provider-proxy

[!NOTE]

This package is for people building TensorZero itself, not for people using TensorZero for their applications.

This is a caching MITM proxy, used to cache (notoriously flaky) model provider requests for our E2E tests. This is a caching MITM proxy, used to cache model provider requests during tensorzero e2e test

Usage

The proxy can be started with cargo run. By default, it runs on port 3003 and writes cache entries to ./request_cache. Use cargo run -- --help for more information.

To use this proxy with the e2e tests, set TENSORZERO_E2E_PROXY="http://localhost:3003" when running e2e or batch tests (e.g. TENSORZERO_E2E_PROXY="http://localhost:3003" cargo run-e2e)