logic; transparency is higher than traditional black-box systems.

3. Quick Deployment

Few dependencies and simple architecture; can be deployed in a few minutes (install dependencies + configure API key).

4. Scalability

Modular design; can integrate IMDB API to get real-time ratings or add user rating functions.

This project is an entry-level case for LLM application development, covering:

• Prompt Engineering: Design structured prompts to guide model output
• API Integration: Securely call external LLM APIs
• Application Development: Quickly build interactive AI interfaces
• Environment Management: Protect sensitive information with environment variables

The idea can be extended to recommendation scenarios such as books, music, and restaurants.

• Automatically display movie posters
• Support recommending multiple movies at once (Top 5)
• Integrate IMDB API to get real-time data
• Add user rating filtering function
• Deploy to Streamlit Cloud for online access

The langchain-movie-recommendation project demonstrates the potential of LLM in practical applications, proving that even without complex machine learning backgrounds, one can quickly build useful AI applications using LangChain and Streamlit. Its open-source nature allows the community to improve it together, making it a project worth learning from.