Here you'll find many examples built with GraVersal. Some of the flows will help you learn and understand GraVersal. Others are real-world examples. Simply click on one of the flows to load and edit it directly in the builder.
This example demonstrates how the flow derives multiple intents from a single input prompt and cleanly processes them sequentially. Each task knows its position in the queue. This allows linguistic transitions between the intents to be modeled cleanly.
Thanks to the Graph-driven engine, GraVersal always knows which elements in the UI make sense at any given moment. If forms, buttons, or other input fields are no longer needed, GraVersal automatically removes them.
In GraVersal, users can jump back and change their minds at any time. Most answers can be changed afterwards using the icon in the chat history. This flow reacts automatically when a user frequently switches back and forth between two product branches and then offers help. This demo shows a selection of products suitable for online shops, for example.
Imagine this: What would happen if you were in the middle of configuring a product, and—right at that moment—the very last unit was sold? The stock drops to zero. The chatbot interrupts the configuration, tidies up the UI, explains what happened, and offers you alternatives.
A tiny ice cream configurator you can actually talk to. Because a GraVersal flow is a visual graph and executable data at the same time, a reflector node can export its full structure, edges and current state included, straight into a GPT node. The result is a configurator that stays strictly deterministic while an LLM explains its own logic to you. Ask why you cannot have a spoon, and it will tell you: no cup, no spoon.
Pick your dough, your sauce, your toppings. View the summary. Jump back. Change your mind. Break it if you can. Backtracking, follow up questions, and data invalidation all happen on their own, without a single line of code, so the configuration always stays consistent, no matter how often you reshape it.
Tell me something and I'll add it to my to-do list.
This flow takes your input from a prompt, then sends a message and resets itself to wait for input again.
This example shows how you can quickly turn a public API into a chatbot.
Various buttons are generated from your text input. So you can decide for yourself what you want to start with.
You can display multiple questions and answer options at once if they are active simultaneously.
This chat will wait until you type the correct password. Can you find it?
This example shows you how repeatedly changing the answers to a question causes the count nodes in the branches to count their activations. You can use this to address users differently if they have already visited a branch before.
This example generates a button list. You can always revisit the selection and correct it.
This example shows a simple calculator demo. Select the two numbers and see the result. Go back in the chat, change the numbers, and get a new result.
This is a simple contact form. It provides a form and sends an email.
This example shows you how to fill the fields of your form with initial data.
In this example, I'll show you how to send a message if a form field has been filled out incorrectly.
A little cat quiz. Can you find the right answer before time runs out? If you like, you can use the 50% joker.
I built this little text adventure for the dev.to June Solstice Game Jam 2026. It's about Alan Turing, who is reborn as a browser-native AI, logical gates, and bananas.
This flow is part of the onboarding chatbot on the homepage of graversal.io.
This example shows a conversational AND. The process only continues if both questions are answered with "yes".
Sometimes, in a conversation, it's necessary to check if exactly one condition is met. It doesn't matter which one condition is met, as long as it's one.
Der PHQ-9 (Patient Health Questionnaire-9) ist ein wissenschaftlich validiertes Screening-Instrument zur Erfassung depressiver Symptome. Auf dieser Website können Sie den Test vollständig anonym als interaktiven Chat durchführen – alle Daten bleiben ausschließlich in Ihrem Browser und werden beim Schließen automatisch gelöscht. Der Test umfasst neun Fragen zu typischen Depressionssymptomen der letzten zwei Wochen und liefert eine sofortige Auswertung mit Schweregradeinschätzung. Mit einer Sensitivität und Spezifität von jeweils 88% ist der PHQ-9 ein zuverlässiges Werkzeug für die Ersteinschätzung, ersetzt jedoch keine professionelle Diagnostik durch Fachpersonal.
Finde in nur 2-3 Minuten heraus, wie gestresst du wirklich bist. Dieser kostenlose Selbsttest basiert auf der wissenschaftlich validierten Perceived Stress Scale (PSS-10) und misst dein subjektiv wahrgenommenes Stresslevel der letzten 30 Tage.
A board is like a bulletin board. You can display data and text there using Markdown. A board is always visible in the interactive part of the chat. Ideal for real-time data or summaries.
A button emits an ultra short signal. Ideal for doing things once for a short moment. Like counting or adding data once.
GPT nodes can be used to generate dynamic prompts from the flow. These are answered by ChatGpt. You can then use the response in the flow.
This example shows you how to clear the chat. This will delete all messages in the history.
Comment nodes don't do anything. You can theoretically use them to forward signals. But in reality, they only contain details and comments about why certain things in the flow are the way they are.
Use conclusion nodes to make conclusions visually understandable.
The contains node checks whether a specific word is present in an input.
This node helps you determine how often a user has been in a specific branch of the flow.
This node allows you to send data from the flow to HTTP APIs and send data from APIs back to the flow.
Use this node to collect form data in the chat. Connect input nodes to this node to create new input fields.
Use the JSON node to define any kind of data. Use handlebars in the data to make it dynamic.
The jump node embeds a button in the chat that allows you to jump to the stored flow.
Lists can store and collect data. Add data, delete it, or iterate over it.
With the Mail node, you can easily send emails from your flow.
You can use the number node to define numbers or to calculate them.
The once node can only be taken once. Ideal for things that can only happen once.
Use the print node to dynamically create documents from Flow data. Give users the ability to print data directly.
Prompt nodes display a prompt in the chat when they are in context. Use the entered data in the flow.
With questions, you can let users actively make decisions. Each question can have multiple suggestion nodes.
Queue and Task nodes help you process complex tasks sequentially.
A race allows for the simultaneous execution of parallel branches — until one arm wins the race.
The Select Node generates dynamic buttons from a data list.
The start node is your central entry point into every flow.
This flow contains a switch. You can use 'call edges' to turn the switch on or off. When this happens is determined by a simple question. You can turn the switch on or off from the chat.
The text node helps you define strings or build them dynamically.
With a timer, you can start things at a later time.
With Try / Catch, you can capture and redirect events deep within a conversation.
A branch structure is like the branches of a tree. They branch out further and further.
A sequence is simply a chain of nodes that are activated one after the other.
Sometimes you need a structure that repeats itself after it has been completed.
In a star structure, there is a central node from which the signals are sent outward. You can see that the connected nodes are processed in parallel. The respective layers are discovered simultaneously and sorted and executed according to the order of the widest edges.
This example demonstrates the advantages of state merging over crossing edges.
This flow demonstrates a logical AND gate
This flow demonstrates a logical NAND gate
A logical NOT gate inverts the input. If the input is enabled the output is disabled for example.
An OR gate is a logical structure. It describes how different conditions can occur for the same event to be triggered. OR gates are represented in this system using 'may edges'.
This flow demonstrates how a logical XOR gate works.
This flow was designed to test the core engine (RGS) and ensure all functions are working as expected.
This flow is part of the tutorial, which can be loaded within the builder.