fix(python-cli): allow qualitative goal creation without metric options

[devloper961-maker]

Problem

omi --json goal create "Build stronger relationships" exits 2 with a missing --target usage error before any HTTP request is sent.

The developer API explicitly supports qualitative goals by omitting all metric fields — CreateGoalRequest in backend/routers/developer.py declares goal_type, target_value, current_value, min_value, max_value as Optional (all default None). The CLI is stricter than the API it wraps, so qualitative goals are impossible to create from the CLI.

Fixes #13085

Root cause

create_goal in sdks/python-cli/omi_cli/commands/goal.py declared --target as required (typer.Option(...)) and always sent every metric field in the request body.

Implementation

• Make --target, --type, --current, --min, --max optional (Optional[float] / Optional[GoalType], default None).
• If no metric option is provided, send only {"title": ...} — a qualitative goal, exactly what the API accepts.
• If any metric option is provided, preserve the historical defaults (type=scale, current=0, min=0, max=10) so existing scripts behave identically.
• Validate metric combinations (cubic review follow-up):
  • --unit without any metric option is rejected (exit 1) — the backend only persists unit on metric-backed goals, so it would be silently dropped on a qualitative goal.
  • Metric options without --target are rejected (exit 1) — preserving the historical contract that every metric goal has a target, instead of fabricating a target-less scale goal.
• --unit continues to be attached whenever provided alongside a metric goal.

Testing

• New: test_goal_create_qualitative_omits_metrics — qualitative request body contains only the title.
• New: test_goal_create_metric_defaults_preserved — --target 2 alone still produces the historical scale defaults (backward compat).
• New: test_goal_create_unit_without_metrics_rejected — --unit alone exits 1 with guidance.
• New: test_goal_create_partial_metrics_rejected — --current 5 without --target exits 1 with guidance.
• New: test_goal_create_unit_attached_to_metric_goal — --target 2 --unit liters sends unit through.
• Full Python CLI suite: 134 passed, 1 skipped.
• The 4 failures in tests/test_openapi_contract.py on older runs are pre-existing on current main and unrelated to this change.
• black clean on both touched files.

Compatibility

No breaking change for valid invocations: every command that worked before produces the same request body. The only behavior changes are that omitting metric options now succeeds (qualitative goal) instead of exiting with a usage error, and previously broken partial-metric invocations now fail fast with a clear client-side message instead of creating a malformed goal server-side.

[cubic-dev-ai]

All reported issues were addressed across 2 files

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>

[devloper961-maker]

Both review findings are fixed in daf249e:

1. --unit without metric flags — now rejected client-side with exit 1 ("--unit requires a metric goal") instead of silently sending a unit the backend drops for qualitative goals. The --unit help text documents the requirement.

2. Metric options without --target — now rejected client-side with exit 1 ("--target is required when using metric options"). This preserves the historical contract where --target was mandatory for every metric goal and prevents the fabricated target-less scale goal (_safe_float(None, 0.0) in backend/database/goals.py would have coerced target_value=None to 0.0). This also matches the scope stated in Python CLI rejects qualitative goal creation supported by the API #13085 ("rejecting other metric options without a target").

New tests: test_goal_create_unit_without_metrics_rejected, test_goal_create_partial_metrics_rejected, test_goal_create_unit_attached_to_metric_goal. Full CLI suite: 134 passed; black clean.

[kodjima33]

Real bug: CLI requires --target even though the backend's CreateGoalRequest supports qualitative goals with no metric fields. Confirmed --target still required on main. Scoped fix (130 loc) with 5 new tests covering both qualitative and backward-compat metric paths. Confidence 5/5.