Prompt Store

Prompt Store is a secure, collaborative CLI and Rust library for managing, deploying, and orchestrating AI prompts using workspaces and Git-based workflows. Built with AES-256-GCM encryption, it provides a robust vault for your prompt templates, allowing you to organize, version, and execute them across various LLM backends.

With a fluent, builder-style API and a powerful interactive CLI, you can create simple prompts or complex multi-step chains with advanced logic like conditional execution, parallel processing, and error fallbacks.

Key Features

• Secure Vault: All prompts are encrypted at rest with AES-256-GCM, optionally protected by a master password that never leaves your machine.

• Workspaces & Git-based Deployment ("PromptOps"):

  • Isolate prompts into distinct workspaces (e.g., default for personal, seo-pack for a deployed set).
  • Deploy "Prompt Packs" directly from public or private Git repositories using prompt-store deploy.
  • Keep packs up-to-date with a simple prompt-store update command.
  • Create and share team prompts with password-protected bundles (prompts.bundle) using prompt-store pack export.

• Schema Definition: Optionally define input/output JSON schemas for your prompts to ensure reliable, structured data from LLMs.

• Advanced Orchestration:

  • Build multi-step, multi-provider prompt chains programmatically.
  • Execute steps in parallel for improved performance.
  • Use conditional steps (step_if) for dynamic workflow logic.
  • Define fallbacks (on_error) for robust error handling.

• Powerful CLI: Manage every aspect of your prompts with intuitive commands, including an interactive REPL mode (prompt-store interactive).

• Fluent Library API: A developer-friendly, chainable API for integrating prompt execution directly into your Rust applications.

• Version History: Automatically creates backups on edits, allowing you to view history and revert to previous versions.

Installation

CLI Tool

You can install the command-line tool directly from the repository:

cargo install prompt-store

Ensure ~/.cargo/bin is in your shell's PATH.

Library

To use prompt-store as a library in your Rust project, add this to your Cargo.toml:

[dependencies]
prompt-store = "0.1.1"

CLI Usage

The prompt-store CLI provides a comprehensive set of commands to manage your prompt vault.

• prompt-store new: Interactively create a new standalone prompt.
• prompt-store chain new: Interactively create a new multi-step prompt chain.
• prompt-store list: Display all prompts and chains.
  • prompt-store list --tag rust --tag api: Filter standalone prompts by tags.
• prompt-store get <id>: Display the content of a specific prompt.
• prompt-store run <id> --var key=value: Render a prompt with variables.
• prompt-store stats: Show statistics about your vault.
• prompt-store interactive: Start an interactive REPL session.

For a full list of commands, run prompt-store --help.

Library Usage

The library offers a powerful, fluent API for prompt execution and chaining, designed to be clean and intuitive.

Basic Prompt Execution

use prompt_store::{PromptStore, RunOutput};
use llm::builder::{LLMBuilder, LLMBackend};

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    // Initialize the store once. For non-interactive use, see `PromptStore::with_password()`.
    let store = PromptStore::init()?;

    let openai_llm = LLMBuilder::new()
        .api_key(std::env::var("OPENAI_API_KEY")?)
        .model("gpt-4o-mini")
        .build()?;

    // Run a prompt stored with the ID 'welcome-prompt'
    let output = store.prompt("welcome-prompt")
        .vars([("name", "Alice")])
        .backend(openai_llm.as_ref())
        .run()
        .await?;

    if let RunOutput::Prompt(text) = output {
        println!("LLM Output: {}", text);
    }

    Ok(())
}

Advanced Chain Execution

Dynamically build and run a chain with parallel steps, conditional logic, and error fallbacks.

use prompt_store::{PromptStore, RunOutput};
use llm::chain::LLMRegistry;
use llm::builder::LLMBuilder;

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let store = PromptStore::init()?;

    // Setup multiple providers
    let mut registry = LLMRegistry::new();
    registry.insert("openai", LLMBuilder::new().api_key(std::env::var("OPENAI_API_KEY")?).build()?);
    registry.insert("failing_provider", llm::NoopLLM::new()); // For fallback demo

    let user_query = "Explain Rust's ownership model.";

    let outputs = store
        .chain(&registry)
        // 1. Sequential step
        .step("topic", "extract-topic-prompt")
            .with_provider("openai")
        
        // 2. Parallel steps
        .parallel(|group| {
            group
                .step("summary", "summarizer-prompt")
                .step("keywords", "keyword-extractor-prompt")
                    .with_provider("failing_provider") // This step will fail
        })
        .with_provider("openai") // Default provider for the parallel group
        
        // 3. Fallback for the failed "keywords" step
        .on_error_stored("basic-keyword-extractor")
            .with_provider("openai")
        
        // 4. Conditional step
        .step_if("tweet", "generate-tweet-prompt", |ctx| {
            ctx.get("summary").map_or(false, |s| s.len() > 50)
        })
            .with_provider("openai")
        
        .vars([("query", user_query)])
        .run()
        .await?;

    if let RunOutput::Chain(map) = outputs {
        println!("Summary: {}", map.get("summary").unwrap_or(&"N/A".into()));
        println!("Keywords (from fallback): {}", map.get("keywords").unwrap_or(&"N/A".into()));
        if let Some(tweet) = map.get("tweet") {
            println!("Generated Tweet: {}", tweet);
        }
    }

    Ok(())
}

Examples

The examples/ directory contains functional code demonstrating various features: