Data sources are the first step in building a retrieval-augmented generation (RAG) application.

Once you create a data source, the information it contains will automatically be extracted, transformed, and loaded into Griptape Cloud.

You can then add your data to knowledge bases that LLM-powered assistants and structures can access. Learn more

Knowledge bases are collections of data sources. They organize your information and make it accessible for retrieval-augmented generation (RAG).

To get started, first create a knowledge base and select the data source(s) that you want to include.

You can then configure an assistant to chat about your data, or deploy a structure to retrieve it. Learn more

Griptape structures are your code. They are Python agents, pipelines, and workflows that execute your application logic and use LLMs to perform tasks.

To create a structure, you can deploy your own code from a GitHub repository or get started quickly by deploying a structure from our sample repo.

You can run structures directly from the Griptape Cloud web console, via API, or through third party software applications. Learn more

Structure runs are individual executions of a structure.

Once you have created and deployed a structure, you can then run it by providing the necessary input arguments and environment variables.

After your run has completed, you can then view its output, logs, and other details.

Assistants are a special type of Griptape agent structure. Use them to create a chatbot in less than a minute, without writing any code.

To get started, simply create an assistant and optionally add a knowledge base, ruleset, and tools. You can then begin chatting with your assistant by sending a message.

You can also query assistants directly via API, or integrate them into other software applications. Learn more

Assistant runs are individual executions of an assistant.

Once you have created an assistant, you can run it by sending messages. You can then view details for your runs, including sent and received messages and run configurations.

Messages are the inputs that you send to assistants, and the outputs that they send back to you.

Messages can be grouped together into conversation threads.

Threads contain the conversation history of messages sent to and from assistants or structures.

To get started, first create a thread and then select it when chatting with an assistant or running a structure. To start a fresh conversation, simply create and select a new thread.

You can return here any time to view the history of messages.

Rulesets guide the behavior of assistants and structures. Think of them as instructions or guidelines written in natural, conversational language.

To get started, first create a ruleset and then add rules to it. Then, when you create an assistant or structure, you can add your ruleset to it.

Assistants and structures will apply these rules when generating responses to your messages.

Integrations enable you to run Griptape structures and assistants from within other software applications, without needing to write custom integration code.

For example, you can integrate a Griptape assistant into a Slack app to create a Griptape-powered Slackbot.

Secrets are encrypted key-value pairs that you can use to store API keys, authentication tokens, encryption keys, cloud provider access credentials, and database connection details.

To get started, create a secret and then enter a key and value for it.

You can then securely add secrets to structures as environment variables, so that you can use them without exposing sensitive information in your code.

Griptape API keys are used to authenticate requests to Griptape Cloud APIs.

For example, you can use them to search a knowledge base or to run a structure. Create and delete API keys to ensure your interactions are secure.

Note: You don’t need an API key to use the Griptape Cloud web console; simply log in with your Griptape Cloud account.