Show HN：面向语音 Agent 的开源模拟测试基础设施 Egma

Egma 是一个用于测试语音 Agent 并在生产环境中监控其表现的开源平台。

文档 · Egma Cloud · Discord

• 发布前测试。 通过模拟语音和文本对话以及模拟的工具响应，构建回归测试套件。将测试保存在 Agent 的代码仓库中，并使用 Egma CLI 或编程 Agent 运行测试。
• 监控生产环境。 审查真实对话、工具调用和指标。使用 grader 检查 Agent 的行为，并调查故障。
• 自带密钥。 无论使用 Egma Cloud 还是自行托管，均可使用自己的模型提供商 API 密钥。你直接向提供商付费，Egma 不会对模型使用收取加价费用。
• 自定义测试。 从支持的模型中进行选择，并自定义来电者角色设定。为希望测试的行为设置评分指令和通过阈值。

Egma 支持 LiveKit（JS/TS）和 Retell。如需支持其他平台或功能，请通过 Discord 联系我们。

最快只需五分钟，即可运行你的第一次语音模拟测试。

在编程 Agent 中打开语音 Agent 的代码仓库，并粘贴以下内容：

Set up Egma in this repo. Install its skills with
npx skills add egma-ai/egma, then use them to connect
my voice agent and run my first voice simulation.

如需使用 Egma Cloud，请创建账户，然后使用上面的提示词。若要自行托管，请遵循自托管指南，并在提示词中包含你的实例 URL。请为安装和首次构建预留额外时间。

在自己的机器上运行 Egma

git clone https://github.com/egma-ai/egma.git
cd egma
npm install --global egma-cli
cp .env.example .env
chmod 600 .env

按照自托管指南配置 .env，然后启动 Egma：

打开 localhost:3101 并创建账户。有关连接 Agent、服务器设置和故障排查，请参阅自托管指南。

Egma 目前支持测试和生产环境监控。我们计划增加自动修复功能：Egma 将利用生产环境故障，为你的语音 Agent harness 提出修改建议，进行测试，并创建一个 Pull Request 供你审查。

阅读文档

加入 Discord

创建 GitHub issue

除 ee 文件夹外，本仓库采用 MIT 许可证。详情请参阅 LICENSE。

Egma is an open-source platform for testing voice agents and monitoring them in production.

Docs · Egma Cloud · Discord

• Test before you ship. Build regression suites with simulated voice and text conversations and mocked tool responses. Keep tests in your agent's repository and run them with the Egma CLI or a coding agent.
• Monitor production. Review real conversations, tool calls, and metrics. Use graders to check your agent's behavior and investigate failures.
• Bring your own keys. Use your own API keys for model providers on Egma Cloud or when you self-host. You pay providers directly, with no Egma markup on model usage.
• Customize your tests. Choose from supported models and customize your caller personas. Set grading instructions and pass thresholds for the behavior you want to test.

Egma supports LiveKit (JS/TS) and Retell. Contact us on Discord to request another platform or feature.

Run your first voice simulation in as little as five minutes.

Open your voice agent's repository in your coding agent and paste:

Set up Egma in this repo. Install its skills with
npx skills add egma-ai/egma, then use them to connect
my voice agent and run my first voice simulation.

For Egma Cloud, create an account and use the prompt above. To self-host, follow the self-hosting guide and include your instance's URL in the prompt. Allow extra time for installation and the first build.

Run Egma on your own machine

git clone https://github.com/egma-ai/egma.git
cd egma
npm install --global egma-cli
cp .env.example .env
chmod 600 .env

Configure .env using the self-hosting guide, then start Egma:

Open localhost:3101 and create your account. See the self-hosting guide for connecting your agent, server setup, and troubleshooting.

Egma supports testing and production monitoring today. We plan to add automatic fixes: Egma would use a production failure to propose a change to your voice agent's harness, test it, and open a pull request for you to review.

Read the docs

Join Discord

Open a GitHub issue

This repository is MIT licensed, except for the ee folders. See LICENSE for more details.