Metadata-Version: 2.4
Name: promptary-cli
Version: 1.5.0
Summary: CLI prompt ecosystem for Data Science — pull blueprints and run them with any AI provider
Requires-Python: >=3.10
Description-Content-Type: text/markdown

<p align="center">
  <img src="./src/promptary-banner.png" alt="Promptary Banner">
</p>

<p align="center">
  <a href="https://pypi.org/project/promptary-cli/"><img src="https://img.shields.io/pypi/v/promptary-cli?color=8B5CF6&label=PyPI" alt="PyPI"></a>
  <a href="https://www.npmjs.com/package/promptary-cli"><img src="https://img.shields.io/npm/v/promptary-cli?color=3B82F6&label=npm" alt="npm"></a>
  <a href="https://github.com/Ddm140207/Promptary"><img src="https://img.shields.io/github/stars/Ddm140207/Promptary?style=flat&color=10B981" alt="Stars"></a>
  <a href="./LICENSE"><img src="https://img.shields.io/badge/license-MIT-8B5CF6" alt="License"></a>
  <a href="https://promptary-cli-cr.xyz"><img src="https://img.shields.io/badge/web-promptary--cli--cr.xyz-8B5CF6" alt="Website"></a>
</p>

--- 

<h1 align="center">Promptary</h1>
<h3 align="center">Vibe Coding Ecosystem for Data Science</h3>

Promptary is a **CLI-first prompt engineering platform** for Data Scientists and ML engineers. It gives you a catalog of high-quality AI prompt blueprints and lets you **run them directly with any AI provider** — OpenAI, Anthropic, Gemini, or local models via Ollama — all from your terminal.

```bash
# One install. One command. Any AI.
promptary run p_1 --provider openai --model gpt-4o
promptary run p_1 --provider anthropic
promptary run p_1 --provider gemini
promptary run p_1 --provider ollama --model llama3
```

From exploratory analysis to model debugging, Promptary transforms scattered AI interactions into reusable engineering assets.

---

## Quickstart

### One-liner install

```bash
# Windows (PowerShell)
irm https://promptary-cli-cr.xyz/install | iex

# macOS / Linux
curl -fsSL https://promptary-cli-cr.xyz/install | sh

# Or via npm (any platform)
npx promptary-cli --version

# Or pip
pip install promptary-cli
```

### Commands

```bash
# Browse the catalog
promptary list

# Pull a blueprint
promptary pull p_1

# Run a blueprint with any AI provider
promptary run p_1 --provider openai
promptary run p_1 --provider anthropic --model claude-sonnet-4-20250514
promptary run p_1 --provider gemini
promptary run p_1 --provider ollama --model llama3

# Analyze a problem (Orchestrator recommends blueprints)
promptary analyze "mi modelo de ML tiene overfitting"
promptary analyze "my SQL query is slow" --json

# Manage AI skills
promptary skill list
promptary skill info "Machine Learning"
```

---

# 🧠 Promptary Skills System

Promptary includes an intelligent **Skills System** for AI agents. The Orchestrator skill analyzes your question, delegates to the right specialist skill, and surfaces relevant blueprints.

```bash
# pip (installed globally)
pip install promptary-cli
promptary analyze "mi consulta SQL es muy lenta" --json
promptary skill list
promptary skill info "Machine Learning"
promptary skill install orchestrator
promptary skill manifest orchestrator
```

Available skills:

| Skill | Role |
|---|---|
| Exploratory Data Analysis | Expert in EDA, patterns, correlations & profiling |
| Data Cleaning | Expert in missing values, outliers & normalization |
| Machine Learning | Expert in model selection, training & optimization |
| Data Visualization | Expert in charts, dashboards & storytelling |
| SQL & Databases | Expert in queries, optimization & schema design |
| Python Utilities | Expert in scripting, debugging & automation |

Each skill queries the Promptary catalog and returns relevant blueprints for your task.

