Ollama Grid Search
A cross-platform desktop app, written in Rust and React, that iterates over combinations of local Ollama models, prompts and inference parameters and lets a user visually compare and A/B test the results.
Ollama Grid Search automates selecting the best model, prompt or inference parameters for a use case by iterating over their combinations against Ollama endpoints, either on localhost or on a remote server, and letting a user visually inspect the results. It supports A/B testing different prompts across several models at once, multiple iterations per parameter combination, and limited concurrency or fully synchronous calls to avoid overloading a server.
It automatically fetches the models available from a configured Ollama server, keeps an archive of past prompts, logs each experiment, and can optionally output inference parameters and response metadata such as inference time and token counts alongside the generated text.
It is distributed as a desktop application for macOS, Windows and Linux, and can also point at llmman, a local model runner that serves a drop-in-compatible Ollama API.