Multi-Step AI Pipeline

Tesseract doesn’t just send your message to a single AI model and return the result. Each response goes through a multi-step pipeline that plans, executes tools, composes block documents, and validates the output.

This means responses are more thoughtful and structured, but may take a few extra seconds compared to a basic chatbot.

What You See While Waiting

Thinking indicator

While the AI is working, you’ll see an animated thinking indicator with dynamic labels that change as the pipeline progresses:

1. Planning — The AI is analyzing your question and deciding how to structure the response.
2. Using tools — The AI is running internal tools (e.g., generating chart data, laying out a flow diagram, or creating an image).
3. Composing — The AI is assembling the final block document.

Planning panel

A collapsible panel appears showing the AI’s plan for responding to your message. Click to expand and see the step-by-step approach the AI has chosen. This is useful for understanding why the AI chose certain block types.

Reasoning panel

A collapsible panel that shows the AI’s chain-of-thought reasoning. Expand it to see the detailed thinking process behind the response.

Streaming & Typewriter Effect

As the response is composed, blocks stream into view progressively:

• Text, callout, and code blocks animate with a typewriter effect — characters appear one by one.
• Tables, charts, and flow diagrams appear fully formed once their data is complete — there’s no partial rendering for visual blocks.
• A final reconciliation event ensures the complete response matches what the AI intended.