llm_models_spider

Auto-updated registry of LLM model capabilities (vision, audio, etc.).

Features

• Zero runtime dependencies - Pure compile-time lookups
• Auto-updated - Model lists are fetched daily from OpenRouter via GitHub Actions
• Fallback patterns - Unknown models are matched against common naming patterns
• Multiple capabilities - Vision, audio, video, and file input detection

Installation

[dependencies]
llm_models_spider = "0.1"

Usage

use llm_models_spider::{supports_vision, supports_audio, is_text_only, ModelCapabilities};

// Quick capability checks
assert!(supports_vision("gpt-4o"));
assert!(supports_vision("claude-3-sonnet-20240229"));
assert!(supports_vision("google/gemini-2.0-flash"));
assert!(!supports_vision("gpt-3.5-turbo"));

// Audio support
assert!(supports_audio("gemini-2.0-flash"));

// Text-only check
assert!(is_text_only("gpt-3.5-turbo"));
assert!(is_text_only("llama-3-70b"));

// Full capabilities lookup
if let Some(caps) = ModelCapabilities::lookup("gemini-2.0-flash") {
    println!("Vision: {}", caps.vision);  // true
    println!("Audio: {}", caps.audio);    // true
}

How It Works

1. GitHub Actions runs daily to fetch the latest model data from OpenRouter's API
2. Models are categorized by their input_modalities field:
   • ["text", "image"] → Vision model
   • ["text", "image", "audio"] → Vision + Audio model
   • ["text"] → Text-only model
3. The src/generated.rs file is updated and a new version is published to crates.io

Data Source

Model capabilities are sourced from OpenRouter's API:

• Vision models: input_modalities contains "image"
• Audio models: input_modalities contains "audio"
• Text-only: input_modalities is ["text"] only

Fallback Patterns

For models not yet in OpenRouter, the library falls back to pattern matching:

Manual Update

To manually update the model lists:

cargo run --bin update-models --features updater

License

MIT License - see LICENSE for details.