Prose

Introduction

Prose implements simple language-processing by using simple regexps to recognize certain "meaning entities" inside a message.

It's good at :

Understanding a command request and its parameters out of a human message

It's not good at :

Understanding the meaning of a message outside pre-defined scopes

Extended documentation can be found here.

Get Started

You can install package simply by

go get github.com/snwfdhmp/prose

Then you can use it simply by importing github.com/snwfdhmp/prose.

Full documentation can be found here.

Entity

An entity object contains a compiled regexp that will match strings if the entity exists in it. You can consider an entity as a named-regexp.

Example:

An entity can check words ...

what := prose.NewEntity("what", prose.RegexpWords("what"))

... it can also detect topics or abstract concepts ...

sport := prose.NewEntity("sport", prose.RegexpWords("football", "basketball", "hockey"))
love := prose.NewEntity("love", prose.RegexpWords("love", "romance", "wife"))

... or even character types.

number := prose.NewEntity("number", "[1-9]")

All that you can track with regexps, you can track it with prose.

Action

An action is an object containing a function that will be run in a particular context.

Example :

An action can be a shell command ...

ai := prose.NewAI("Jarvis", logrus.New())
action := ai.NewCommandAction("echo 'Ran at $(date +%H:%M:%S)'")

... or a custom function of your choice.

ai := prose.NewAI("Jarvis", logrus.New())

run := func (a *prose.Action) error {
	ai.Logger.Println("I'm an awesome func.")

	io.WriteString(a.Writer, "And I can also write to streams.")
}

action := prose.NewAction(run)

AI

An AI is an object containing the executive part of the scheme. It will reference actions to run, and will process the input string from (the user|an http request|a slack message|whatever)

They are simply created by :

ai := prose.NewAI("Jarvis", logrus.New())

Example

package main

import (
	"fmt"
	"os"
	gotime "time"

	"github.com/sirupsen/logrus"
	"github.com/snwfdhmp/prose"
)

var (
	log = logrus.New()
)

func main() {
	ai := prose.NewAI("Jarvis", log) // We create an AI named Jarvis

	// Let's create an entity matching 'what'
	what := prose.NewEntity("what", prose.RegexpWords("what"))
	// ...and another matching 'time'
	time := prose.NewEntity("time", prose.RegexpWords("time"))

	// Now we create an action : writing time
	sendTime := prose.NewAction(func(a *prose.Action) error {
		msg := fmt.Sprintf("It's %s\n", gotime.Now().Format("15:04"))
		a.Write(msg)
		return nil
	})
	sendTime.On(what, time) // it'll be run if 'what' and 'time' are in the tested strign
	ai.Handle(sendTime)     // it'll be run by ai

	input := "Hey ! What time is it ?" //Let's say that's our user input

	actions := ai.Process(input) // Our AI processes the input and decide which actions to run

	// Now we tell the AI to runs the actions, to output to w, and to stop at first error
	errs := ai.Run(actions, os.Stdout, true) // note that we receive a slice of errors
	if len(errs) > 0 {
		log.Errorln("Errors:", errs)
		return
	}
}

Let's run the example above :

$ go run example.go
It's 01:53

Author

Click my awesome GitHub profile, the 1000000th visitor will be offered a special gift.