Skip to main content


user "What is the playground?"
bot "It's a web-based environment for editing and running colang models."

Colang models can be created with any text editor or IDE. However, to run a colang model, you need to publish it to a Conversational AI Environment that supports colang models. The Playground is a web-based environment with advanced capabilities for editing a colang model, and is connected to a backend environment that can run your model.

Key Features

Below is a non-comprehensive list of the key features provided by the Playground:

  • Editing: syntax highlight, syntax errors, global model search, sort intents and response, sort training data;
  • IntelliSense: auto-complete, definition lookup, references lookup, refactoring, unused symbol warnings, undefined symbol warnings;
  • Quick Actions: create user intents and bot responses from flow logic;
  • Training: generate training datasets, train your NLU model, evaluate model against a train/test split;
  • Running: run your model using a state-of-the-art conversational AI stack;
  • Chat Widget: interact instantly with the model using the chat widget;
  • Testing: automatic logging of conversations as tests and running tests;
  • Auto Test: test your bot model against GPT-3 i.e. provide a briefing for GPT-3 and let it talk to your bot.
  • Channel Integrations: integration with various channels e.g. Twilio.

Backend Environment


A Conversational AI Environment is a backend environment that implements the Conversational AI Environment API.

In order to run your model, the Playground uses a state-of-the-art stack of conversational AI components:

  1. Hybrid NLU: a hybrid NLU engine using a grammar-based system and a machine-learning system using RASA.
  2. Flows Engine: a conversational flows engine that implements the full semantics of the colang language.
  3. Graph: an in-memory graph database supporting static data import and connectors to other platforms.
  4. Contextual NLG: natural language generation component with support for context variables, widget templates, and contextual responses.


As you will be using the Playground, you will discover all the powerful features it has to offer. In the following sections, we will start looking multiple colang models, one by one, and highlight various language features, modeling techniques and conversation design best practices. At any point, you can also refer to the Reference Documentation which goes over the language features in a comprehensive way.