Otto AI
Otto was my monkey plush, now is my personal AI assistant.
Development
You have to satisfy some dependencies that can be installed via a script based on your platform.
./deps/macos/install.sh
if you run on macOS./deps/pi/install.sh
if you run on a Raspberry Pi
Then:
cp .env.example .env
yarn install
yarn start:dev
If you're gonna work on the client:
cd src-client
yarn start:dev
I/O Drivers
I/O drivers are the the way the AI handles inputs and output.
Every I/O driver must expose these methods:
start
- To start receiving inputsoutput
- To process output
You can configure the I/O driver you want to spawn on your server via config,
using the ioDrivers
keyword.
{
"ioDrivers": ["telegram", "messenger"]
}
You can temporary use a driver without altering your configuration by setting an environment var:
export OTTO_IO_DRIVERS="telegram,test"
There are 4 I/O drivers available at the moment:
- Human: handle input using microphone and speech recognizer and output using a TTS via a speaker
- Telegram: handle I/O for a Telegram bot
- Web: handle I/O via Rest API
IO.Human
This is the main I/O driver.
It uses your microphone to register your voice; once it detects an hot word (example: Hey BOT), it sends the stream through an online speech recognizer and return the speeech.
When you finish to talk, it sends the recognized speech over AI that could return a output speech; it is sent over an online TTS to get an audio file that is played over the speaker.
IO.Telegram
It listens via webhook (or via polling) the chat events of your Telegram bot, send the text over AI that return an output.
The output is used to respond to the user request via Telegram.
IO.Web
It provides a clean Socket.IO interface to interact with the bot.
For every request, you must provide a unique session ID.
Params:
sessionId
: requiredoutputType
: optional, define an additional output type (example:voice
)
I/O Accessories
I/O Accessories are similar to drivers, but don't handle input and output direclty. They can be attached to I/O driver to perform additional things.
You can temporary use a accessory without altering your configuration by setting an environment var:
export OTTO_IO_ACCESSORIES=telegram,test
How to write an action
An action is a responder for an intent that has logic inside.
Your action parameters are:
- The API.AI (Dialogflow) object
- The mongoose session for this request
There is a main difference in actions.
If an action has one return value, it should be a Function or, if you need to do async requests, a Promise / AsyncFunction.
Otherwise, if an action return multiple values over time, it should be a Generator / AsyncGenerator.
Promise/Async Function
export const id = "hello.name";
export async function main({ queryResult }, session) {
let { parameters: p, queryText } = queryResult;
if (p.name == null) throw "Invalid parameters";
return `Hello ${p.name}!`;
}
Generator Function
export const id = "count.to";
export async function* main({ queryResult }, session) {
let { parameters: p, queryText } = queryResult;
for (let i = 1; i < Number(p.to); i++) {
await timeout(1000);
yield String(i);
}
}
Naming
The actions must be placed in the ./src/packages
directory.
If an action name is hello.name
, the final file must be ./src/actions/hello/name.js
;
shorter, if an action name is hello
, the final must be ./src/actions/hello/index.js
.