OnionAI

Chat & Prompts

OnionAI answers questions in natural language. This page explains how to write effective prompts and how context is used across a conversation.

Asking questions

You can ask in plain English (or your preferred language). Be specific about what you want: time ranges, data sources, and the kind of answer (e.g. list, summary, or step-by-step). Examples:

  • “Show me all critical vulnerabilities from NVD affecting our infrastructure in the last 30 days.”
  • “Search Elasticsearch for failed SSH logins in the last 24 hours and summarize by source IP.”
  • “What are the top email threats from SpamAssassin this week?”

Context and follow-ups

Each chat keeps a conversation history. You can refer back to earlier answers (“expand on the second one” or “run that same query for yesterday”) and the AI uses that context. Starting a new chat gives you a clean slate, which helps when switching to a different topic or incident.

Integrations in scope

Answers are based on the integrations you’ve connected and the permissions of your account. If a query touches a source that isn’t connected or isn’t in scope, the AI will say so and suggest next steps.

Tips for better answers

  • Include time ranges when relevant (e.g. “last 7 days”).
  • Name the data source if you have several (e.g. “in Elasticsearch”).
  • Ask for a specific format (e.g. “as a table” or “bullet points”).
  • Use follow-up messages to refine or drill down instead of one very long prompt.

Next steps