Sylor AI:
The Conversational Intelligence Layer of the Orzatty Ecosystem

Abstract

Sylor is Orzatty's flagship AI assistant — a production-grade, multi-provider conversational intelligence system deployed on Cloudflare's global edge network. It combines a resilient provider routing layer, a sovereign model roadmap, and deep integration with the Orzatty ecosystem (OrzattyAccount, OrzattyDrive, O-Suite) to deliver a private, fast, and extensible AI experience across web and native platforms.

1. Motivation

The current AI landscape is dominated by centralized, opaque systems that treat user data as a resource. Sylor is built on a different premise: intelligence should be fast, honest, and accountable to the user — not to the model provider. By routing across multiple open providers and progressively replacing them with a sovereign model trained and controlled by Orzatty, Sylor moves toward an AI that is genuinely owned by its ecosystem.

2. Architecture Overview

Sylor is composed of three layers:

2.1 Edge Worker (Cloudflare Workers)
The API layer runs entirely on Cloudflare Workers — stateless, globally distributed, with zero cold starts. It handles authentication via OrzattyAccount JWT tokens, rate limiting, conversation persistence (D1 + R2), and streaming SSE responses to clients.

2.2 Tool Calling Orchestration
Sylor supports native tool calling with automatic stream interception. When a provider emits a tool call, the router pauses the stream, executes the tool (e.g., search_web), appends the result to the message chain, and recurses to generate the final response. This enables real-time web search, structured data retrieval, and future tool integrations without client-side changes.

3. Sovereign Model Roadmap

Orzatty is developing a proprietary small language model (SLM) — 3.2B parameters, based on the Llama 3.2 architecture with an 8K Context Window — fine-tuned on Orzatty's internal knowledge base, product documentation, and curated instruction datasets. Once deployed, this model will be registered as the highest-priority provider in the router, replacing external dependencies for the majority of requests while maintaining fallback chains for complex or multimodal tasks.

This transition is intentionally invisible to the end user: the interface, the personality, and the API contract remain identical. The only observable change is improved latency, reduced cost, and full data sovereignty.

4. Identity and Personality

Sylor is not a generic chatbot. Its system prompt defines a consistent character: direct, honest, energetic, and grounded in the Orzatty worldview. It speaks Spanish by default, adapts to the user's language, generates conversation titles automatically, and uses web search proactively rather than hallucinating. This personality layer is version-controlled and deployed as part of the worker — not as a user-configurable setting.

5. Client Architecture

Sylor ships as two clients sharing a single codebase:

Web PWA — deployed on Cloudflare Pages, installable on any device, with offline support via Service Worker.

Native Android App — built with Tauri v2, wrapping the same web frontend in a native shell with OTA update support via the Sylor worker's /updater/check endpoint.

Both clients authenticate via OrzattyAccount, share conversation history through R2, and communicate with the same worker API.

6. Privacy Model

Conversation data is stored in Cloudflare R2 under a per-user, per-conversation key structure. No conversation is accessible without a valid OrzattyAccount JWT. Rate limiting is enforced server-side with passive credit regeneration — users receive 100 credits per session, regenerating at 10 per hour. No conversation content is used for model training without explicit user consent.

7. Ecosystem Integration

Sylor is designed as the conversational interface to the entire O-Suite. Current integrations include OrzattyAccount (authentication and user identity) and OrzattyDrive (file context). Planned integrations include O-Mail (email drafting and summarization), OrzattyCloud (infrastructure queries), and the Rany IDE (code context and agent delegation). As each O-Suite service reaches production readiness, Sylor gains a new tool — without requiring a new app.

8. Monetization Strategy

Sylor is offered under a freemium model:

Free tier — 100 messages per session, multi-provider routing, web search, conversation history, PWA and native app access.

OrzattyOne subscribers — unlimited messages, priority routing to the sovereign model, multimodal support (image analysis), and early access to new tools.

Enterprise licensing is available for organizations requiring private deployment, custom system prompts, or dedicated model capacity.

9. Conclusion

Sylor represents Orzatty's thesis on what AI should be: fast, honest, private, and owned by the ecosystem that builds it. The multi-provider architecture provides resilience today; the sovereign model provides independence tomorrow. The integration roadmap ensures that as the O-Suite matures, Sylor becomes the single interface through which users interact with all of it.

This document is published by ORZ Holdings for informational and research purposes. Technical specifications are subject to change as the product evolves.