94

GitHub - asticode/go-astibob: Bob is a framework to create a distributed AI that...

 6 years ago
source link: https://github.com/asticode/go-astibob
Go to the source link to view the article. You can view the picture content, updated content and better typesetting reading experience. If the link is broken, please click the button below to view the snapshot at that time.

Golang framework to build an AI that can understand and speak back to you, and everything else you want.

WARNING: the code below doesn't handle errors for readability purposes, however you SHOULD!

Demos

Here's a list of AIs built with astibob (if you're using astibob and want your project to be listed here, please submit a PR):

How it works

Overview

  • humans operate the AI through the Web UI
  • the Web UI interacts with the AI through the Index
  • the Index keeps an updated list of all Workers and forwards Web UI messages to Workers and vice versa
  • Workers have one or more Abilities and are usually located on different machines
  • Abilities run simple tasks such as reading an audio input (e.g. a microphone), executing speech-to-text analyses or doing speech-synthesis
  • Abilities can communicate directly between each other even if on different Workers
  • all communication is done via JSON messages exchanged through HTTP or Websocket

  • Why split abilities between several workers?

    Because abilities may need to run on different machines located in different part of the world. The simplest example is wanting to read microphones inputs located in several rooms of your house. Each microphone is an ability whereas each room of your house is a worker.

Install the project

Run the following command:

$ go get -u github.com/asticode/go-astibob/...

I want to see some code

Index

// Create index
i, _ := index.New(index.Options{
    Server: astibob.ServerOptions{
        Addr:     "127.0.0.1:4000",
        Password: "admin",
        Username: "admin",
    },
})

// Make sure to properly close the index
defer i.Close()

// Handle signals
i.HandleSignals()

// Serve
i.Serve()

// Blocking pattern
i.Wait()

Worker

// Create worker
w := worker.New("Worker #1", worker.Options{
    Index: astibob.ServerOptions{
        Addr:     "127.0.0.1:4000",
        Password: "admin",
        Username: "admin",
    },
    Server: astibob.ServerOptions{Addr: "127.0.0.1:4001"},
})

// Make sure to properly close the worker
defer w.Close()

// Create runnables
r1 := pkg1.NewRunnable("Runnable #1")
r2 := pkg2.NewRunnable("Runnable #2")

// Register runnables
w.RegisterRunnables(
	worker.Runnable{
        AutoStart: true,
        Runnable:  r1,
    },
	worker.Runnable{
        Runnable:  r2,
    },
)

// Create listenables
l1 := pkg3.NewListenable(pkg3.ListenableOptions{
	OnEvent1: func(arg1 string) { log.Println(arg1) },
})
l2 := pkg4.NewListenable(pkg4.ListenableOptions{
	OnEvent2: func(arg2 string) { log.Println(arg2) },
})

// Register listenables
w.RegisterListenables(
	worker.Listenable{
        Listenable: l1,
        Runnable:   "Runnable #1",
        Worker:     "Worker #1",
    },
	worker.Listenable{
        Listenable: l2,
        Runnable:   "Runnable #3",
        Worker:     "Worker #2",
    },
)

// Handle an event and send a message to one of the runnables
w.On(astibob.DispatchConditions{
    From: astibob.NewRunnableIdentifier("Runnable #1", "Worker #1"),
    Name: astikit.StrPtr("Event #1"),
}, func(m *astibob.Message) (err error) {
    // Send message
    if err = w.SendMessages("Worker #1", "Runnable #1", pkg2.NewMessage1("Hello world")); err != nil {
        err = errors.Wrap(err, "main: sending message failed")
        return
    }
    return
})

// Handle signals
w.HandleSignals()

// Serve
w.Serve()

// Register to index
w.RegisterToIndex()

// Blocking pattern
w.Wait()

Abilities

The framework comes with a few abilities located in the abilities folder:

Audio input

This ability allows you reading from an audio stream e.g. a microphone.

Dependencies

It's strongly recommended to use PortAudio and its astibob wrapper.

To know which devices are available on the machine run:

$ go run abilities/audio_input/portaudio/cmd/main.go

Runnable and operatable

// Create portaudio
p := portaudio.New()

// Initialize portaudio
p.Initialize()

// Make sure to close portaudio
defer p.Close()

// Create default stream
s, _ := p.NewDefaultStream(portaudio.StreamOptions{
    BitDepth:             32,
    BufferLength:         5000,
    MaxSilenceLevel:      5 * 1e6,
    NumInputChannels:     2,
    SampleRate:           44100,
})

// Create runnable
r := audio_input.NewRunnable("Audio input", s)

// Register runnables
w.RegisterRunnables(worker.Runnable{
    AutoStart: true,
    Runnable:  r,
})

// Register listenables
// This is mandatory for the Web UI to work properly
w.RegisterListenables(worker.Listenable{
    Listenable: r,
    Runnable:   "Audio input",
    Worker:     "Worker #1",
})

Listenable

// Register listenables
w.RegisterListenables(
    worker.Listenable{
        Listenable: audio_input.NewListenable(audio_input.ListenableOptions{
            OnSamples: func(from astibob.Identifier, samples []int, bitDepth, numChannels, sampleRate int, maxSilenceLevel float64) (err error) {
                // TODO Do something with the samples
                return
            },
        }),
        Runnable: "Audio input",
        Worker:   "Worker #1",
    },
)

Speech to Text

This ability allows you to execute speech-to-text analyses.

