ИИ-агенты
deepeval
Unit tests for LLMs — evaluate before you ship.
★ 18,7 тыс. звёздApache-2.0
Смотреть руководство →guardrails — проект с открытым исходным кодом из категории «ИИ-агенты» от guardrails-ai (guardrails-ai/guardrails на GitHub). У него 7 496 звёзд на GitHub и 715 форков, он распространяется по лицензии Apache-2.0 и последний раз обновлялся на GitHub 6 октября 2026.
Put guardrails on LLMs: validate outputs before they reach users.
Write down the outcome you want in one sentence.
Clone it locally and check the README.
Get dependencies and tooling in place per the docs.
Follow the minimal example, then adapt it to your needs.
Check LLM output against your rules.
Express constraints declaratively.
Re-ask when the model strays.
Enforce schemas, not hopes.
Catch disallowed content pre-send.
Fits the standard LLM stack.
Download the source and move into the project folder.
git clone https://github.com/guardrails-ai/guardrails.git
cd guardrailsIsolate dependencies, then install requirements.
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txtLaunch using the entry point shown in the README.
python -m guardrails # check README for the exact commandIt is an open-source project at guardrails-ai/guardrails with 715 forks. Check the contributors tab and recent commits to judge how active maintenance is.
Yes — Apache 2.0 allows commercial use and modification. Keep the license and notice files, and note it includes a patent grant.
You define rules for LLM outputs; the framework validates and retries or corrects when outputs break them.
Validation adds some latency — the trade-off for safer outputs. Measure it for your use case.
Clone the repo and follow the README quickstart — the docs at https://www.guardrailsai.com/docs walk through setup.
Unit tests for LLMs — evaluate before you ship.
Terminal security that stops malicious prompts, URLs, and payloads before they run.
Detect prompt injections and jailbreaks before they hit your LLM.
Programmable guardrails to keep LLM systems safe and on rails.
Fast, best-practice evaluations for AI model outputs.
Scan AI models for serialization attacks before you load them.
Другие проекты из категории «ИИ-агенты» с наибольшим числом звёзд в каталоге LibreAI — реалистичный список альтернатив, если guardrails вам не подходит.
A personal AI that really does things — on any OS, with your data staying yours.
An agentic skills framework & software development methodology that works.
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
The agent that grows with you — a Python framework for adaptive AI agents.
DeepSeek's plugin-based agent harness where everything is a plugin.
Соседние проекты из категории «ИИ-агенты» в каталоге, ближайшие по числу звёзд к guardrails — продолжайте изучать полку по обе стороны от него.
Reef — infrastructure for continually self-improving agents: weight-training and harness-evolving deployments.
OpenTelemetry-native observability for LLM apps.
Honcho — Plastic Labs' memory library for stateful agents that learn about their users over time.
Programmable guardrails to keep LLM systems safe and on rails.
A batteries-included browser API so agents can automate the web.
Voyager — the open-ended embodied Minecraft agent that writes its own skills and keeps learning for life.
Бесплатно и открыто — забирайте исходный код, ставьте звезду репозиторию и начинайте создавать.