Subprocess

The subprocess backend runs your prompts by spawning a command.

This is useful when you already have an LLM runner script, want to call a local model, or need full control over auth and tooling.

Config

backend: subprocess

subprocess:
  command: python3
  args:
    - ./run_llm.py
  useStdin: false

How the prompt is passed

• useStdin: false (default): promptmd appends the prompt as the last command-line argument
• useStdin: true: promptmd writes the prompt to stdin

Streaming

Whatever your command writes to stdout is streamed to the terminal in real-time.

Structured output

If your prompt file defines an output: schema, promptmd will try to parse your subprocess stdout as JSON.

That means your command should print a single JSON object to stdout.

{"title":"...","priority":"high"}