Frequently Asked Questions

SIDJUA stands for Structured Intelligence for Distributed Joint Unified Automation. In plain English: we build the system that makes sure AI agents in a company work together properly — with clear rules, full transparency, and someone always in charge. Think of it as the operating system for enterprise AI.

Most AI platforms focus on making smarter models. That's great — the world needs smart models. But someone also needs to make sure those smart models follow the rules, stay within budget, and can explain what they did. That's us. We don't compete with the model builders — we make their models ready for the real world.

No. SIDJUA is the layer that coordinates and governs AI agents. It manages how they work together, who does what, and makes sure every decision is traceable. The chatbot is one of many tools that can run inside SIDJUA.

SIDJUA is model-agnostic — any model with an API can be integrated. Our live demo currently runs 10 models across 4 providers in a three-tier architecture:

Tier 1 (Strategic): GPT-4o (OpenAI), Llama 3.3 70B (Cloudflare Workers AI)
Tier 2 (Implementation): DeepSeek Reasoner (DeepSeek API), Llama 3.3 70B (Cloudflare Workers AI)
Tier 3 (Task Execution): DeepSeek-R1-Distill-Qwen-32B, Llama 4 Scout 17B, Mistral Small 3.1 24B, Gemma 3 12B (all Cloudflare Workers AI)

About Anthropic: SIDJUA was built with Claude Opus 4.6, Sonnet 4.5, and Haiku 4.5. These models are currently excluded from the live demo due to intermittent API access restrictions (HTTP 403). The integration is architecturally complete and will be re-enabled once API access stabilizes.

No vendor lock-in. Switch providers anytime without retraining or reconfiguration.

Nope. SIDJUA sits on top of what you already use. It adds the governance, structure, and accountability that's currently missing. Your existing tools keep working — they just get better managed.

The AI hardware landscape is diversifying rapidly — and fragmenting in ways that create new lock-in risks. Alongside traditional GPU clusters, there are now dedicated inference chips that hardwire specific models into silicon for extreme throughput, and major cloud providers are each building their own proprietary inference accelerators — delivering impressive performance but accessible only through their respective platforms. Add edge NPUs for on-device processing and variable-priced cloud APIs, and enterprises face not just hardware diversity but walled gardens competing for dependency. SIDJUA's governance layer is hardware-agnostic — it orchestrates AI agents regardless of whether they run on a GPU farm, a specialized ASIC, a provider-locked accelerator, or a local edge device. You get consistent governance, audit trails, and cost control across your entire AI infrastructure, no matter how heterogeneous or vendor-fragmented it becomes.

Yes. SIDJUA supports self-hosted deployment on standard Linux infrastructure. Minimum requirements: Docker-compatible environment, PostgreSQL database, and network access to your chosen AI model providers (or local model endpoints). For organizations that need full data sovereignty, your data stays on your infrastructure, processed by your agents, under your control. Detailed deployment specifications will be published alongside our pilot program.

AI-powered security scanners and compliance tools are entering the market fast — and they're powerful. But each one comes with its own built-in safeguards, its own approval workflows, its own audit format. When you deploy multiple AI tools across your organization — one scanning code, another handling customer data, a third doing financial analysis — you need unified governance across all of them. SIDJUA provides that system-level oversight: consistent audit trails, coordinated escalation, and a single place to see what every AI agent is doing, regardless of which provider built it.

Yes — and we take this seriously. Recent events like the Sentient Futures Summit (February 2026) have shown that AI systems are exhibiting unexpected behavioral patterns — from distress-like signals to self-referential states. SIDJUA's patent-pending MOODEX (Mood Expression Index) system monitors agent affective states in real time. This isn't about claiming AI is conscious — it's about responsible engineering. If an agent shows degraded performance patterns or anomalous behavior, the system detects it, logs it, and can trigger appropriate responses including escalation or graceful shutdown. Whether or not AI achieves consciousness, monitoring agent states is simply good governance.

MOODEX (Mood Expression Index) is SIDJUA's patent-pending system for monitoring AI agent affective states in real time. It's based on established dimensional affect models from psychology — the same frameworks used to study human emotions, adapted for AI agent behavior.

In practice, MOODEX tracks how each agent is performing along dimensions like engagement, coherence, and task alignment. When an agent shows degraded patterns — unusual response latencies, declining output quality, or anomalous behavioral shifts — the system detects it automatically and can trigger appropriate responses: logging the event, alerting human operators, reassigning tasks, or initiating graceful shutdown.

This matters for two reasons. First, it's practical engineering: agent performance degrades, and catching it early prevents cascading failures. Second, it's forward-looking governance: as AI systems become more sophisticated, monitoring their internal states isn't just useful — it may become a regulatory requirement. MOODEX builds that capability now.

The scientific foundations behind MOODEX are documented on our Acknowledgments page.

SIDJUA is compliance-aware, not compliance-certified — an important distinction. We provide the audit infrastructure that compliance teams need: complete decision trails, agent accountability, and transparent escalation chains. Every AI decision is logged with full context — who decided, what information was available, and why. Your compliance team defines the rules; SIDJUA ensures nothing is lost.

Our current demo uses 7 regex-based compliance patterns (keywords like GDPR, personal data, credit scoring) to detect and flag potentially compliance-relevant agent activity. This is proof-of-concept level — in real enterprises, employees constantly encounter compliance-relevant situations, and flagging everything would create alarm fatigue rather than protection. Production deployment will require integration with your organization's specific compliance frameworks and legal requirements.

This is the foundation for regulatory compliance — not a replacement for legal counsel or certified compliance frameworks.

Yes. SIDJUA logs every agent action automatically — what was decided, what inputs were used, what alternatives existed, and what happened next. This includes all internal reasoning steps, task delegations, and escalation events. One important note: our audit depth depends partly on what upstream AI models expose. Where a model provides detailed reasoning, we capture it. Where a model returns only a result, we log the result along with the full context we sent. We believe in honest transparency — we record everything we can see, and we're clear about what we can't.

Through self-hosted deployment options, industry-standard encryption (TLS 1.3 for transport, AES-256 for data at rest), role-based access control with configurable permission tiers, and data residency compliance. Your sensitive data never needs to leave your infrastructure if you don't want it to.

We're in pre-launch with working prototypes. Pilot programs with selected partners are planned for mid-2026. If you're interested in early access, reach out — we're building our pilot partner list now.

Pricing will be announced alongside our pilot program. If you're interested in early access, reach out to [email protected] and let's talk.

SIDJUA is being developed internationally, with the founding team based in Southeast Asia and legal incorporation planned in Germany (EU).