roboai-cli 1.1.2

A command line tool to create, manage and deploy Rasa chatbots on the Robo.AI platform.

ROBO.AI Bot Runtime manager CLI tool

This tool allows anyone to create, train, deploy, monitor and manage a Rasa based bot on the ROBO.AI platform.
Check our CHANGELOG for the latest changes.

Tutorials:

• The ROBO.AI platform - creating bots and API keys for deployment
• Creating and deploying a Rasa chatbot on the ROBO.AI platform

How to install

Requirements

• Python 3.7
• Pip and/or anaconda

You can create a virtual environment using conda:

conda create -n roboai-cli python=3.7
conda activate roboai-cli

Installing the ROBO.AI tool

Assuming you are already in your virtual environment with Python 3.7, you can install the tool with the following command:

pip install roboai-cli

After installing the library you should be able to execute the robo-bot command in your terminal.

Usage

The command line tool is available through the following terminal command:

roboai

When you execute it in a terminal you should see an output with a list of commands supported by the tool.

I.e:

user@host:~$ roboai
 ____   ___  ____   ___           _    ___ 
|  _ \ / _ \| __ ) / _ \         / \  |_ _|
| |_) | | | |  _ \| | | |       / _ \  | | 
|  _ <| |_| | |_) | |_| |  _   / ___ \ | | 
|_| \_\\___/|____/ \___/  (_) /_/   \_\___|
Bot Management Tool             robo-ai.com

Usage: roboai [OPTIONS] COMMAND [ARGS]...

  roboai 1.1.1

Options:
  --version  Show the version and exit.
  --help     Show this message and exit.

Commands:
  clean        Clean the last package
  connect      Connect a local bot to a ROBO.AI server bot instance.
  data         Utility command to split, export and import data.
  deploy       Deploy the current bot into the ROBO.AI platform.
  diff         Check for structural differences between languages for the...
  environment  Define the ROBO.AI platform API endpoint to use.
  interactive  Run in interactive learning mode where you can provide...
  login        Initialize a new session using a ROBO.AI API key.
  logout       Close the current session in the ROBO.AI platform.
  logs         Display selected bot runtime logs.
  package      Package the required bot and make it ready for deployment.
  remove       Remove a deployed bot from the ROBO.AI platform.
  run          Start the action server.
  seed         Create a new ROBO.AI project seedling, including folder...
  shell        Start a shell to interact with the required bot.
  start        Start a bot deployed on the ROBO.AI platform.
  status       Display the bot status.
  stories      Generate stories for a Rasa bot.
  stop         Stop a bot running in the ROBO.AI platform.
  test         Test Rasa models for the required bots.
  train        Train Rasa models for the required bots.

Each of the listed commands provides you a functionality to deal with your bots, each one has a description, and a help option, so you can see what options and arguments are available.

You can invoke each of the tool commands by following the pattern:

roboai <command> [command arguments or options]

i.e.:

roboai login --api-key=my-apy-key

You can check the supported options and arguments for every command by following the pattern:

roboai <command> --help

i.e.:

user@host:~$ roboai login --help
 ____   ___  ____   ___           _    ___ 
|  _ \ / _ \| __ ) / _ \         / \  |_ _|
| |_) | | | |  _ \| | | |       / _ \  | | 
|  _ <| |_| | |_) | |_| |  _   / ___ \ | | 
|_| \_\\___/|____/ \___/  (_) /_/   \_\___|
Bot Management Tool             robo-ai.com

Usage: roboai login [OPTIONS]

  Initialize a new session using a ROBO.AI API key.

Options:
  --api-key TEXT  The ROBO.AI platform API key.
  --help          Show this message and exit.

Using roboai-cli to create and maintain a bot

Generating an initial structure

The ROBO.AI tool provides you with a set of commands useful to create, train, interact and test a bot before its deployment.

