effect-uai-multi-model-compare

Parallel multi-model comparison dispatches a single prompt to multiple providers concurrently and streams tagged outputs as they arrive for side-by-side analysis. Each provider's response is emitted with an explicit member label.

Per-member error isolation generates a standalone error event for the failed provider instead of terminating the entire stream, allowing other providers to continue streaming their outputs.

Yes, you can build ensemble verdicts by dispatching a single prompt to multiple providers concurrently, evaluating the reasoning quality across their streamed tagged outputs, and auditing the responses.

You need a LanguageModelService instance configured for each provider you intend to include in the parallel comparison, which enables the merged stream to emit delta events per member.

Tagged output streaming emits a delta event per member with an explicit provider label, allowing you to audit model behavior and track individual reasoning outputs as they arrive in the merged stream.

Dispatch the same prompt to multiple providers concurrently and compare their tagged streamed outputs to quickly evaluate reasoning quality, audit responses, and make faster cross-model decisions.