promptcmd

promptcmd is a manager and executor for programmable prompts.

translate.prompt

---
input:
  schema:
    to: string, Target language
    summarize?: boolean, Whether to also summarize the text
---

{{#if summarize}}
  Translate the following to {{to}} in a summarized form:
{{else}}
  Translate the following to {{to}}:
{{/if}}

{{stdin}}

generated command

Usage: translate [OPTIONS] --to <to>

Prompt inputs:
      --to <to>       Target language
      --text <text>   Text to translate. will use stdin if not specified
      --summarize     Whether to also summarize the text

Why promptcmd?

Prompt Management

At the simplest form, promptcmd provides a powerful interface for managing your prompts in one place. Prompts can be created, edited, and recalled using the simple promptctl interface.

Programmable Prompts

Prompt files use DotPrompt format. The prompts are essentially templates with parameterized inputs that control the generation of a final prompt. This enables an efficient management of prompts, preventing duplications and reducing work when multiple prompts share parts in common.

Execution & Pipeline Integration

In addition to managing prompts, promptcmd can transform a prompt into a familiar command-line program, where the arguments are used in generating the final prompt. This enables seamless integration of prompts as seemingly native commands into pipelines. Execution further comes with useful features such as load balancing across several LLM providers and caching responses.

Inversion of Control

Inverting control by pre-fetching and embedding necessary information directly into prompts offers advantages over tool-calling for certain use cases. For many it provides provides peace of mind through explicit control over what data the model accesses, eliminating concerns about unintended API calls or resource consumption. It also improves reproducibility: with all context predetermined, the same prompt yields consistent results, enabling reliable testing and debugging.