Storage, AI & Real-time

Storage

Storage is a facade over object_store 0.14 (S3/GCS/Azure) with a local disk on tokio::fs.

cargo artisan make:controller UploadController

AI SDK

rustasea-ai defines a provider-agnostic trait with real HTTP adapters via reqwest. provider_from_env() supports OpenAI, Anthropic, Azure, Groq, xAI, DeepSeek, Mistral, Ollama, OpenRouter, and OpenAI-compatible endpoints. Gemini and Bedrock have no adapter yet and return AiError::UnknownProvider.

Agents and tools are first-class:

cargo artisan make:agent SupportAgent
cargo artisan make:tool SearchOrders

let agent = SupportAgent::new()
    .with_tool(SearchOrders)
    .stream("Where is my order?")
    .await?;

JSON:API resources

JsonApiResource is serde-based and renders Content-Type: application/vnd.api+json, with sparse fieldsets, relationship inclusion, links, and headers.

Broadcasting and real-time

config/broadcasting.toml selects the connection; the in-process hub is the zero-dependency default, and pusher/redis connections are feature-gated.

Presentation layers

Mail and notifications

rustasea-mail provides Mailable, Mailer implementations (ArrayMailer, LogMailer, SmtpMailer), and queued notifications with #[deleteWhenMissingModels].

Observability

Both are off by default:

• rustasea-debugbar — per-request SQL/cache/event capture into an in-memory ring buffer with a JSON/HTML surface.
• rustasea-queue-dashboard — live queue depth and age, queue_metrics history with retention, failed-job retry/forget, and worker heartbeats.