[BotshareAI]

Summary

Brings decisions.yaml to the current authority: P-03 rev19.6 (18 October 2026), E-01 rev1.3 and BOM Issue 7.10. It builds on the rev19.4 citation update from the first commit (f20cfea). It supersedes #51.

Authority

Key	Title as recorded
in_force	P-03-rev19.6, 2026-10-18 (was P-03-rev19.4, 2026-10-16)
P-03-rev19.6 (new)	OpenAMRobot 2.0 Decision Addendum rev19.6, 18 October 2026 (the addendum in force; item 22: Gigavac GX11CAC contactors and the item 22 controller I/O)
E-01-rev1.3 (new)	E-01 Safety and Power Distribution Baseline rev1.3, 18 October 2026 (detail for P-03 rev19.6 item 22)
BOM-Issue-7.10 (new)	OpenAMRobot 2.0 Detailed Hardware BOM, Issue 7.10, 18 October 2026 (canonical per P-03 rev19.6)
P-03-rev19.4, P-03-rev19.2, BOM-Issue-7.8	marked as earlier revision or earlier issue

What changed

c8842c1: P-03 rev19.6 item 22

Entry	Change	Source
BOM-ISSUE-IN-FORCE	Issue 7.10 is canonical; Issue 7.8 moves to supersedes. The check accepts 7.10 and flags Issue 7 to 7.9 presented as canonical.	P-03-rev19.6, BOM-Issue-7.10
SAFETY-PROCUREMENT	K1 and K2 are Gigavac GX11CAC (150 A, 24 V coil with internal suppression), normally open, in series. Each has one SPST-NC auxiliary, mechanically linked, in the PNOZ reset loop. BUY-32 is gated on the datasheet checks. LEV100H (with its EDM-STOP gate) moves to supersedes. Separate K1/K2 feedback read by the firmware is not an active value and is recorded as superseded. The two E-stops and the PNOZ s3 are unchanged. New check: LEV100H outside superseded or historical context. The EDM-variant check now also accepts GX11CAC and BUY-32, and its message names the GX11CAC. The automotive-relay message names the GX11CAC.	P-03-rev19.6 item 22; E-stops and PNOZ per P-03-rev19.2 items 5 and 20; E-01-rev1.3
SAFETY-CHAIN-IO (new)	Item 22 controller I/O: Y32 through one ISO1212 module; base E-stop spare NO contact as a diagnostic input; RESET NC input; charger fault input; TPL7407L outputs for the RESET lamp and KCHG. The superseded separate K1/K2 firmware feedback is also recorded here. New check: K1/K2 or contactor feedback to the firmware, STM32, MCU or a GPIO outside superseded or historical context.	P-03-rev19.6 item 22, E-01-rev1.3
MAIN-PROTECTION	Adds F-LIFT 15 A with 2.5 mm2 cable; the rest is unchanged.	P-03-rev19.6 item 22; detail in E-01-rev1.3

Owner of every changed entry: electrical-owner. Approval (verification reviewer): platform-lead.

f20cfea: P-03 rev19.4 citations (first commit, unchanged)

• New source P-03-rev19.4.
• The base camera (D3), LiDAR mount (D4) and imu_link (D5) decisions cite P-03 rev19.4 item 19.
• The Alex decision 2026-10-16 source reads "recorded in P-03 rev19.4 item 19 D3, D4 and D5".

Work package

Register update for P-03 rev19.6 item 22 and BOM Issue 7.10, on top of the rev19.4 citation update. Replaces #51.

Integration Gate

• Updated PR branch from latest main: main SHA 3e26c10fd4e37891b9ad62d89785b40f413f8bca is the second parent of merge commit ae68f925f3874165145973c700c52536c56445eb.
• Previous PR head: c8842c1800d297db8c79fead8e14256b4e5e89b6. The merged tree was tested locally and matches the current PR head tree.
• docs: align decision register with P-03 rev19.4 #51 stays rejected and closed.
• No other open PR touches decisions.yaml.
• Rejected: patterns on the new part numbers themselves (GX11CAC, ISO1212, TPL7407L). They would only restate the values. The new checks flag stale wording only: LEV100H, and K1/K2 feedback to the firmware.

