Data stays where it belongs
If it can run locally, it does not go to the cloud. StreamCopilot runs entirely on the machine and never uploads captured material — that is the product, not a setting.
01What we do
Scope of work
Not every AI-shaped request deserves an AI solution. Our order of judgement: can a rule handle it, can an existing tool handle it, and only then — do we call a model.
- AI product development — from scenario definition to launch, including model selection, prompt and output-structure design, and failure fallbacks.
- Automation and integration — repetitive browser work, data movement and deployment flows collapsed into re-runnable scripts and services.
- Aggregation systems — multi-source ingestion, de-duplication, incremental updates and retrieval, delivered as something you can filter and subscribe to.
- Deployment and operations — a single ingress, container orchestration, access logs and traffic statistics, so services stay observable and reversible.
02Engineering principles
Four things we hold to
Contracts before implementation
Mismatched field names are the most expensive class of bug because nothing catches them at compile time. Fixing the interface first removes most integration rework.
Every release is reversible
Static sites are files on a read-only mount; services are containers and versioned directories. A bad change should cost minutes to undo, not hours.
Observable or it is not live
Without access logs, traffic statistics and event tracking you cannot tell whether a service is healthy. Observability comes before optimisation.
03Contact
How to reach us
Enquiries
| Channel | Use it for |
|---|---|
| contact@zyaisoft.xyz | Business, project enquiries, problem reports |
| zhisai.zyaisoft.xyz | Zhisai usage questions and data corrections |
| zbzs.zyaisoft.xyz | StreamCopilot trials and training enquiries |
04FAQ
Questions we get asked
Can you do only part of a project?
Yes. Front-end implementation, back-end services, AI integration, automation scripts and deployment can each be taken on separately — there is no requirement to buy the whole chain.
Does our data leave our environment?
Not by default. Products built around local inference, such as StreamCopilot, are designed so captured material never leaves the machine. Where a hosted model is genuinely needed, the data flow is spelled out first and you decide.
Can we change things after delivery?
Yes. Code and configuration live in repositories you control, and changes are scheduled per task or per cycle. Build and deploy procedures are documented on delivery day — no lock-in by design.
Why does every product get its own domain?
It decouples each product lifecycle from the main site, so retiring or rewriting one does not disturb the others, and their search performance stays independent.