Autopilot Resource

Note: This capability is now provided as an installable component. See the Components guide for how to install and use it.

Install: kdeps component install autopilot

Usage: run: { component: { name: autopilot, with: { task: "...", context: "...", model: "gpt-4o" } } }

The Autopilot component is kdeps' goal-directed task execution engine. Describe what you want to achieve in plain language; autopilot uses an LLM to plan and execute the task.

Component Inputs

Input	Type	Required	Default	Description
task	string	yes	—	Plain-language description of the goal
context	string	no	—	Additional context or constraints for the task
model	string	no	gpt-4o	LLM model to use for planning and execution

Using the Autopilot Component

run:
  component:
    name: autopilot
    with:
      task: "Research the top 5 open-source LLM frameworks and summarize their strengths"
      context: "Focus on frameworks that support local inference"
      model: "gpt-4o"

Access the result via output('<callerActionId>').

When to use Autopilot

Use autopilot when the exact sequence of steps is not known ahead of time. For well-understood, repeatable pipelines, declarative resources are faster and more predictable.

---

Result Map

Field	Type	Description
success	bool	true if the task completed successfully
result	any	Output of the executed task
model	string	Model used for planning

---

Expression Support

All fields support KDeps expressions:

run:
  component:
    name: autopilot
    with:
      task: "{{ get('user_task') }}"
      context: "{{ get('user_context') }}"
      model: gpt-4o

---

Examples

Research and Summarize

apiVersion: kdeps.io/v1
kind: Resource

metadata:
  actionId: research
  name: Autopilot Researcher

run:
  component:
    name: autopilot
    with:
      task: "Search the web for '{{ get('q') }}' and return a 3-paragraph summary."
      model: "gpt-4o"

Data Analysis

apiVersion: kdeps.io/v1
kind: Resource

metadata:
  actionId: analyze
  name: Autopilot Data Analysis

run:
  component:
    name: autopilot
    with:
      task: |
        Analyze the following data and return key insights as JSON:
        {{ get('data') }}
      context: "Focus on trends and anomalies."
      model: gpt-4o

---

Differences from agent:

	agent:	autopilot:
Workflow	Pre-written, static	LLM-directed at runtime
Steps	Known ahead of time	Determined by model
Predictability	High	Variable
Use case	Production pipelines	Exploratory / dynamic tasks

See Also

• Agency and Multi-Agent - Static agent delegation
• LLM Resource - Direct LLM interaction
• Error Handling - onError: for deterministic fallbacks