This document summarizes a presentation about chatbots and the Hubot framework. It introduces Hubot, describing it as open source, written in CoffeeScript, running on Node.js, and easy to deploy. Hubot is described as "self-contained, adaptable, listens and hears, and can communicate over HTTP/JSON." While some bots are "smart" through machine learning, Hubot and most bots are "kinda dumb" but have context and utility. The document discusses how Hubot works using regex matching, callbacks, middleware, external scripts, and task running. It also addresses common problems with chatbot packages and frameworks, and describes tools the presenter is developing called Conversation, Dynamic Conversation, and Conditioner
2. Introductions
I’m Tim - product designer
at 4thParty.
Background in front-end
dev… kinda.
4thParty started with an
energy comparison site →
Grouply.co
3. Introductions
Now we’ve made a
platform for apartment
residents, with chat bots!
Access services and
decision support, with a
residential assistant called
Concierge.
4. Introductions
We’re doing UX research,
interviews, beta testing
and development, using
Hubot.
Looking for feedback to
validate (or not) our
framework.
5. How many have made a
bot?
Worked with Hubot?
Other frameworks...
Chatscript, Rivescript,
AIML, Wit.Ai, MBF,
Watson, Botkit, Viv?
Let’s Talk About
Hubot
6. Let’s Talk About
Hubot ● Made by Github
● Written in Coffeescript
● Runs on Node
● Easy to deploy
10. Setting Expectations
Some bots are kinda dumb.
Hubots are in that group.
But they have context and utility.
Alexa / Siri are a bit of both, due to learning from mass use /
massive content.
That’s not really an option.
11. How Hubot
Works
● Listens and Hears
● Regex matching
● Callbacks, response
● Middleware
● Adaptors
● Brain (Redis, or not)
● External scripts
● Requests / task running
12. The Problem
with Packages There’s a package for that.
Sometimes you need things
to work together.
Which are essential, which
are conditional.
14. What We’re Making
Lenny Markus
made Conversation,
it’s great
Eugene Asiedu Gary Chapman
made Dynamic
Conversation, with us
extended support for
Models, with YAML
Tim Kinnane
Stitching it all together,
with Conditioner for Regex
Some other packages: welcome, ignore direct, announcements.
Demo models: https://github.com/timkinnane/playbook-runner/blob/master/models
16. What’s Next
The Playbook library / framework /
just another module?
● Dialogue - expect the expected
● Scene - engage the audience
● Act - choose your adventure
● Outline - start with a plan
● Transcript - always take notes