kdeps

Appearance

Reusing and Remixing AI Agents ​

Kdeps makes it easy to integrate existing AI agents, enabling reusing and remixing of pre-built AI agents into your own AI workflows.

Installing a Kdeps AI Agent ​

To begin, install the AI agent using the kdeps install command:

bash
kdeps install conveyour_counting_ai-1.2.5.kdeps

Once installed, the agent is ready to be registered in your workflow.

Registering AI Agents in Your Workflow ​

After installation, you must register the AI agent in your workflow.pkl file. External workflows are referenced using @ followed by the agent name.

For example, to include the latest version of the AI agent in your workflow:

apl
workflows {
  "@conveyour_counting_ai"
}

This will include to all the resources provided by the conveyour_counting_ai agent. If you prefer a specific version of the agent, include the :version specifier, like this:

apl
workflows {
  "@conveyour_counting_ai:1.2.5"
}

Utilizing an External AI Agent ​

Once the agent is registered in your workflow.pkl file, you can include it in the requires block of your resources:

apl
requires {
  "@conveyour_counting_ai/countImageLLM:1.2.5"
  "@conveyour_counting_ai/sortImageItemsLLM:1.2.5"
}

After specifying the required resources, you can use a function or retrieve output through file. Here’s an example:

apl
local sortedItemsJsonPath = "@(llm.file("@conveyour_counting_ai/sortImageItemsLLM:1.2.5"))"
local sortedItemsJsonData = "@(read?("\(sortedItemsJsonPath)")?.text)"

local report = new Mapping {
  ["fruit_count"] = "@(document.JSONParser(sortedItemsJsonData)?.fruit_count_integer)"
  ["vegetable_count"] = "@(document.JSONParser(sortedItemsJsonData)?.vegetable_count_integer)"
  ["stock_analysis_report"] = "@(document.JSONParser(stockAnalysisLLMReporter)?.report_markdown)"
}