Dependencies

It's strongly recommended to install DeepSpeech and its astibob wrapper.

I don't want to train a new model

  • create a working directory (for simplicity purposes, we'll assume its absolute path is /path/to/deepspeech)

  • download a client native_client.<your system>.tar.xz" matching your system at the bottom of this page

  • create the /path/to/deepspeech/lib directory and extract the client content inside it

  • create the /path/to/deepspeech/include directory and download deepspeech.h inside it

  • create the /path/to/deepspeech/model/en directory, and download and extract the english model inside it

  • whenever you run a worker that needs deepspeech, make sure to have the following environment variables:

      CGO_CXXFLAGS="-I/path/to/deepspeech/include"
  LIBRARY_PATH=/path/to/deepspeech/lib:$LIBRARY_PATH
  LD_LIBRARY_PATH=/path/to/deepspeech/lib:$LD_LIBRARY_PATH

I want to train a new model

In addition to the steps above:

  • create the /path/to/deepspeech/model/custom directory
  • run git clone https://github.com/mozilla/DeepSpeech inside /path/to/deepspeech
  • install the dependencies

Runnable and Operatable

// Create deepspeech
mp := "/path/to/deepspeech/model/en"
d := deepspeech.New(deepspeech.Options{
    AlphabetPath:   mp + "/alphabet.txt",
    BeamWidth:      1024,
    ClientPath:     "/path/to/deepspeech/DeepSpeech/DeepSpeech.py",
    LMPath:         mp + "/lm.binary",
    LMWeight:       0.75,
    ModelPath:      mp + "/output_graph.pb",
    PrepareDirPath: "/path/to/deepspeech/prepare",
    TrainingArgs: map[string]string{
        "checkpoint_dir":   "/path/to/deepspeech/model/custom/checkpoints",
        "dev_batch_size":   "4",
        "export_dir":       "/path/to/deepspeech/model/custom",
        "noearly_stop":     "",
        "test_batch_size":  "4",
        "train_batch_size": "20",

        // Mozilla values
        "learning_rate": "0.0001",
        "dropout_rate":  "0.15",
        "lm_alpha":      "0.75",
        "lm_beta":       "1.85",
    },
    TriePath:             mp + "/trie",
    ValidWordCountWeight: 1.85,
})

// Make sure to close deepspeech
defer d.Close()

// Initialize deepspeech
d.Init()

// Create runnable
r := speech_to_text.NewRunnable("Speech to Text", d, speech_to_text.RunnableOptions{
    SpeechesDirPath: "/path/to/speech_to_text/speeches",
})

// Initialize runnable
r.Init()

// Make sure to close the runnable
defer r.Close()

// Register runnables
w.RegisterRunnables(worker.Runnable{
    AutoStart: true,
    Runnable:  r,
})

// Send samples
w.SendMessage(worker.MessageOptions{
    Message:  speech_to_text.NewSamplesMessage(
        from,
        samples,
        bitDepth,
        numChannels,
        sampleRate,
        maxSilenceLevel,
    ),
    Runnable: "Speech to Text",
    Worker:   "Worker #3",
})

Listenable

// Register listenables
w.RegisterListenables(
    worker.Listenable{
        Listenable: speech_to_text.NewListenable(speech_to_text.ListenableOptions{
            OnText: func(from astibob.Identifier, text string) (err error) {
                // TODO Do something with the text
                return
            },
        }),
        Runnable: "Speech to Text",
        Worker:   "Worker #3",
    },
)

Text to Speech

This ability allows you to run speech synthesis.

Dependencies

It's strongly recommended to use astibob wrapper.

If you're using Linux it's strongly recommended to use ESpeak.

Runnable

// Create speaker
s := speak.New(speak.Options{})

// Initialize speaker
s.Initialize()

// Make sure to close speaker
defer s.Close()

// Register runnables
w.RegisterRunnables(worker.Runnable{
    AutoStart: true,
    Runnable:  text_to_speech.NewRunnable("Text to Speech", s),
})

// Say something
w.SendMessage(worker.MessageOptions{
    Message:  text_to_speech.NewSayMessage("Hello world"),
    Runnable: "Text to Speech",
    Worker:   "Worker #1",
})

Create your own ability

Creating your own ability is pretty straight-forward: you need to create an object that implements the astibob.Runnable interface. Optionally it can implement the astibob.Operatable interface as well.

If you want other abilities to be able to interact with it you'll need to create another object that implements the astibob.Listenable interface.

I strongly recommend checking out how provided abilities are built and trying to copy them first.

Runnable

The quickest way to implement the astibob.Runnable interface is to add an embedded astibob.BaseRunnable attribute to your object.

You can then use astibob.NewBaseRunnable to initialize it which allows you providing the proper options.

Operatable

The quickest way to implement the astibob.Operatable interface is to add an embedded astibob.BaseOperatable attribute to your object.

You can then use the cmd/operatable command to generate an operatable.go file binding your resources folder containing your static and template files. You can finally add custom routes manually to the astibob.BaseOperatable using the AddRoute method.

Listenable

No shortcut here, you need to create an object that implements the astibob.Listenable interface yourself.

Contribute

If you've created an awesome Ability and you feel it could be of interest to the community, create a PR here.


report

Recommend

About Joyk


Aggregate valuable and interesting links.
Joyk means Joy of geeK