Visual answers
Not just text: diagrams that become knowledge.
The chat does not answer with text alone. It renders Mermaid diagrams and Syncfusion charts — safely and declaratively rather than as free HTML artefacts. What appears in the chat looks the same in the knowledge base, can be exported as an image and saved with one click as a searchable, lasting note.
A wall of text explains some things poorly
Most LLM chats deliver a wall of text. For an explanation that is often enough; for a process, an architecture or a row of figures it rarely is — there a diagram says more than three paragraphs.
The common way to produce visuals is free HTML or JavaScript artefacts. That is flexible, but it runs foreign code from a model or knowledge-base answer in the browser — an open door for security problems. And even a good graphic is lost once you close the chat.
Safely declarative, consistent, lasting
CompanyWizard describes visual content declaratively, not as executable code: Mermaid for processes and schemas, a lean chart notation for figures. Every output also passes a sanitisation step that strips active parts such as scripts while keeping embedded images. If the model describes a chart invalidly, the output degrades cleanly to a readable code block instead of breaking the answer.
Chat and knowledge base use the same render component — what you see in the chat looks identical once saved. Via "Save to KB" the whole answer, diagrams included, becomes a searchable note (vault picker, role-based write path), and on export it lands as an image in your corporate layout in PDF or Word.
From answer to knowledge
-
Visualise
Mermaid for processes, charts for figures — autonomously when the question calls for it, with no need to ask first.
-
Render safely
Declarative instead of executable, output sanitised; an invalid chart degrades to a readable code block.
-
Show consistently
One shared render component for chat and knowledge base — no surprise on reopening.
-
Save to KB
The whole answer including diagrams becomes a note via a vault picker and the role-based write path — re-rendered and searchable on reopening.
-
Export
As an image to PDF or Word in your corporate layout.
Free HTML/JS artefacts or declarative visuals?
| Criterion | Free HTML/JS artefacts | Declarative visuals (CompanyWizard) |
|---|---|---|
| Security | Foreign code runs in the browser | No executable code, output sanitised |
| Design freedom | Arbitrary, interactive too | Common types: flowchart, bar, line, pie |
| Chat vs. knowledge base | Differs by renderer | One shared component, identical |
| On error | Can break the answer | Degrades cleanly to a readable code block |
| Export | Inconsistent | As an image to PDF/Word in corporate layout |
| Permanence | Fleeting in the chat log | One click to a searchable KB note |
| Searchability (RAG) | Not provided for | Saved answer is full-text searchable |
What this means for you
- Diagrams and charts right in the chat — Mermaid for processes, Syncfusion charts for figures.
- Safe by design: declarative instead of executable code, output sanitised.
- One rendering for chat and knowledge base — no surprise when you reopen a saved answer.
- From image to knowledge: one click to a searchable KB note, exportable to PDF and Word.
Honestly: the limits
-
Declarative does not mean anything goes.
The chart notation covers the common types (bar, line, pie), not every conceivable special visualisation. For complex, interactive or pixel-perfect graphics a specialised tool serves you better — that is the price of the safe path.
-
No free interactive artefacts.
Unlike assistants that generate arbitrary HTML/JS surfaces, CompanyWizard stays deliberately declarative. For interactive mini-applications in the chat this is not the right place.
-
Quality depends on the model.
A diagram or chart is built from what the LLM describes. If the description is incomplete or invalid, you see a readable code block instead of the graphic — that prevents errors but occasionally calls for a follow-up or a corrected instruction.
Sounds right for your organisation?
In a 30-minute demo we show the concept on your concrete use case — no obligation, with concrete suggestions at the end.