Playgrounds
ChatBotKit comes with a set of tools that let you experiment with conversational AI in a safe sandbox before you roll changes into production. We call these tools Playgrounds. Each Playground is designed around a specific kind of task, so you can test prompts, inspect structured data, debug extraction logic, and validate how your AI behaves under different conditions.
Playgrounds are useful when you want fast feedback. Instead of editing a live bot and guessing how changes will behave, you can prototype ideas in isolation, compare outputs, and refine the setup until it works the way you expect. This makes Playgrounds especially valuable for prompt design, data transformation, API troubleshooting, and content extraction.
Conversation
The Conversation Playground is the core Playground in ChatBotKit. It gives you an interactive environment where you can talk to your bot, try different inputs, and observe how the bot responds in real time.
Use it when you want to test prompts, compare models, swap in different datasets or skillsets, and understand how a bot behaves before making changes in production. It is usually the first Playground to use when you are developing a new conversational experience.
Situation
The Situation Playground helps you test how a bot behaves under different conditions. You can simulate a conversation, adjust configuration, and inspect how the model responds when the surrounding context changes.
Use it when you want to troubleshoot a developed conversation, compare response strategies, or identify where a configuration needs improvement before you save those changes elsewhere.
Entity
The Entity Playground helps you inspect how personal and structured information is detected and transformed. It is useful for validating privacy behavior and understanding how sensitive user data is handled.
Use it when you want to test PII detection, verify redaction behavior, or make sure entity processing is doing what you expect before that logic affects production traffic.
Backstory
The Backstory Playground helps you write and refine the personality, tone, and context of your bot. It gives you a focused place to shape how a bot presents itself and how it frames responses.
Use it when you want to create a stronger assistant persona, improve consistency in tone, or experiment with prompt variations before updating a live bot.
Record
The Record Playground helps you work with dataset records and turn them into more useful formats. It is especially helpful when you want to improve raw source material before using it in a dataset.
Use it to convert rough notes into cleaner knowledge entries, FAQs, or structured content that is easier for your AI workflows to consume.
Ability
The Ability Playground gives you a sandbox for creating and testing bot abilities before deployment. It helps you validate how an instruction turns into a tool-call shape and whether the intended behavior is clear enough.
Use it when you are designing custom abilities, tightening instructions, or checking that a generated tool definition matches the action you want the AI to perform.
Image
The Image Playground lets you generate images with different prompts and model settings. It is useful for experimenting with visual output before you incorporate generated images into a workflow or experience.
Use it when you want to compare models, refine prompts, and quickly see how changes affect the generated result.
Widget
The Widget Playground helps you preview and refine the appearance of embedded widgets. It is the right place to test themes, colors, and layout details before you ship a widget to your site.
Use it when you want to align the widget with your brand, compare built-in and custom themes, or validate a visual design before deployment.
Message
The Message Playground helps you create and preview message content across multiple rendering targets. It is useful when you want to verify how the same content appears in widget, Slack, Discord, WhatsApp, Messenger, or Telegram contexts.
Use it to test formatting, compare renderers, and make sure a message looks correct before you use it in a live integration.
API
The API Playground lets you test REST API endpoints directly in the browser. You can send requests, inspect responses, and work through example operations without building a separate client first.
Use it when you want to validate request shapes, debug authentication or payload issues, or learn the API surface through hands-on experimentation.
GraphQL
The GraphQL Playground gives you an interface for testing GraphQL queries and mutations. You can explore the schema, run operations with variables, and inspect response data as you build integrations.
Use it when you want to prototype GraphQL operations quickly, verify returned fields, or debug variables and payload structure before wiring them into an application.
JSONPath
The JSONPath Playground helps you evaluate JSONPath expressions against sample JSON data. It is useful when you need to test extraction logic before embedding it into a broader automation or integration.
Use it to check query syntax, confirm the selected values are correct, and iterate on transformations without having to run a full workflow.
JMESPath
The JMESPath Playground works like the JSONPath Playground, but for JMESPath expressions. It helps you filter, reshape, and extract data from structured JSON documents before those expressions are used elsewhere.
Use it to validate query behavior, compare outputs, and make sure your JSON transformation logic is correct before deployment.
HTML
The HTML Playground lets you test how HTML content is converted into simplified text. It is useful for debugging content extraction and understanding what text your conversational workflows will actually consume.
Use it when you are tuning selectors, skipping specific tags, or validating how rich web content is reduced for AI processing.
The PDF Playground helps you inspect how PDF documents are converted into text. It is useful when you need to validate document extraction quality before relying on PDFs in search, knowledge, or automation workflows.
Use it to test real files, review extracted output, and catch formatting or parsing issues early.