ibm-watsonx-ai-cli 0.1.8

CLI for deploying AI services with Agents to watsonx.ai.

ibm-watsonx-ai-cli

CLI tool that enables developers to deploy AI services to watsonx.ai directly from their local environment.

IBM watsonx.ai CLI for Agents

Repository containing source code for ibm-watsonx-ai-cli package.

Contact info

Maintainers

• Łukasz Ćmielowski - [e-mail]
• Daniel Ryszka – [e-mail]
• Mateusz Szewczyk – [e-mail]
• Mateusz Świtała – [e-mail]

Installation

pip install -U ibm-watsonx-ai-cli

[!WARNING] The ibm_watsonx_ai_cli package requires poetry to be installed. To install Poetry, follow the instructions on the Poetry installation page.

Usage

Run CLI

$ watsonx-ai [OPTIONS] COMMAND [ARGS]...

[OPTIONS]

--version  -v           Print the current CLI version.
--help                  Show this message and exit.

COMMAND

- template              Explore, download and try-out the template.
- service               Work with deployed templates.
- app                   Build & run an UI playground.

TEMPLATE

- list                  List all available templates.
- new                   Creates a selected template in a local env.  [ARGS: name, target]
- invoke                Executes the template code locally with demo data.  [OPTIONAL ARGS: query]

SERVICE

- list                   List all deployed AI services.
- new                    Create & deploy a new ai service from a root template directory. [OPTIONAL ARGS: name]
- get                    Get service details. [OPTIONAL ARGS: deployment_id (deployment_id)]
- delete                 Deletes ai service. [ARGS: deployment_id]
- invoke                 Calls the service by providing the test record. [OPTIONAL ARGS: deployment_id, query]

APP

- list                  List playground app samples.
- new                   Creates a demo playground app for the service.  [OPTIONAL ARGS: name, target_dir]
- run                   Start the playground app.  [OPTIONAL ARGS: target_dir]

---

Template

watsonx-ai template list

List all available templates.

Usage:

$ watsonx-ai template list [OPTIONS]

Options:

• --help: Show this message and exit.

watsonx-ai template new

Creates a selected template in a local env.

Usage:

$ watsonx-ai template new [OPTIONS] name target

Arguments:

• name: Name of the template to download or its index in the list of available templates.
• target: Target directory where template will be saved.

Options:

• --help: Show this message and exit.

watsonx-ai template invoke

Executes the template code locally with demo data.

Usage:

$ watsonx-ai template invoke [OPTIONS] query

Arguments:

• query:[Optional] Input query for template code. If not specified, searches for the cli.options.payload_path file to be specified in config.toml. Content of cli.options.payload_path will be directly send to the AI service and should agree with AI service request schema.

Options:

• --help: Show this message and exit.

Service

watsonx-ai service list

List all deployed AI services.

Usage:

$ watsonx-ai service list [OPTIONS]

Options:

• --help: Show this message and exit.

watsonx-ai service new

Create & deploy a new AI service from a root template directory. After successful deployment creation the deployment_id will be saved automatically in section [deployment] in config.toml.

When running this command a package distribution with implemented agent will be created in folder dist. Warning! If dist already exists, distribution creation is skipped.

Usage:

$ watsonx-ai service new [OPTIONS] name

Arguments:

• name:[Optional] Name for the deployed AI service. If not provided, name for the deployment will be the same as name of current directory.

Options:

• --help: Show this message and exit.

watsonx-ai service get

Get deployed service details.

Usage:

$ watsonx-ai service get [OPTIONS] deployment_id

Arguments:

• deployment_id: Id of the deployed AI service.

Options:

• --help: Show this message and exit.

watsonx-ai service delete

Deletes ai service.

Usage:

$ watsonx-ai service delete [OPTIONS] deployment_id

Arguments:

• deployment_id: Id of the deployed AI service.

Options:

• --help: Show this message and exit.

watsonx-ai service invoke

Calls the service by providing the test record.

Usage:

$ watsonx-ai service invoke [OPTIONS] query

Arguments:

• query:[Optional] Test data to send to deployed AI service. If not specified, searches for the cli.options.payload_path file to be specified in config.toml. Content of cli.options.payload_path will be directly send to the deployed AI service and should agree with AI service request schema.

Options:

• --deployment_id:[Optional] Id of a deployed AI service. If not provided, taken from deployment.deployment_id field from config.toml.
• --help: Show this message and exit.

App

watsonx-ai app list

List playground app samples.

Usage:

$ watsonx-ai app list [OPTIONS]

Options:

• --help: Show this message and exit.

watsonx-ai app new

Creates a demo playground app for the service.

Usage:

$ watsonx-ai app new [OPTIONS] name target_dir

Arguments:

• name:[Optional] The name of the app to use. If not provided, the user will be prompted to choose one.
• target_dir:[Optional] The target folder where the app will be downloaded. If not provided, the user will be prompted to enter one.

Options:

• --help: Show this message and exit.

watsonx-ai app run

Start the playground app.

Usage:

$ watsonx-ai app run [OPTIONS] target_dir

Arguments:

• target_dir:[Optional] The directory to the app.

Options:

• --help: Show this message and exit.
• --dev | -d:[Optional] A flag indicating either app should be deployed in developer mode.

