Summary

docs/INTEGRATION-TESTING.md describes flags, install method and model IDs that don't exist (or no longer exist) in the current binary (waza version 0.33.0).

Repro / evidence

$ waza version
waza version 0.33.0

$ waza run --help | grep -E -- '--(executor|log)'
# nothing — those flags don't exist

$ waza models | head -3
MODEL ID           NAME               VISION    CONTEXT WINDOW
──────────────────────────────────────────────────────────────────
auto               Auto               no        -
# claude-sonnet-4-20250514 is not present; sonnet shows up as claude-sonnet-4.5

What the doc currently says (and what's wrong)

In docs/INTEGRATION-TESTING.md:

1. Install:

   pip install waza[copilot]

   waza is a Go binary now (installed via install.sh or go install). There is no pip package.

2. --executor CLI flag:

   waza run eval.yaml --executor copilot-sdk --model claude-sonnet-4-20250514

   --executor was removed; the executor is now set only via config.executor in eval.yaml. Passing --executor errors with unknown flag.

3. --log transcript.json:

   waza run eval.yaml --executor copilot-sdk --log transcript.json

   --log doesn't exist. Closest equivalent is --transcript-dir <dir> (which writes one file per task, not a single --log file) and --session-log.

4. Model ID claude-sonnet-4-20250514 — not in waza models. Trying to use it produces:

   JSON-RPC Error -32603: Request session.create failed with message:
   Model "claude-sonnet-4-20250514" is not available.
   

   The current Sonnet IDs are short: claude-sonnet-4.5, claude-sonnet-4.6.

Suggested fix

• Replace pip install step with the install.sh / go install snippet from docs/GUIDE.md.
• Drop --executor from all examples; document that executor lives in config.executor.
• Replace --log examples with --transcript-dir / --session-log (or remove if those aren't intended for user-facing docs).
• Replace claude-sonnet-4-20250514 with claude-sonnet-4.5 (and recommend waza models to discover IDs).

Happy to send a PR if useful.