KDeps

Appearance

Data Folder ​

The data folder serves as a dedicated space within your AI agent project to store files and resources that the agent requires during its operations. This structure ensures access to these assets during runtime.

Packaging and Deployment ​

When you package your AI agent, the data folder is automatically included in the resulting .kdeps package. During the Docker image build process, the folder's contents are extracted, guaranteeing that all necessary files are available and ready for use when the agent runs in its containerized environment.

File Organization ​

To maintain clarity and consistency, the data folder is organized based on the AI agent's name and version. For instance:

bash
data/aiagentx/1.0.0/<files_here>

Including Dependencies ​

If your AI agent depends on other AI agents through the Workflows {...} configuration, the data folders for those agents are also packaged. Each agent’s files are stored separately, structured by its name and version:

bash
data/anotheragent/2.1.3/<files_here>
data/aiagentx/1.0.0/<files_here>

Accessing data Files ​

To simplify file access within the resources and the containerized environment, KDeps provides a helper function: data.filepath("agentName/version", "filename"). This function takes two parameters: the agent name with its version and the file name.

For example, to access a file.txt from a dependent AI agent named anotheraiagent:

apl
local fileTxt = "@(data.filepath("anotheraiagent/2.1.3", "file.txt"))"

You can use the retrieved file path with the read function to access its content. For instance, you might want to source a python or shell script stored in the data folder and execute it within each resource.

apl
local fileContent = """
@(read("\(fileTxt)")?.text)
"""