Tests

• test_bom_issue_7_10_in_force replaces the 7.8 test.
• test_automotive_relays_are_not_the_2_0_estop now uses "K1 and K2 are Gigavac GX11CAC contactors." as its negative example; the LEV100H sentence is now a finding.
• New tests, each with a superseded or historical negative case:
  • test_linked_auxiliary_variant_is_the_gx11cac
  • test_lev100h_is_superseded
  • test_no_separate_k1_k2_firmware_feedback
• Revert run: these four tests against the register at f20cfea give 4 failed, 71 passed. On c8842c1 the file gives 75 passed.
• Full suite on the merged tree: 266 tests passed, 0 skipped.
• The local runner needed PyYAML in a temporary virtual environment; no dependency manifest or repository file was changed.

Evidence

Base SHA: 3e26c10fd4e37891b9ad62d89785b40f413f8bca
Head SHA: ae68f925f3874165145973c700c52536c56445eb

python3 tools/check_decisions.py --decisions decisions.yaml --maintainers maintainers.yaml --validate-only
  decisions: 36 loaded, 32 recorded and scanned, 1 open, 3 superseded (citations scanned)   exit 0
python3 tools/check_decisions.py --decisions decisions.yaml --maintainers maintainers.yaml --root . --repository .github
  result: 0 contradiction(s), 0 allowed, scope full checkout   exit 0
python3 -m venv .venv && .venv/bin/python -m pip install 'PyYAML==6.0.2'
VERIFY_PATH="$PWD/.venv/bin:/usr/local/bin:/usr/bin:/bin" bash rollout/verify.sh
  Tests executed: 266 of 266 (skipped 0); PASS: all detected verification stages   exit 0
python3 tools/check_agent_rules.py --canonical agent-rules/SHARED_RULES.md --file AGENTS.md
  PASS AGENTS.md (v2)   exit 0
python3 tools/watchdog.py --root . --repository .github --report-only
  Total: 0 finding(s)
python3 tools/watchdog.py --accepted-words
  regenerated the accepted-words section from decisions.yaml
git diff --check
  PASS

Dependencies

None.

Safety impact

This records safety-chain decisions (contactors, PNOZ reset loop, controller I/O) as register text and text checks only. No E-stop, brake, contactor, watchdog, motor-enable or charge-inhibit logic is authored or changed in any repository. A clean Watchdog result is textual consistency, not safety evidence. The platform lead approves the substance, with two human approvals per the repository ruleset.

STATE.md

No change: this repository has no STATE.md.

Not verified

• P-03 rev19.6 item 22, E-01 rev1.3 revision line, and BOM Issue 7.10 Overview were checked for the F-LIFT attribution and the two source dates. Other source entries were not revalidated in this follow-up.
• The GX11CAC datasheet checks (BUY-32) remain open; no hardware or datasheet checks were performed.
• Consumer repositories were not scanned against this register in this run.

AI disclosure

Claude Code made the original register and test updates described above. Codex verified the named source entries and dates, updated the source citations, synchronized the branch with main, and ran the checks listed above. The commits include the requested Alex Reznichenko sign-off.

Contribution terms

• Every commit is signed off under the DCO.
• I am covered by an accepted Contributor Agreement.
• Third-party material is identified with source and licence.
• No confidential, personal, credential or export-controlled information is included.
• No partner, customer or private person is named; the application is Use_Case_1.

https://claude.ai/code/session_01TPiK6pUmmjNhPceECkKpR2

---

Generated by Claude Code

[panthera-momagdii]

Approved for the citation-only register update.

Reviewed snapshot: openAMRobot/.github, branch audit/p03-rev19-4-register-2026-10-16-v2, head f20cfea264a22a2eadbe9fb58e272c042982f72d, base b7b992ba42e89babe72e42b7907466120d5adc57. Expected outcome: approve the reference/date update only; no technical values, enforcement settings or runtime behavior change.

The complete one-file diff changes the in-force revision/date, adds the rev19.4 source, and updates the three D3/D4/D5 citations. It leaves the IMU values and supersedes history from #50, check patterns, and the rev19.2 item-20 references intact.

I inspected workflow 38956940302 and the actual quality/test log: 256 tests executed, 0 skipped, OK; register validation loaded 35 decisions successfully. This ran on GitHub's PR merge ref 85f142d for the reviewed head/base pair. The separate ROS-verifier smoke and repository-quality jobs succeeded; the reusable repository-quality / quality/test job was skipped, but the dedicated quality/test job did run. License/cla is successful.

Source-provenance limit: I have not independently read the private P-03 rev19.4 document. The revision title, effective date and item attribution are accepted here as Alex's supplied source statement, not an independent verification of that document or the physical geometry. This approval does not change or certify the technical decisions already recorded.

ChatGPT-assisted source/hosted-CI review at Mohamed's request. No local test execution, code edits, merge or settings changes.