# Create Actions with AI If you require an acton that is not yet included in unSkript, you can create your own. The quickest way to go from Zero to Action is to use ChatGPT's generative AI to create an Action. {% embed url="" %} ## Adding your ChatGPT API Key In order to interact with ChatGPT's API, you will need an [API key](https://platform.openai.com/account/api-keys) from OpenAI. Adding your API key: {% tabs %} {% tab title="Docker" %} Your API Key and your organization are entered as parameters when starting your Docker image:
``` docker run -it -p 8888:8888 \ -v $HOME/Awesome-CloudOps-Automation/custom:/data \ -v $HOME/.unskript:/unskript \ -e ACA_AWESOME_MODE=1 \ -e OPENAI_ORGANIZATION_ID= \ -e OPENAI_API_KEY= \ -e OPENAI_MODEL=GPT-4 \ --user root \ docker.io/unskript/awesome-runbooks:1221 ``` If you do not include the OPENAI\_ORGANIZATION\_ID or OPENAI\_API\_KEY, unSkript will run with all the features and tools, except the ChatGPT GenAI tool. The default model is GPT-4. You can find the latest models at [OpenAI](https://platform.openai.com/docs/models/overview). {% endtab %} {% tab title="SAAS/Online" %} In the online version of unSkript, choose More->Settings from the top navigation. Under Integrations, add your organization and API key. {% endtab %} {% endtabs %} ## Creating an Action with ChatGPT In an open RunBook, Click the **+ Add GenAI Action**. This will add a new Action in your RunBook.

A new GenAI Action

In addition to the Configurations and Run Action buttons, there is a new Chat GenAI button. When this button is clicked, a new interface opens in thr right column of the dashboard: ![](/files/6jsckDteYfuY5INoNr38) At the top of this Navigation are two selections: * Action Type: This specifies the Connector to be used in your Action * HealthCheck Action: When checked, the Action will be built as a HealthCheck Action. The output will be a Tuple, first value is Boolean (pass/fail) and the 2nd value is the result. At the bottom of the Navigation is the chat box where we described the Action that is being built. ### Creating an Action Once you have selected the type of action, we can begin generating code. * Type a complete sentence or two of what you would like your Action to do. * Describe any input parameters, whether or not they are required or not, and their initial values. > Example: For an AWS Action, you could say *"list all the S3 buckets that are in region, where region is a required parameter."* Chat GPT will take this input and generate the Python code to complete this action. When the code has been created, open the Configurations tab, and add the configurations needed. In this example, we need to add an AWS credential, and a region. Upon execution, the Action will return a list of Buckets in the region:

Results from a generated Action

### Modifying your Action If you generated Action is not exactly what you need, continue the chat to modify your Action. You may wish to add another input parameter, or slightly change the output of the Action. For example, we could make two modifications to the Action above with the following prompt: > "Return each bucket as a dict with the bucket name and the created date. Add a filter input that returns buckets with that string in the name." This returns an Action with 2 inputs - Region and Filter. On running this Action we can get all Buckets with 'unskript' in the name:

Modified Action

## Testing your Action ChatGPT is not perfect, and yur Action may not do exactly what you expected. This is why testing your Action is very important. {% embed url="" %} Resolving Common issues {% endembed %} Common errors: * AWS: ChatGPT often misses that a query requires a region\_name parameter. If you know that your AWS query is region specific, include region details in your prompt. * Syntax errors: These can often be resolved by feeding the error into the chat. * F Strings: ChatGPT can try some elegant things, but sometimes generates code that reuse the same quote, breaking the f statement: > `print(f' - {dashboard['title']}')` If your generated Action throws an error, often youc an just paste in the error message, and the next version will resolve the issue. ## ChatGPT Tips --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://docs.unskript.com/unskript-documentation/actions/create-actions-with-ai.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.