Laravel Filament Solaris Laravel Package

Product Decisions This Supports

• Build vs. buy on AI-in-the-admin-panel. This package is the "buy" option for putting generated content (drafts, summaries, classifications) inside a Filament back office without building a provider integration from scratch. It directly informs whether AI assist is a one-sprint add-on or a bespoke platform effort.
• AI provider / hosting selection. Solaris is a Statik-operated offering with EU-oriented positioning. If your roadmap includes an AI feature touching customer or employee data, this package is a concrete way to satisfy data-residency and procurement questions rather than a generic OpenAI wrapper.
• Provider and model governance as a product policy, not a code detail. Release 0.4.1 lets AiGenerator carry a default provider, model, and timeout. That converts three previously per-call parameters into a configuration decision: which model tier each environment gets (cheap/fast in staging, stronger in production), what default latency budget the UX assumes, and where cost controls live. This is the kind of thing that otherwise gets hardcoded across features and becomes an audit finding.
• Timeout as a UX contract. A configurable default timeout forces an explicit product answer to "how long will an editor stare at a spinner before we degrade?" — synchronous generation with a hard ceiling versus queued/async with notification. Getting this into config early keeps that decision reversible.
• Roadmap sequencing for AI features. Because it is Filament-native, the cheapest AI wins are the ones living inside existing admin forms — copy drafting, record summarization, tagging suggestions, translation of CMS content. This package supports shipping those as experiments before committing to a customer-facing AI surface.
• Vendor-concentration risk tracking. Adopting it makes Solaris a named dependency in your architecture; the assessment should be revisited if the roadmap expands to public-facing or high-volume inference, where a thin Filament-oriented adapter stops being the right shape.

When to Consider This Package

Good fit when:

• You already run Filament and want AI generation inside admin forms with minimal glue code.
• You have (or are willing to take) a Solaris account and are comfortable with that vendor relationship.
• Volume is modest and interaction is editor-initiated — human-in-the-loop drafting rather than automated pipelines.
• You want central configuration of provider/model/timeout (0.4.1) so environments and cost tiers diverge without code changes.
• Your organization values EU-hosted inference and a smaller, auditable surface over breadth of capabilities.

Look elsewhere when:

• AI is a core product path rather than an admin convenience. A pre-1.0 package at 0.4.x with a narrow maintainer base is not the foundation for a revenue-critical feature; expect API churn between minor versions.
• You need streaming chat, tool/function calling, structured-output guarantees, embeddings, retrieval, or multi-provider failover — reach for Prism, the Laravel AI SDK, or direct provider SDKs, and keep the provider boundary under your own abstraction.
• You need public-facing, latency-sensitive inference or high-volume batch jobs. Nothing here is aimed at throughput, queue orchestration, or observability at scale.
• You want provider neutrality. This is a coupling to Solaris by design; migrating later means rewriting the call sites, so isolate it behind your own service interface if there's any chance the vendor changes.
• You need mature operational tooling — retries, cost dashboards, prompt versioning, evaluation harnesses. Not present at this stage.

How to Pitch It (Stakeholders)

For executives: This is a small, low-cost integration that puts AI-assisted drafting inside the tools our editors already use, with inference hosted in the EU via a known vendor. The 0.4.1 release centralizes which model we use and how long we wait, which means spend and performance are controlled from configuration rather than scattered through the codebase — good for cost governance and for answering procurement and data-handling questions. It is an experiment-scale commitment, not a platform bet.

For engineering: A thin, Filament-native adapter over the Solaris API. The 0.4.1 change moves default provider, model, and timeout onto AiGenerator, so call sites stop repeating those arguments and environment-level policy is expressed once. Dependency footprint is small and the integration surface is narrow — but it is 0.4.x, so pin the version, wrap the generator behind an internal interface, and treat the public API as unstable. If we later need streaming, retrieval, or a second provider, we replace the adapter rather than the calling code.

Recommended next step: Timebox a spike on a single non-critical admin form, pin the version, and set explicit default model and timeout values per environment before any wider rollout.