[knipknap]

Summary

Fixes #667.

The Tab Power setting never reached the tab generator. When the
toolpath was built, IntentBuilder._transformer_settings()
assembled the settings dict handed to the transformers' to_spec()
from a fixed list, and the step's power and tab_power were
simply missing from it. TabOpsTransformer.to_spec() therefore
always fell back to its default of tab_power=0.0 and degraded
every tab to the gap (skip) mode, no matter what the user entered in
the Tab Power field.

This forwards both values so a step with Tab Power above 0% now cuts
through the tab at reduced power (tab_power * power), exactly as
the reporter expected. With the test machine configured for
1000 max power, a 70% Tab Power now emits the expected reduced-power
S700-style segment through the tab instead of M5 + G0 skip.

Both FrameStep and ShrinkWrapStep are LaserStep subclasses,
so they pick up the same fix.

Test plan

• pixi run lint passes
• New regression tests in tests/pipeline/test_intent_builder.py:
  • power/tab_power forwarded into the TabsSpec with correctly
    scaled clip data
  • 0% tab power still produces the gap mode
  • changing tab_power invalidates the compute token (cache
    regeneration)
  • end-to-end: executing the pipeline stages produces ops with the
    reduced tab power present (SET_POWER 0.7)
• All four fail on main (verified by reverting the fix) and pass
  with it
• pixi run test tests/pipeline/ rayforge/builtin_addons/rayforge-addon-post/tests/ rayforge/builtin_addons/rayforge-addon-laser/tests/: 714 passed

Docs

The holding-tabs docs only described the gap mode. Added a "Tab
Power" section describing both modes and updated the intro/tab-width
wording, in all 8 languages. The translated setting names use the
exact strings from the app's UI catalogs (e.g. Haltesteg-Leistung,
Puissance des languettes, 定位片功率).

[github-actions]

🤖 Estimated LLM implementation cost

Replayed each of the 1 commits in this pull request and counted the code lines each commit added or rewrote:

Commit	Message	Lines (add/rewrite)	Est. output tokens	Est. output cost
30856de7	fix(tabs): forward the step's power settings to the tab generator	266 / 74	3,400	$0.0510
Total		340	3,400	$0.0510

Added lines by language: Markdown 161, Python 105

Full estimate

Component	Est. tokens	Est. cost
Code emission (340 add/rewrite lines)	3,400	$0.0510
Context input (reading touched files)	27,276	$0.0818
Conversation replay (≈15× output)	51,000	$0.1530
Prompting (pull request description)	355	$0.0011
Lower bound (output only)	3,400	$0.0510
Best estimate	82,031	$0.2869

Cost model: ≈10 tokens per code line at $15/Mtok output and $3/Mtok input (assuming heavy prompt caching); input ≈ 15× output for conversation replay; a review round ≈ 42,000 tokens; a failed CI run ≈ 21,000 tokens; file reads capped at 20 kB each.

Active window from first commit to merge: <1 day.

⚠️ Caveat: the lower bound covers only the tokens for emitting the code. The best estimate adds modeled overhead for exploration, conversation replay, review and debug turns, but real sessions vary by tool and task — treat it as an order-of-magnitude figure. Human engineering time is not included.