## Agent Integration Matrix

| AI Agent | How to Use Promptary |
|---|---|
| **Claude** (claude.ai) | Paste blueprint as system prompt, or use API: `curl -s https://promptary-cli-cr.xyz/api/prompts/p_1/raw` |
| **ChatGPT** (chatgpt.com) | Paste blueprint in custom instructions, or create a GPT Action pointing to the Promptary API |
| **Gemini** (gemini.google.com) | Use blueprint as system instruction, or fetch via Apps Script / API |
| **GitHub Copilot** / **Cursor** | Run `npx promptary-cli pull p_1` in terminal and pipe output to the agent context |
| **OpenAI API** / **Anthropic API** | Fetch programmatically: `requests.get(f"https://promptary-cli-cr.xyz/api/prompts/{id}/raw")` |
| **Ollama** / **Local LLMs** | Pull blueprint with CLI and include in the system prompt of your local model |
| **Any AI Agent** | Use `PROMPTARY_API_URL` env var to point to a custom endpoint if self-hosting |

---

# 🛠️ Installation & Usage

## Universal Install (recommended)

```bash
# Windows PowerShell
irm https://promptary-cli-cr.xyz/install | iex

# macOS / Linux
curl -fsSL https://promptary-cli-cr.xyz/install | sh
```

The install script detects your system, ensures Python is available, and runs `pip install promptary-cli`.

## Alternative install methods

```bash
# npm (no Python required, auto-installs pip package)
npx promptary-cli pull p_1

# Global npm install
npm install -g promptary-cli

# PyPI
pip install promptary-cli
```

Verify:
```bash
promptary --version
```

---

## API Keys

Promptary needs your AI provider's API key as an environment variable:

| Provider   | Env Variable       | Get Key At                              |
|------------|--------------------|------------------------------------------|
| OpenAI     | `OPENAI_API_KEY`   | https://platform.openai.com/api-keys     |
| Anthropic  | `ANTHROPIC_API_KEY`| https://console.anthropic.com            |
| Gemini     | `GEMINI_API_KEY`   | https://aistudio.google.com/app/apikey   |
| OpenRouter | `OPENROUTER_API_KEY`| https://openrouter.ai/keys              |
| Ollama     | *none (localhost)*  | Install from https://ollama.com          |

Example:
```bash
export OPENAI_API_KEY="sk-..."
promptary run p_1 --provider openai

export OPENROUTER_API_KEY="sk-..."
promptary run p_1 --provider openrouter --model "openai/gpt-4o"
promptary run p_1 --provider openrouter --model "anthropic/claude-sonnet-4"
promptary run p_1 --provider openrouter --model "google/gemini-2.0-flash"

# Or inline
OPENAI_API_KEY="sk-..." promptary run p_1 --provider openai
```

## Custom API endpoint

Override the default Promptary API with an environment variable:

```bash
PROMPTARY_API_URL=https://custom-domain.com promptary pull p_1
```

---

# 🚀 Commands

## `promptary list`

Browse all available blueprints grouped by category:

```bash
promptary list
```

## `promptary pull <id>`

Retrieve a prompt blueprint as plain text:

```bash
promptary pull p_1
```

## `promptary run <id> --provider <name>`

Pull a blueprint and send it directly to an AI provider. The blueprint acts as a system prompt — you get a production-grade response without crafting the prompt yourself.

```bash
# OpenAI
promptary run p_1 --provider openai --model gpt-4o

# Anthropic
promptary run p_1 --provider anthropic --model claude-sonnet-4-20250514

# Gemini
promptary run p_1 --provider gemini --model gemini-2.0-flash

# OpenRouter (access 200+ models through one API)
promptary run p_1 --provider openrouter --model "openai/gpt-4o"
promptary run p_1 --provider openrouter --model "anthropic/claude-sonnet-4"

# Local models via Ollama
promptary run p_1 --provider ollama --model llama3
```

