HackerNews中文版

我想将用户查询输入到运行小型模型的 Ollama 中，但 Ollama 的分词管道相当复杂。多年来，大多数处理用户输入的软件都暴露出缺陷，有什么方法可以在将用户输入传递给 Ollama 之前对其进行清理吗？或者，我还可以做些什么来避免这个问题？或者，我可以认为 Ollama 的解析本身能够免疫于格式错误的输入所带来的错误吗？

查看原文

I'd like to give user queries to ollama running a small model, but ollama does have a complex tokenization pipeline. Most software that processes user input has turned out to have flaws over the years, is there anything I can do to sanitize users' input before I pass it on to ollama, or anything else I can do to avoid this? Or, can I consider ollama's parsing itself to be immune to errors from malformed user input?

Ask HN：将用户输入传递给 Ollama 安全吗？有什么方法可以对其进行清理吗？