# Galileo

> End-to-end platform for generative AI evaluation, observability, and real-time protection that helps teams test, monitor, and guard production AI applications.

Galileo provides an enterprise-focused GenAI Studio for evaluating, observing, and protecting generative AI systems in development and production. It centralizes evaluation metrics, real-time observability, and runtime guardrails so teams can iterate on prompts, models, and retrieval strategies with measurable feedback. Galileo runs low-latency evaluators (Luna models), hosts inference for live monitoring, and offers SDKs and APIs to integrate logging and traces into existing applications.

- **Evaluate** — Rapidly run and compare combinations of prompts, models, embedding params, and chain nodes to find the right configuration; use the platform UI or SDKs to log experiments and golden test sets.
- **Observe** — Monitor live traffic, traces, and sessions with pre-built and custom metrics to detect drift, latency, and accuracy issues in production systems; connect via SDKs or API to stream logs and traces.
- **Protect** — Intercept requests and responses in real time with guardrail policies and threat detection to block harmful outputs or attacks before they reach users.
- **Luna evaluation models** — Use Galileo’s low-latency evaluator models to run automated judgements (<200ms typical) for production monitoring and inexpensive continuous evaluation.
- **SDKs & integrations** — Install Python or TypeScript SDKs, initialize with an API key, and add the log decorator or GalileoLogger to capture prompts, responses, traces, and spans.

Getting started: sign up for the hosted console, install the Python or TypeScript SDK, set GALILEO_API_KEY, and either use the log decorator or GalileoLogger to begin sending traces and running evaluations.

## Features
- AI evaluation workflows for prompts, models, and RAG systems
- Real-time observability of traces, sessions, and metrics
- Runtime protection and guardrail policies
- Prebuilt and custom evaluator metrics (including hallucination detection)
- Luna low-latency evaluation models and hosted inference server
- Python and TypeScript SDKs and a public API
- Auto-tune evaluators with continuous learning (CLHF)

## Integrations
OpenAI (wrapper), NVIDIA NeMo, Python SDK, TypeScript SDK

## Platforms
WEB, API, DEVELOPER_SDK

## Pricing
Freemium — Free tier available with paid upgrades

## Links
- Website: https://galileo.ai
- Documentation: https://v2docs.galileo.ai/
- Repository: https://github.com/rungalileo
- EveryDev.ai: https://www.everydev.ai/tools/galileo