Secrets

Environment Variables for IBM watsonx.ai for IBM Cloud

For security reasons, it's recommended to not hard-code your API key or other secrets directly in your scripts or config.toml. Instead, set it up as an environment variable. Below variables will be used to authenticate to watsonx.ai APIs if analogous ones not provided in configuration file.

import os

WATSONX_URL = os.environ.get("WATSONX_URL", "")
WATSONX_APIKEY = os.environ.get("WATSONX_APIKEY", "")
WATSONX_TOKEN = os.environ.get("WATSONX_TOKEN", "")
WATSONX_SPACE_ID = os.environ.get("WATSONX_SPACE_ID", "")
WATSONX_PROJECT_ID = os.environ.get("WATSONX_PROJECT_ID", "")

Config file (config.toml)

config.toml is an optional file that can be defined in template's directory. Please note, that some templates will require a correctly prepared config.toml since template's content may depends on some of the available configuration options. If a downloaded template has config.toml with placeholders, please read its description carefully and fill in required fields.

Available configuration options

Below are all the sections and options you can have in your config.toml

[cli.options]

[cli.options]
  # If true, cli `invoke` command is trying to use `ai_service.generate_stream` function for local tests, and `ai_service.generate` otherwise.
  # Default: true
  stream = true

  # Path to json file with a complete payload that will be send to proper AI service generate function.
  # Note that, the payload file will be only used when no `query` is provided when running `invoke ` command
  # Default: None
  payload_path = ""

[deployment]

If present, its fields will be used for authentication to watsonx.ai APIs.

[deployment]
  # One of the below is required.
  # To determine your `api_key`, refer to `IBM Cloud console API keys <https://cloud.ibm.com/iam/apikeys>`_.
  watsonx_apikey = "PLACEHOLDER FOR YOUR APIKEY"
  watsonx_token = "PLACEHOLDER FOR YOUR TOKEN"

  # should follow the format: `https://{REGION}.ml.cloud.ibm.com`
  watsonx_url = ""

  # Deployment space id is required to create deployment with AI service content.
  space_id = "PLACEHOLDER FOR YOUR SPACE ID"

  # variable, that is populated with last created deployment_id every time when command `watsonx-ai service new` finish successfully
  deployment_id = ""

[deployment.online.parameters]

This section is recommended for users of IBM watsonx.ai for IBM Cloud and when some additional parameters need to be passed to the AI service function during deployment creation. When creating a deployment, additional parameters can be passed inside the ONLINE.PARAMETERS object to further reference in the content of the AI service function. The exact set of parameters may depend on the template used.

[deployment.online.parameters]

# Example set of parameters that must be passed during deployment creation of particular template to guarantee its correct functioning
  url = ""  # should follow the format: `https://{REGION}.ml.cloud.ibm.com`
  space_id = "PLACEHOLDER FOR YOUR SPACE ID"
  model_id = "mistralai/mistral-large"  # underlying model for inference

[deployment.software_specification]

During deployment creation, the custom software specification with template package distribution is stored for use when running AI service inference. For more details about software specifications you can find in IBM watsonx.ai Runtime Supported frameworks and software specifications documentation and Customizing deployment runtimes documentation.

[deployment.software_specification]
  # Name for derived software specification. If not provided, default one is used that will be build based on the package name: "{pkg_name}-sw-spec"
  name = ""

  # Whether to overwrite (delete existing and create new with the same name) watsonx derived software specification
  # Default: false
  overwrite = false

  # The base software specification used to deploy the AI service. The template dependencies will be installed based on the packages included in the selected base software specification
  # Default: "runtime-24.1-py3.11"
  base_sw_spec = "runtime-24.1-py3.11"

Sample flow

1. List available templates

   watsonx-ai template list
   

2. Create the template in my IDE

   watsonx-ai template new "base/langgraph-react-agent" "langgraph-react-agent"
   

3. Go into template directory

   cd langgraph-react-agent
   cp config.toml.example config.toml
   

4. Customize the template code and config.toml to my needs
5. Test the template code by calling invoke:

   watsonx-ai template invoke "question"
   

6. Create a service

   watsonx-ai service new
   

7. Test the service

   watsonx-ai service invoke "question"
   

8. List available apps

   watsonx-ai app list
   

9. Create an app in my IDE

   watsonx-ai app new "base/nextjs-chat-with-ai-service" "chat-with-ai-service_app"
   

10. Go into app directory and copy your environment variable file.

cd chat-with-ai-service_app
cp template.env .env

11. Update .env file with your credentials
12. Start the playground app.

watsonx-ai app run

Support for Forks or Copies of watsonx-developer-hub

You can now configure the CLI to retrieve agent templates from a forked or mirrored version of the official watsonx-developer-hub repository. To do this, set the following environment variables before invoking any commands:

TEMPLATE_REPO_URL

• Description: URL of the Git repository containing your customized agent templates.
• Default: https://github.com/IBM/watsonx-developer-hub
• Example:

  export TEMPLATE_REPO_URL="https://github.com/your-org/your-forked-repo"
  

TEMPLATE_REPO_BRANCH

• Description: Branch name within the repository where the agent templates reside.
• Default: main
• Example:

  export TEMPLATE_REPO_BRANCH="development"
  

TEMPLATE_REPO_TOKEN

• Description: Personal access token (PAT) or other valid credential used to authenticate against a private repository.
• None: If the target repository is public, this variable is optional.
• Example:

  export TEMPLATE_REPO_TOKEN="<YOUR_TOKEN>"
  

Supported Git Providers

• GitHub
• GitLab

If you configure TEMPLATE_REPO_URL to point to a host other than GitHub or GitLab, the CLI will return an error indicating that the provider is unsupported.