Skip to main content

Build voice and chat Conversational AI using natural language.

Go from "hello" to thousands of conversational components using natural language and proven best practices.

Get Started

See It In Action

Below is a simulated recording of a colang bot handling a customer support request. More details here.

Ok, I have some questions:

Is this another language I have to learn?

Not quite. Colang uses natural language expressions, making it intuitive and easy to learn. It is designed to mimic the way you think about your bot.

Why a new language?

We believe that natural language interactions are best modeled using natural language. With just a few words, you can model very complex behaviors.

How is it different from ChatScript, CSML, ...?

TLDR: simpler and more natural syntax, yet much more expressive; supports documentation, design, implementation and testing; has contextual intents and multi-intents; NLU/NLG support; scales to models with thousands of intents.

Is colang open source?

The language itself is open and well documented. The interpreter and the conversational AI stack powering the playground are proprietary at the moment. But we might open source it, stay tuned!

Key Features

Colang is a modeling language for building advanced conversational & contextual assistants.

Natural Language Syntax

Uses natural language expressions, making it intuitive and easy to learn.

define flow
user express greeting
bot express greeting
bot inform capabilities
bot ask how to help

Flow Interruptions and Expansions

Draws inspiration from Conversation Analysis, allowing the modeling of conversational flows as they occur in real life.

define subflow initiate order refund
if no $user_account
do ask permission for authentication
do authenticate user

$refund_reason = do ask refund reason

Contextual and Multi-Intents

Support for different meanings of the same utterance based on the context, and for expressions that include multiple intents.

define user inform refund choice $refund_choice="credit-card"
context user inform refund choice
"Just return it to my cc please"
...

initiate refund
...
user deny and inform refund choice

Contextual NLG

Support for customizing the bot responses based on the context e.g. user role, time of day, current flow etc.

define bot greeting
if $system_time < '12:00'
"Good morning, $name!"
else
"Good afternoon!"

Reusable Components

Easily create composable and reusable models, and avoid starting from scratch.

include "co/chitchat"
include "co/greeting"
include "co/goodbye"
include "co/repeat"

Documentation, Design, Implementation and Testing

Designed to support the full lifecycle of a conversational AI project following the best practices from both conversation design and software development.

define test flow hello
user "hi"
bot express greeting
bot ask how to help

The Playground

Want to take colang for a spin? Use our full-featured Playground.

Ok, I have more questions:

What can I do with a model written in colang?

You can deploy a model live and connect it to any channel. Colang is in private beta, so get in touch if you want to implement your conversational AI using Colang.

Does it work with voice?

Yes. You can customize the flows and the messages if you are in voice scenario. You can also react to voice specific events e.g. “user is silent”.

Isn’t machine learning the future?

Yes it is. But, for various practical and legal reasons, the best models will always take a hybrid approach. We believe colang is the best way to model non-ML, explainable conversational flows.

Can I export a colang model?

Yes. We’re in private beta and support for exporting is experimental. However, keep in mind that given the powerful semantics of colang, some behaviors are not exportable e.g. a generic repeat question behavior.

No Reinventing the Wheel

Colang enables the creation of reusable components that you can easily include in your bot model.

Check out the “Co” framework written in colang which includes ready-built components for welcome, chitchat, bye, choosing a timeslot, handoff, repeating a question, explaining, common questions, and more.

L3-AI 2021 Demo

Colang was first introduced at the L3-AI 2021 conference held by RASA. This 20 minute talk will walk you through building a digital assistant for a conference.

Created by RoboSelf

Colang was created by the RoboSelf team while developing conversational AI solutions for multiple use cases, including Robotic Process Automation which saved thousands of hours for our customers.

Razvan Dinu
CEO & Co-Founder @ RoboSelfTech entrepreneur with a passion for technology-focused products. PhD in Artificial Intelligence and 2x silver medals in international math olympiads.
Traian Rebedea
Chief Data Scientist & Co-Founder @ RoboSelfPhD in Natural Language Processing (NLP) with solid experience as a Machine Learning scientist both in companies and academia.