✨ prep

Transform messy prompts into precise, powerful instructions for AI assistants.

Features • Installation • Quick Start • Providers • Configuration • Documentation

---

Why Prep?

Ever type a vague prompt like "make a website" and get mediocre results? Prep transforms your casual thoughts into precise, actionable prompts that get better results from any AI assistant.

$ prep "make a website for my cat photos"

# Output:
Build a responsive single-page website to showcase cat photos with the following 
requirements: a masonry-style photo grid layout, lightbox functionality for 
enlarged viewing, lazy loading for performance, mobile-first responsive design, 
and a simple navigation header. Use HTML5, CSS3 with Flexbox/Grid, and vanilla 
JavaScript. Include alt text placeholders for accessibility.

Features

🔄 Multi-Provider Support — Local Ollama, Ollama Cloud, OpenAI, Anthropic
🎨 Beautiful Terminal UI — Colored output, progress spinners, interactive prompts
⚙️ Persistent Configuration — Set your defaults once, use everywhere
📋 Clipboard Integration — Copy refined prompts with --copy
📜 History Tracking — SQLite-backed history with search
📝 Prompt Templates — 10 built-in templates for common tasks
🐚 Shell Completions — Bash, Zsh, Fish, PowerShell

Installation

From Source (Recommended)

# Clone the repository
git clone https://github.com/caseybarajas/prep.git
cd prep

# Build and install
cargo install --path .

Prerequisites

• Rust 1.70+ — Install Rust
• AI Provider — One of:
  • Ollama running locally, OR
  • API key for Ollama Cloud, OpenAI, or Anthropic

Quick Start

Basic Usage

# Simple prompt refinement
prep "write a python script to parse json"

# Pipe from file or other commands
cat rough_idea.txt | prep

# Use with a specific provider
prep --provider openai "build a REST API"

With Ollama (Local)

# Start Ollama
ollama serve

# Pull a model
ollama pull llama3.2

# Use prep
prep "create a dockerfile"

With Cloud Providers

# Set your API key
export OPENAI_API_KEY="sk-..."
# OR
export ANTHROPIC_API_KEY="sk-ant-..."
# OR
export OLLAMA_API_KEY="..."

# Use the cloud provider
prep --provider openai "design a database schema"
prep --provider anthropic "write unit tests"
prep --provider ollama-cloud "explain microservices"

Providers

Provider	Flag	Model Default	API Key Variable
Ollama (Local)	--provider ollama	llama3.2	Not required
Ollama Cloud	--provider ollama-cloud	llama3.2	OLLAMA_API_KEY
OpenAI	--provider openai	gpt-4o	OPENAI_API_KEY
Anthropic	--provider anthropic	claude-3-5-sonnet	ANTHROPIC_API_KEY

Configuration

Initialize Config File

prep config init

Configuration is stored at ~/.config/prep/config.toml:

[default]
provider = "ollama"           # Default provider
model = "llama3.2"            # Default model
output_format = "text"        # text, json, or markdown
copy_to_clipboard = false     # Auto-copy results

[providers.ollama-local]
endpoint = "http://localhost:11434"

[providers.openai]
endpoint = "https://api.openai.com/v1"
model = "gpt-4o"

[providers.anthropic]
endpoint = "https://api.anthropic.com/v1"
model = "claude-3-5-sonnet"

[ui]
color = true                  # Colored output
spinner = true                # Show progress spinners

[history]
enabled = true                # Track refinement history
max_entries = 1000            # Max history entries

Config Commands

prep config show              # Display current config
prep config set KEY VALUE     # Set a value
prep config get KEY           # Get a value
prep config path              # Show config file location

Documentation

Command Reference

prep [OPTIONS] [PROMPT]

Arguments:
  [PROMPT]  Raw prompt to refine (reads from stdin if not provided)

Options:
  -p, --provider <NAME>    AI provider (ollama, ollama-cloud, openai, anthropic)
  -m, --model <NAME>       Model to use (overrides config)
  -o, --output <FORMAT>    Output format: text, json, markdown
  -C, --copy               Copy result to clipboard
  -t, --template <NAME>    Use a prompt template
      --context <FILE>     Include file as additional context
      --dry-run            Preview without calling API
  -v, --verbose            Show diagnostic output
      --no-color           Disable colored output
      --no-history         Don't save to history
  -h, --help               Print help
  -V, --version            Print version

Subcommands:
  config       Manage configuration
  history      View and manage refinement history
  templates    Work with prompt templates
  completions  Generate shell completions

Templates

Built-in templates optimize prompts for specific tasks:

Template	Description
code	Code generation requests
debug	Debugging assistance
docs	Documentation writing
explain	Concept explanations
review	Code review requests
refactor	Refactoring tasks
test	Test writing
api	API design
security	Security analysis
architecture	System architecture

# List all templates
prep templates list

# Use a template
prep --template code "parse CSV files in rust"
prep --template debug "my function returns null"

History

# List recent refinements
prep history list

# Show specific entry
prep history show 42

# Search history
prep history search "python"

# Clear all history
prep history clear

Shell Completions

# Bash
prep completions bash > ~/.local/share/bash-completion/completions/prep

# Zsh (add ~/.zfunc to fpath in .zshrc first)
prep completions zsh > ~/.zfunc/_prep

# Fish
prep completions fish > ~/.config/fish/completions/prep.fish

# PowerShell
prep completions powershell >> $PROFILE

Examples

Basic Refinement

$ prep "make api"

# Output:
Design and implement a RESTful API with the following specifications: 
define the resource endpoints (GET, POST, PUT, DELETE), implement proper 
HTTP status codes, add request validation, include error handling with 
meaningful messages, and document the endpoints. Specify the programming 
language and framework to use.

With Context File

$ prep --context src/main.rs "add error handling"

# The context file content is sent along with your prompt
# for more relevant refinements

Different Output Formats

# JSON output (for scripting)
prep --output json "write tests" | jq .refined_prompt

# Markdown (for documentation)
prep --output markdown "explain oauth"

Pipeline Integration

# Refine and pass to another tool
prep "write a greeting function" | pbcopy

# Use with AI coding assistants
prep "add caching to this function" --context utils.py | claude

Security

• API keys are never stored in config files
• Keys are loaded from environment variables only
• History is stored locally in SQLite
• See SECURITY.md for details

Contributing

Contributions are welcome! Please read CONTRIBUTING.md for guidelines.

License

MIT License — see LICENSE for details.

---

⬆ Back to Top

Made with ❤️ and Rust