Skip to content

Compatible AI Services

Jump to bottom
smart-coder997 edited this page Jul 21, 2025 · 2 revisions

Compatible AI Services

Recommendarr is designed to work with any AI service that offers an OpenAI-compatible API endpoint for chat completions. This provides flexibility in choosing the AI model that best suits your needs and budget.

Supported Service Types

  • OpenAI API: The standard integration, supporting models like GPT-3.5 Turbo, GPT-4, GPT-4o, etc. Requires an API key and uses the standard https://api.openai.com/v1 endpoint.
  • Ollama: Allows running various open-source models locally. Provides an OpenAI-compatible endpoint (usually http://localhost:11434/v1).
  • LM Studio: Another popular tool for running models locally on your computer, also offering an OpenAI-compatible server (often http://localhost:1234/v1).
  • OpenRouter: A service that aggregates multiple AI models (including free and premium ones) behind a single OpenAI-compatible API. Useful for trying different models easily.
  • Anthropic Claude: Models like Claude 3 Opus, Sonnet, and Haiku can be accessed via OpenAI-compatible endpoints provided by Anthropic or through services like OpenRouter.
  • Other Self-hosted Models: Any self-hosted Large Language Model (LLM) that exposes an API conforming to the OpenAI chat completions standard can be used.
  • See AI Service Setup for instructions on configuring the API URL, key, and model in Recommendarr.

Recommended Models

Choosing the right model can impact the quality and cost of recommendations. Here are some suggestions, categorized for convenience (especially relevant if using OpenRouter):

Free / Low-Cost Options (via OpenRouter or Local)

  • Meta Llama 3.3 70B Instruct: Often available for free or very low cost on OpenRouter, offering great performance.
  • Google Gemini 2.0 Flash: Excellent recommendation quality, often with generous free tiers or low pricing via OpenRouter.
  • Mistral Small / Large: Good balance of performance and cost.
  • Local Models (Ollama/LM Studio): Models like Llama3, Mistral, Phi-3 can run entirely locally (performance depends on your hardware).

Premium Models (Higher Cost, Potentially Higher Quality)

  • Anthropic Claude 3.7 Sonnet/Haiku: Known for strong reasoning and understanding nuanced preferences. Haiku is often very cost-effective.
  • OpenAI GPT-4o / GPT-4o mini: State-of-the-art models offering excellent performance, with mini being a more affordable option.
  • Google Gemini 2.5 Pro: High-quality model available via OpenRouter or Google AI Studio.

Configuration Tips

  • Max Tokens: Setting this higher (e.g., 4000 or more) allows the AI to generate more detailed recommendations and reasoning.
  • Temperature: Controls randomness. Lower values (e.g., 0.2 - 0.5) produce more focused, predictable recommendations. Higher values (e.g., 0.7 - 1.0) encourage more creative or diverse suggestions. A good starting point is often around 0.6 - 0.8. Experiment with different models and settings in the AI Service Setup page to find what works best for your library and preferences.
Clone this wiki locally