global-flags

Global flag placement demo

The loopback fixture reflects the received OpenAI-Project header in a synthetic model ID. The recorder uses matching commands and settings for both binaries. Each command places --project before, between, or after the command words. Existing --transform=id -r options print the synthetic model ID on one line. The recorder preserves actual CLI output and labels each placement for comparison.

Build the fixture:

go build -o /tmp/global-flags-demo-api ./scripts/demos/global-flags

Record the comparison with pinned binaries and commits:

DEMO_API_BINARY=/tmp/global-flags-demo-api \
  bash scripts/demos/global-flags/record.sh \
  /absolute/path/to/before/openai /absolute/path/to/after/openai \
  BEFORE_COMMIT_SHA AFTER_COMMIT_SHA /absolute/path/to/empty-output-directory

Put asciinema, agg, ffmpeg, ffprobe, and python3 on PATH first. The shared recorder requires an empty output directory outside the repository. It runs isolated Bash sessions with a synthetic key and a loopback base URL.

The baseline must accept only the root placement. The candidate must accept all three placements. The recorder checks command exit statuses, extracted output, and actual request headers. It retains these checks in statuses.tsv, requests.jsonl, and validation.txt.

Inspect before.png, after.png, and comparison.gif before sharing the captures. The captures demonstrate terminal replay, not native terminal graphics or other platforms. Keep all binary media outside Git.