Revalvo

Revalvo is a local-first workbench for prompt engineering and model evaluation. Instead of switching between chat UIs or running one-off tests, you write a single prompt and dispatch it to multiple models at once — OpenAI, Anthropic, Groq, Ollama, and more — all in parallel. The results appear side by side with latency, token count, and cost per call. You can score outputs automatically using built-in evaluators that check for structure, semantic similarity, hallucination, and even safety. Every run is saved as an immutable snapshot, so you can track how a prompt evolves over time.

What makes Revalvo different is its focus on versioning and reproducibility. Prompts are treated like code: you can diff any two versions, roll back to an earlier one, and export a winning prompt as a ready-to-paste API snippet in Python, TypeScript, cURL, or Go. The batch runner lets you test a prompt against a full dataset — upload a CSV or let AI generate test rows — and get a structured report with pass rates, model rankings, and per-case pass/fail. The tool runs entirely in your browser using IndexedDB; there's no Revalvo account, no server-side storage, and your API keys never leave your machine. You can even run fully offline with local models via Ollama or LM Studio.

Revalvo is built for anyone who needs to make informed decisions about which model or prompt to use in production. Prompt engineers can iterate quickly with real-time scoring and version history. AI researchers can run controlled experiments across models and datasets. Developers shipping AI features can batch-test prompts against edge cases before deployment. The tool is free to use — you only pay for the API calls you make through your own provider keys.