Key facts
| API compatibility | OpenAI-compatible /v1/chat/completions - change base_url and model name |
| Models | 30+ models behind one API, from free tiers to frontier reasoning |
| Nearest data plane | GCC (UAE); region-locked, so prompts, completions, embeddings and logs stay in-region |
| Deployment | Plugsky cloud, your VPC, on-prem or air-gapped |
| RAG and embeddings | Embeddings and retrieval are live for private document search |
| Pricing model | Flat monthly self-serve plans; no per-token billing on self-serve |
| Free tier | Free plan with 2 free AI models (plugsky-micro, plugsky-lite), no card |
TL;DR
- Pin the GCC (UAE) plane so prompts, completions, embeddings and logs stay in one jurisdiction.
- An OpenAI-compatible API means a base-URL change, not a rewrite, for Riffa teams.
- 30+ models behind one API, from free tiers for pilots to frontier reasoning models.
- Deploy in a VPC, on-prem or air-gapped when Bahrain's data-protection law obligations or internal policy require it.
- Evaluate on the free plan or the 14-day full-access trial before committing.
How it works, step by step
- Map which Riffa workloads touch personal or regulated data, and classify each one.
- Choose a data plane; for Bahrain the nearest documented option is GCC (UAE).
- Create a Plugsky account and API key on the free plan (two models, no card), or use the 14-day full-access trial.
- Change base_url to api.plugsky.com, map model names, and run your existing tests and evals.
- Measure latency from your Riffa network and set timeouts and retries for your workload.
- Add scoped keys and logging, then decide whether a VPC, on-prem or air-gapped deployment is required.
- Document the residency decision - region, sub-processors and log retention - for auditors.
Original data
Try it yourself
Open the AI data residency checklist →
Why Riffa teams choose where AI runs
Riffa is Bahrain's largest city by area in the Southern Governorate, combining residential districts, sports facilities and growing commercial activity. AI adoption is mostly internal: service desks, HR and document search, with some customer-facing assistants. Bahrain's personal data protection law applies, so even internal tools need scoped keys and a documented region.
Production workloads in Riffa usually sit in internal operations, retail and services and sports and events. Bahrain's personal data protection law applies to personal data. That turns region choice into an architecture decision: inference, embeddings, retrieval and logs should be pinned to the same place, and the choice should be provable with documentation rather than assumed from a vendor page.
How an AI cloud deployment in Riffa works on Plugsky
Plugsky exposes an OpenAI-compatible /v1/chat/completions endpoint, so migration is a base URL change plus model mapping: your SDK, streaming, function calling and JSON mode keep working. The catalogue holds 30+ models, from the free plugsky-micro and plugsky-lite tiers to frontier reasoning models. Test Arabic and English prompts on internal documents.
Plugsky supports region-locked data planes; see /data-residency. For Bahrain, the nearest documented plane is GCC (UAE), and prompts, completions, embeddings and logs stay in-region by architecture. Enterprise adds customer-managed keys, zero-knowledge mode and right-to-audit clauses; where in-country processing is mandatory, the same API runs in your VPC, on-prem or air-gapped.
Latency, failover and multi-region design
Riffa reaches the GCC plane in the UAE over short Gulf routes; latency is low enough for interactive internal assistants.
Keep continuity same-region: exported logs, documented RTO and RPO, and a written stance on cross-border fallback.
A rollout plan that fits Riffa
Start with one internal operations workflow and a fifty-question evaluation set drawn from real cases. Pin the GCC (UAE) plane while you evaluate on the free plan (plugsky-micro and plugsky-lite, no card), or use the 14-day full-access trial for larger models, then measure answer quality and round-trip time from your own network. Before production, settle scoped keys with rotation, log retention, a model allow-list tied to evaluations, and a deployment choice that follows from Bahrain's data-protection law obligations.
Honest comparison
| Capability | Plugsky | Single-endpoint AI API | Building in-house |
|---|---|---|---|
| Region options | EU, GCC, APAC and US planes; Riyadh on Enterprise | Usually one global endpoint | You operate every region |
| Residency enforcement | Region-locked by architecture | Often contractual only | You build the controls |
| Migration effort | Base URL plus model mapping | Vendor-specific SDK | Full rewrite |
| Model choice | 30+ models behind one API | Single-vendor catalogue | You host each model |
| Pricing model | Flat monthly self-serve, no per-token billing | Per-token metering | GPU plus operations cost |
| Endpoint roadmap | Audio, images, moderation, files, batch and fine-tuning coming soon | Varies by vendor | Separate pipelines to maintain |
Frequently asked questions
Which region should a Riffa workload use?
The nearest documented plane for Bahrain is GCC (UAE); verify current availability at /data-residency before pinning production.
Does Plugsky have a data center in Riffa?
Plugsky does not publish city-level data center locations. Region availability is listed at /data-residency, and enterprise teams can deploy in their own VPC, on-prem or air-gapped environment.
Is there a free plan?
Yes - the free plan includes two free AI models (plugsky-micro and plugsky-lite) with no credit card required.
How does pricing work?
Self-serve plans are flat monthly with unlimited fair-use usage and no per-token billing; see the live pricing page for current plans.
Can we keep our existing OpenAI SDK code?
Yes. Change the base URL to api.plugsky.com and map model names; streaming, function calling and JSON mode continue to work.
Can data stay inside our own network?
Yes. VPC, on-prem and air-gapped deployments are available, with customer-managed keys and audit-log export on Enterprise.
Which endpoints are still coming soon?
Audio, images, moderation, files, batch and fine-tuning are coming soon; chat, streaming, JSON mode, function calling, embeddings and RAG are live.
Is a trial available?
Yes - a 14-day full-access trial is available alongside the free plan for evaluation.