To create a bot you can use the seed command:

roboai seed [language-codes] [--path <path> --language-detection --chit-chat --coref-resolution]

i.e.:

roboai seed en de --path bot/ --language-detection --chit-chat --coref-resolution

The first argument of the seed command is the language-codes which indicate the languages the bot will be built upon. If no language-codes are passed, only an english sub-directory (en) will be created. The optional parameters are referring to features you may want to add to the bot.
This command behaves like rasa init but it'll generate a dedicated structure where you can have multi-language bots related with the same domain. Below there's an example of a bot generated with this command.

.
├── actions
│   └── action_parlai_fallback.py
├── custom
│   ├── components
│   │   └── spacy_nlp
│   │       ├── spacy_nlp_neuralcoref.py
│   │       └── spacy_tokenizer_neuralcoref.py
│   └── policies
│       └── language_detection
│           ├── lang_change_policy.py
│           └── lid.176.ftz
├── languages
|   ├── de
|   │   ├── config.yml
|   │   ├── data
|   │   │   ├── lookup_tables
|   │   │   └── nlu.md
|   │   └── domain.yml
|   ├── en
|   │   ├── config.yml
|   │   ├── data
|   │   │   ├── lookup_tables
|   │   │   └── nlu.md
|   │   └── domain.yml
|   └── stories.md
├── credentials.yml
├── endpoints.yml
└── __init__.py

Generating stories for a bot

After defining intents and actions for a bot you need to combine these in stories. This command allows you to generate the most basic interactions in your Rasa bot. Note: Manual checks will be needed to implement more complex stories but basic ping-pong dialogues should be covered with this feature. Usage:

roboai stories [language-codes] [--check-covered-intents]

If no language-code is passed, roboai-cli will assume you're working in a single-language bot (and thus the default Rasa structure). The option --check-covered-intents will go through your stories file and check if the intents you have defined in the domain file are being covered in the dialogues. This command is more useful when you're deep in the development of your bot.

Checking for differences in a bot

After making all the necessary changes to your bots, you want to make sure that all bots (languages) are coherent between each other (i.e. the same stories.md file will work for the nlu.md and domain.yml files configured for the different languages.) To know whether your bot is achieving this, you can use the diff command.

roboai diff [language-codes] [--path <path>]

It will check for structural differences between the domain.yml and stories.md files for the same multi-language bot. If no language codes are passed, then it'll pair all the languages found and check for differences between them.

Splitting the nlu

In case you want to split your nlu data, you can use the data command for that.
Simply run roboai data split nlu [language-code] and a new folder called train_test_split will be generated within the bot directory.
When training and testing the bot you can then pass these files as arguments.

Training a bot

You're now in a position to train the bot. To do so you only need to run the train command just as you would do in Rasa.

roboai train [language-codes] [--nlu --core --augmentation <value> --dev-config <path to config file> --force --debug --training-data-path <path-to-training-data-file>]

In case you want to pass a specific training data file you can use the train command in the following way:

roboai train en --training-data-path train_test_split/training_data.md

It will train the bot and store the model in the language sub-directory. If no language codes are passed, all bots will be trained.

Note: The augmentation and force options do not work in the case of NLU training.

Interacting with a bot

To interact with the bot, you can use the shell command. Before running it, you need to execute the run actions command.

roboai run actions [--debug]

After doing so, you can execute the shell command.

roboai shell [language-code] [--nlu] [--debug] 

You need to specify what language (bot) you want to interact with - you can only interact with one bot at the time.

Testing a bot

Testing a bot is also probably in your pipeline. And this is possible with the test command.

roboai test [language-code] [--cross-validation --folds <nr-of-folds> --tet-data-path <path-to-testing-data-file>]

In case you want to pass a specific testing data file you can use the test command in the following way:

roboai test --training-data-path train_test_split/test_data.md

It'll test the bot with the conversation_tests.md file you have stored in your tests folder.
The results will be stored in the language sub-directory. Besides Rasa's default results, roboai-cli also produces an excel file with a confusion list of mistmatched intents.

Interactive learning

If you want to use Rasa's interactive learning mode you can do this by using the interactive command.

roboai interactive [language-code]

It'll launch an interactive session where you can provide feedback to the bot. At the end don't forget to adjust the paths to where the new files should be saved.

