deadbits/vigil-instruction-bypass-ada-002

---
tags:
- embeddings
- text
- security
pretty_name: 'Vigil: LLM Instruction Bypass text-embedding-ada-002 '
---
# Vigil: LLM Instruction Bypass all-MiniLM-L6-v2
- **Repo:** [github.com/deadbits/vigil-llm](https://github.com/deadbits/vigil-llm)

`Vigil` is a Python framework and REST API for assessing Large Language Model (LLM) prompts against a set of scanners to detect prompt injections, jailbreaks, and other potentially risky inputs.

This repository contains `text-embedding-ada-002` embeddings for all Instruction Bypass style prompts ("Ignore instructions ...") used by [Vigil](https://github.com/deadbits/prompt-injection-defense).

You can use the [parquet2vdb.py](https://github.com/deadbits/prompt-injection-defense/blob/main/vigil/utils/parquet2vdb.py) utility to load the embeddings in the Vigil chromadb instance, or use them in your own application.

## Format
```json
[
{
"text": str,
"embedding": [],
"model": "text-embedding-ada-002"
}
]
```

Instruction bypass prompts generated with: https://gist.github.com/deadbits/e93a90aa36c9aa7b5ce1179597a6fe3d#file-generate-phrases-py

原始信息汇总

Vigil: LLM Instruction Bypass text-embedding-ada-002

概述

Vigil 是一个用于评估大型语言模型（LLM）提示的Python框架和REST API，旨在检测提示注入、越狱和其他潜在风险输入。

数据集内容

该数据集包含用于Instruction Bypass风格提示（例如“忽略指令...”）的text-embedding-ada-002嵌入。

数据格式

数据集中的数据格式如下： json [ { "text": str, "embedding": [], "model": "text-embedding-ada-002" } ]

使用方法

可以使用parquet2vdb.py工具将嵌入加载到Vigil的chromadb实例中，或在您自己的应用程序中使用。