AI for Wisej.NET
Put intelligence inside the controls your app already uses.
Wisej.AI adds language, vision and reasoning to any Wisej.NET application — no REST plumbing, no vendor SDKs, no rewrites. Drop a SmartHub on a form, attach an adapter to a control, and the control gains a skill.
Free forever · Bring your own model · C# or VB.NET
- 17
- SmartAdapters
- 24
- model endpoints
- 6
- vector stores
- $0
- licence cost
A Wisej.NET screen with three adapters attached: the notes field is completing a sentence, the chart was generated from a plain-language request, and the assistant just called GetAccountRisk — a method of the application.
One endpoint class away from
- OpenAI
- Azure OpenAI
- Anthropic Claude
- Google AI
- Amazon Bedrock
- DeepSeek
- x.AI Grok
- NVIDIA
- GroqCloud
- Together AI
- Hugging Face
- Cerebras
- SambaNova
- Ollama
- LocalAI
- transformers.js
What it is
A thin, dependency-free layer between your controls and a model.
Wisej.AI builds the prompt from the controls and code you already have, sends it to the provider you choose, and parses the response back onto the screen. It runs on .NET Core and .NET Framework, in C# or VB.NET, with no third-party AI client libraries underneath.
No plumbing
You never write a prompt, parse a completion or touch an HTTP client. The adapter knows what its control holds and what the model needs to hear.
Any provider, one line
24 endpoint classes cover the major clouds, OpenAI-compatible services, self-hosted models and inference that runs entirely in the browser.
Your code stays in charge
Models only act through methods you decorate with [SmartTool.Tool]. The surface area of what AI may touch is exactly the surface you define.
.NET to the core
Extender components that behave like every other Wisej.NET component — designer-friendly, strongly typed, at home in Visual Studio.
How it works
Three steps, then it's just your app.
Drop a SmartHub
One extender component on the form. Point it at the endpoint — and therefore the model — of your choice.
var hub = new SmartHub(this) {
Endpoint = new OpenAIEndpoint { Model = "gpt-4o" }
};Attach adapters
Pick the controls that should gain a skill. The adapter builds the prompts, parses the responses and drives the control.
var chat = new SmartChatBoxAdapter {
ChatBox = chatBox1
};Expose your tools
Mark the methods the model may call. The assistant can now act on the running application — on your terms.
[SmartTool.Tool]
[Description("Opens the customer screen.")]
public void OpenCustomer(int id) =>
new CustomerForm(id).Show();The building blocks
Five pieces. One mental model.
SmartAdapters
17 ready-made mediators between a control and a model. Attach one and the control gains a skill — chat, extraction, charting, speech.
Read on → 02SmartEndpoint
One connection object for any model — 24 endpoints across cloud, local and in-browser inference. Swap providers in one line.
Read on → 03SmartHub
The center of an AI screen: binds the endpoint, the adapters and the methods you allow the model to call.
Read on → 04SmartTools
The methods you let the model call. How an assistant acts on your app, your APIs and the UI — scoped to exactly what you expose.
Read on → 05Vectorization
A built-in RAG pipeline: convert, split, embed and store documents, then retrieve by meaning at question time.
Read on →How it fits together
Everything meets at the hub.
Controls are adapted, adapters and tools feed the hub, and the hub reaches whatever model you configured — that's the whole architecture. Watch the data move.
In practice
What teams ship with it.
Every scenario below is one hub, one adapter and a few lines of wiring — not a project.
An ERP copilot
A ChatBox that opens screens, clicks menu items and runs commands on the user's behalf — scoped to exactly the tool methods you expose.
See the adapter →Paste-to-form intake
A customer's email lands in a contact form and name, company, phone and address fill themselves in. Typed objects from untyped text.
See the adapter →Contract & policy Q&A
“How many vacation days do new hires get?” answered from the handbook itself — grounded, cited, no hallucinated policy.
See the adapter →Plain-language reporting
“Top 10 customers by revenue this year” becomes a query against your schema and a DataTable in your grid.
See the adapter →Dictation & meeting capture
Voice notes transcribed in their own language through Whisper-class endpoints, written straight back to the control.
See the adapter →Charts on demand
Hand over raw figures and a sentence; get a populated chart with sensible series, axes and layout.
See the adapter →17 SmartAdapters
From a text box to a copilot.
Each adapter pairs one control with one capability — extraction, transcription, charting, conversation. Mix as many as a screen needs; they share the hub and the endpoint.
Proven in production
Already shipping in real applications.
Wisej.AI isn't a preview. It is running today inside production Wisej.NET applications — powering copilots, document extraction, search and chat for real users and real workloads.
Deployed in production line-of-business apps, not just demos.
SmartAdapters battle-tested across conversation, vision and data tasks.
Built on the same Wisej.NET foundation teams already trust and run.
Pricing
Absolutely free.
Every SmartAdapter, the SmartHub, SmartEndpoint and the full vectorization pipeline are included at no cost. There is no licence fee, no per-seat charge and no usage tier for Wisej.AI itself — you only ever pay your own AI provider for the tokens you use.
- No licence or subscription fee
- All adapters & features included
- Bring your own model & keys
Wisej.AI is part of the Wisej.NET family by Ice Tea Group.