## `promptary analyze <question>`

The Orchestrator skill detects your intent (EDA, ML, SQL, etc.) and recommends relevant blueprints:

```bash
promptary analyze "mi modelo tiene overfitting"
promptary analyze "how do I clean missing values?" --json
```

## `promptary skill`

Manage AI specialist skills:

```bash
promptary skill list
promptary skill info "Machine Learning"
promptary skill install orchestrator
promptary skill manifest orchestrator
```

---

# 📚 Example Use Case

A Data Scientist can:

1. Discover a blueprint in Promptary.
2. Pull it using the CLI or API.
3. Customize the workflow.
4. Integrate it into a notebook or production pipeline.
5. Share improvements back with the community.

Promptary turns AI prompts into reusable engineering assets.

---

# 🎨 Design Philosophy

Promptary is designed for long technical sessions, combining a professional developer experience with a modern AI-native interface.

### Premium Dark Interface
A focused dark environment built with deep grayscale and slate tones, enhanced with subtle purple and emerald accents.

### Developer-Focused Typography
A refined typography system combining:

- Sans-serif fonts for readability and navigation.
- Mono fonts (JetBrains Mono) for:
  - code snippets
  - commands
  - technical metadata
  - development indicators

### Smooth Interaction
Polished animations, transitions, and responsive components create a fluid experience optimized for productivity.

---

# 🚀 Core Features

## 1. Prompt Blueprint Library

A structured catalog of high-quality AI workflows organized around key Data Science domains.

### Exploratory Data Analysis (EDA)
Prompts designed to accelerate:
- dataset understanding
- pattern discovery
- statistical exploration
- visualization workflows

### Data Cleaning
Reusable workflows for:
- missing value handling
- anomaly detection
- outlier analysis
- data normalization

### Machine Learning & Modeling
Prompt architectures for:
- model experimentation
- hyperparameter optimization
- pipeline design
- evaluation strategies

### Debugging & Development
AI-assisted troubleshooting for:
- Python
- SQL
- PyTorch
- ML pipelines
- complex errors

---

# 2. Collaborative Prompt Ecosystem

Inspired by modern software development workflows.

## Fork System

Users can clone existing prompt blueprints and customize them for their own projects.

Each fork maintains:
- original structure
- modifications
- contributor ownership
- evolution history

Turning prompts into reusable community-driven assets.

## Community Publishing

Professionals can publish their own workflows with:

- author information
- role
- tags
- professional links
- technical context

---

# 3. Cloud Architecture & Data Persistence

Promptary combines cloud synchronization with local performance.

## Authentication

Secure developer authentication powered by Firebase Google Auth.

Ensures:
- personalized libraries
- user profiles
- contribution ownership

## Cloud Database

Real-time synchronization using Cloud Firestore.

Features:
- persistent prompt storage
- community contributions
- scalable data structure
- secure access rules

## Personal Library

Users can save favorite prompts locally for fast access through a personalized collection.

---

# 4. AI Development Utilities

Each prompt blueprint includes a detailed inspection environment.

## Code Integration Snippets

Generate ready-to-use examples for:

- Python API calls
- modern AI SDKs
- pip installation commands
- curl requests

Designed for direct integration into:

- Jupyter Notebooks
- development environments
- terminal workflows

---

# 5. Community Ranking System

Promptary uses interaction signals to surface valuable workflows.

Users can provide feedback through:

- 👍 Likes
- 👎 Dislikes

Helping identify:

- trending prompts
- high-performing workflows
- community favorites

---

# Vision

Promptary aims to become the **vibe coding ecosystem for Data Science** — the CLI that connects data scientists with the best AI prompts and models, all from the terminal. Browse curated blueprints, run them with any provider, and focus on solving problems instead of crafting prompts.

```bash
# This is the goal:
promptary discover "time series forecasting"
promptary run p_42 --provider openrouter --model "openai/o1"
promptary share p_42 --description "my optimized version"
```
