reloaded-code-models-dev

Sync the online models.dev catalog into a compact ModelCatalog with ETag caching, zstd compression, and offline fallback. ~3000 models in ~24 KiB.

Documentation · API Reference

Why this exists

If you run coding agents against many providers, you want to have fresh data. models.dev is one such source of data.

This crate downloads from models.dev, keeps only the fields we need, and builds a reloaded_code_core::models::ModelCatalog.

Usage

Load flow (simple)

1. Read cache header (if present) and get the old ETag.
2. Send request to models.dev with If-None-Match when ETag exists.
3. If server returns 304 Not Modified, load catalog from cache.
4. If server returns 200 OK, parse JSON, map it into catalog sources, write fresh cache, then build catalog.
5. If network fails, try cached data as fallback; if no valid cache exists, return an error.

Non-blocking (tokio)

use reloaded_code_models_dev::{CatalogLoadSource, ModelsDevCatalog};

#[cfg(feature = "tokio")]
async fn load_catalog() -> Result<(), Box<dyn std::error::Error>> {
    let result = ModelsDevCatalog::load().await?;

    match result.source {
        CatalogLoadSource::Downloaded => {
            println!("Downloaded fresh catalog data.")
        }
        CatalogLoadSource::NotModifiedCache => {
            println!("Cache is already up to date.")
        }
        CatalogLoadSource::FallbackCache => {
            println!("Network unavailable, using cached catalog data.")
        }
    }

    if let Some(entry) = result.catalog.lookup("openai", "gpt-4") {
        println!("provider api url: {}", entry.0.api_url);
        println!("max input tokens: {}", entry.1.max_input);
    }

    Ok(())
}

Blocking (blocking)

use reloaded_code_models_dev::{CatalogLoadSource, ModelsDevCatalog};

#[cfg(feature = "blocking")]
fn load_catalog() -> Result<(), Box<dyn std::error::Error>> {
    let result = ModelsDevCatalog::load()?;

    match result.source {
        CatalogLoadSource::Downloaded => {
            println!("Downloaded fresh catalog data.")
        }
        CatalogLoadSource::NotModifiedCache => {
            println!("Cache is already up to date.")
        }
        CatalogLoadSource::FallbackCache => {
            println!("Network unavailable, using cached catalog data.")
        }
    }

    if let Some(entry) = result.catalog.lookup("openai", "gpt-4") {
        println!("provider api url: {}", entry.0.api_url);
        println!("max input tokens: {}", entry.1.max_input);
    }

    Ok(())
}

Load from a custom cache path

use reloaded_code_models_dev::ModelsDevCatalog;
use std::path::PathBuf;

#[cfg(feature = "tokio")]
async fn load_catalog() -> Result<(), Box<dyn std::error::Error>> {
    let cache_path = PathBuf::from("/tmp/models-dev.cache");
    let _result = ModelsDevCatalog::load_at(&cache_path).await?;
    Ok(())
}

#[cfg(feature = "blocking")]
fn load_catalog() -> Result<(), Box<dyn std::error::Error>> {
    let cache_path = PathBuf::from("/tmp/models-dev.cache");
    let _result = ModelsDevCatalog::load_at(&cache_path)?;
    Ok(())
}

Resolve the shared cache path

use reloaded_code_models_dev::shared_cache_path;

fn print_cache_path() -> Result<(), Box<dyn std::error::Error>> {
    let path = shared_cache_path()?;
    println!("{}", path.display());
    Ok(())
}

Cache location

By default, cache is stored in the platform cache directory:

• Linux: ~/.cache/reloaded-code/models.dev.catalog.v1.cache
• macOS: ~/Library/Caches/reloaded-code/models.dev.catalog.v1.cache
• Windows: %LOCALAPPDATA%\reloaded-code\models.dev.catalog.v1.cache

Set RELOADED_CODE_MODELS_DEV_CACHE_PATH to override this path.

Cache size and performance

Current ballpark from a recent models.dev/api.json snapshot:

• Size: about 1.31 MiB JSON -> 109 KiB serialized payload -> 23.7 KiB compressed cache
• Compression: about 10.1 ms with current zstd level 17
• Decompression: about 0.057 ms (57 us) in --release
• Cache load into ModelCatalog: about 0.31 ms (read + decompress + decode + build)

Measured on a single core of a Ryzen 9950X3D; these are rough guidance numbers and will drift as the upstream catalog changes.

Feature flags

• tokio (default): async runtime support.
• blocking: synchronous runtime support.

Exactly one runtime mode must be enabled.

License

Apache-2.0