Back to Blog

Probefish 🐟 is open source now!

  • Create test suites for your prompts and endpoints (for example your AI Assistant)
  • Automated static validations (regex, JSON schema, response time)
  • AI-validation rules, use LLM-as-judge for quality scoring
  • Compare prompt execution results across GPT-4, Claude, and Gemini side-by-side. Select model fits better
  • Track regression history over time
  • Self-hosted, your API keys stay with you

If you're building with LLMs and want confidence your prompts actually work - check it out.

Can be connected to Gitlab-Ci and provides webhooks out-of-box.