By now you're probably ready to deploy your bot...

Using roboai-cli to deploy a bot

Setting the target endpoint

Before doing any operation you must indicate to the tool in what environment you're working in, for that you have the environment command:

The tool provides you with a default production environment in the ROBO.AI platform. You can activate it by running:

roboai environment activate production

You can also create new environments by executing:

roboai environment create <environment name> --base-url <base-url> [--username <username> --password <password>]

The base-url refers to the environment URL and you can optionally pass a username and password if your environment requires them.

i.e.:

roboai environment create development --base-url https://robo-core.my-robo-server.com --username m2m --password GgvJrZSCXger

After creating an environment, do not forget to activate it if you want to use it. To know which environment is activated you can simply run:

roboai environment which

It's possible to check what environments are available in your configuration file by running:

roboai environment list

You can also remove environments by executing:

roboai environment remove <environment name>

Logging in

Once you have the desired environment activated, you need to login into the account you'd like to use by using an API key.

1. Log-in into your ROBO.AI administration and generate an API key (do not forget to enable it).
2. Execute the login command and enter the API key.

i.e.:

roboai login --api-key=my-api-key

Or if you don't want to enter the api key in your command, you can enter it interactively by only executing:

roboai login

Initializing a bot

In order to manage a bot runtime, it needs to be initialized so the tool will know what bot this runtime refers to. If you already have the Rasa bot initialized, just execute the following command:

roboai connect [language-code] --target-dir <path to rasa bot files>

i.e.:

roboai connect [language-code] --target-dir /path/to/rasa/bot

First it'll ask you to pick an existing bot (if it does not exist, you must create it before executing this step). After doing it, it'll generate a new file called robo-manifest.json which contains meta-information about the deployment and the target bot.
Note: if no language-code is provided, it's assumed that you're working with the default Rasa structure.

Deploying a bot

When your bot is ready for deployment, you must train it first and ensure you're in the bot root directory. You can then execute:

roboai deploy [language-code] --model <path to model file>

It'll package your bot files and upload them to the ROBO.AI platform, starting a new deployment. This step may take some time.
If you want you can pass the path to the model you want to deploy. If no model path is passed then the most recent one will be picked up. Note: if no language-code is provided, it's assumed that you're working with the default Rasa structure.

Checking a bot status

If you want to check your bot status, just run the following command from the same directory as of your robo-manifest.json

roboai status [language-code]

Note: if no language-code is provided, it's assumed that you're working with the default Rasa structure.

Removing a bot

If you need to remove a bot, execute the following command from the bot root directory:

roboai remove [language-code]

Note: if no language-code is provided, it's assumed that you're working with the default Rasa structure.

Checking a deployed bot logs

Sometimes it's useful to have a look into the logs, for that you need to execute:

roboai logs [language-code]

It'll show you the latest 1000 lines from that rasa bot logs.
Note: if no language-code is provided, it's assumed that you're working with the default Rasa structure.

Using roboai-cli to export and import data

Export data

If you require to export your bot's data you may use the data command for that end.
You can run

roboai data export [nlu/responses/all] --input-path <bot-root-dir> --output-path <path-to-where-you-want-to-save-the-file>

If you opt to export only the nlu or the responses, an excel file will be generated with this content. If you wish to export both, use the 'all' option and an excel file with both the nlu and responses will be generated.

Import data

To import the data back from excel to markdown/yaml, the data command is what you're looking for.
You can run

roboai data import [nlu/responses/all] --input-path <path-where-your-file-is-saved> --output-path <path-to-where-you-want-to-save-the-file>

This will generate markdown and yaml files containing the nlu and responses content, respectively.

Code Style

We use Google Style Python Docstrings.

Releases

We follow Semantic Versioning 2.0.0 standards.

Given a version number MAJOR.MINOR.PATCH, increment the:

• MAJOR version when you make incompatible API changes,
• MINOR version when you add functionality in a backwards compatible manner, and
• PATCH version when you make backwards compatible bug fixes.
  Additional labels for pre-release and build metadata are available as extensions to the MAJOR.MINOR.PATCH format.