Run Deterministic Benchmarks

Run the deterministic smoke tools first. They are useful for comparing changes in a controlled local environment, but their timings are measurements from that machine and run, not universal performance promises.

uv run miniproto-bench-acceptance --mode smoke --json .tmp/bench/runtime-acceptance.json
uv run miniproto-bench-imports --runs 3
uv run miniproto-bench-media-scheduler
uv run miniproto-bench-native-fallback-crypto --mode smoke --json .tmp/bench/native-fallback.json
uv run miniproto-bench-runtime-paths
uv run miniproto-bench-session-crypto --mode smoke --json .tmp/bench/session-crypto.json

The command names are installed project scripts and are documented with their focused inputs in Development Commands. Keep JSON output under .tmp/ or another task-owned output directory.

Credentialed media and matrix benchmarks are intentionally separate. They depend on Telegram permissions, account type, data-center placement, rate limits, network conditions, and live credentials. Do not enable a live benchmark by accident or use one noisy run as a deterministic regression gate.