AI Chat overview
Conversation-driven workflow — tell the agent what you want, and it acts on your document directly: creating widgets, editing configs, reading documents.
AI Chat is VizChat's primary interaction surface. Unlike single-turn Q&A chatbots, VizChat's agent can operate your document directly — create widgets, modify configurations, read files, invoke tools — turning a conversation into an actual workflow change.
How it works
You describe intent in natural language
↓
Agent reads context (the current document, the selected widget, referenced files)
↓
Agent plans tool calls (create widgets / modify config / read files / search)
↓
Executes on canvas → awaits your confirmation or auto-applies
↓
You follow up; the agent remembers the threadFour core capabilities
| Capability | Example instruction |
|---|---|
| Create widgets | "Make a line chart for revenue over the last 12 months" |
| Edit config | "Change the X axis of this bar chart to months" |
| Read docs | "Look at my uploaded Q3 report, pull out 3 highlights" |
| Web search | "Find this week's AI industry news and make an infographic" |
Relationship to other features
- Canvas (Canvas): the stage the agent operates on;
- Slash commands (Slash commands): shortcuts for high-frequency actions;
- Storyboard (Storyboard): conversation output goes straight into presentation;
- BYOK (BYOK): use your own API key, skip platform credits entirely.
Consumption
By default, using platform-hosted models costs credits. If you bring your own OpenAI / Anthropic / DeepSeek / etc. API key (BYOK), the cost falls on your provider account and does not consume platform credits.