promptfoo

Overview

Promptfoo is a CLI and library for evaluating and red-teaming LLM-based applications. It supports automated evaluations of prompts and models, red teaming and vulnerability scanning, model comparisons, CI/CD checks, code scanning, and sharing results with a team. It is open source and MIT licensed.

Key Features

• Automated evaluations for prompts, models, and RAG pipelines
• Red teaming and vulnerability scanning for LLM applications
• Side-by-side comparison of model providers including OpenAI, Anthropic, Azure, Bedrock, and Ollama
• CI/CD integration and code scanning for LLM-related security and compliance issues in pull requests
• Local execution for LLM evals so prompts do not leave your machine
• Declarative configs with command line and library usage

Use Cases

• Test and improve prompts, agents, and RAGs with automated evaluations
• Secure AI applications with red teaming and vulnerability scanning
• Compare model performance across providers
• Add automated LLM checks to CI/CD and review pull requests for security and compliance issues

Getting Started

• Install with npm install -g promptfoo, or use brew install promptfoo, pip install promptfoo, or npx promptfoo@latest.
• Run promptfoo init --example getting-started.
• Set an API key for most LLM providers as an environment variable, for example export OPENAI_API_KEY=sk-abc123.
• Run promptfoo eval and promptfoo view in the getting-started directory.

Deployment & Requirements

• Can be installed via npm, Homebrew, pip, or run with npx.
• Most LLM providers require an API key set as an environment variable.
• Supports on-premise or cloud integration.

Before You Adopt

• License: MIT. Review its terms before using, modifying, or distributing the project.
• Most LLM providers require an API key configured as an environment variable.
• LLM evaluations can run locally and are described as private.