Port CT002/CT003 emulator to ESPHome as external component

[tomquist]

Ports the AstraMeter CT002/CT003 grid-meter emulator, load balancer, and cross-phase filter pipeline from Python to C++ as an ESPHome external component, so the full stack can run directly on an ESP32 — no separate Home Assistant add-on or server required. Grid power comes from any ESPHome sensor source (Modbus, M-Bus, Tasmota, Shelly, Envoy, …).

What ships

A single ESPHome external component, ct002, with two optional nested sub-blocks under the same ct002: key:

external_components:
  - source: github://tomquist/astrameter@main
    components: [ct002]

ct002:
  id: ct002_main
  power_sensor_l1: grid_l1          # + l2/l3 for three-phase
  ct_type: HME-4                    # HME-4 = CT002, HME-3 = CT003
  filters:                          # cross-phase pipeline (all optional)
    hampel:    { window: 7, n_sigma: 3.0, min_threshold: 50 }
    smoothing: { alpha: 0.3, max_step: 200 }
    deadband:  { deadband: 10 }
    pid:       { kp: 0.5, ki: 0.1, output_max: 800, mode: bias }
  balancer:    { fair_distribution: true, balance_gain: 0.2, ... }
  saturation:  { enabled: true, alpha: 0.15, ... }

  # Optional sub-block: HA Device Discovery + Marstek-app MQTT responder.
  # Requires an mqtt: block in the YAML.
  mqtt_insights:
    base_topic: astrameter
    ha_discovery: true
    marstek_mqtt_enabled: true
    marstek_mqtt_interval: 300s

  # Optional sub-block: register a managed device with the Marstek cloud
  # on first boot, persist the MAC, feed it back to ct002. Requires an
  # http_request: block.
  marstek_registration:
    base_url: https://eu.hamedata.com
    mailbox: you@example.com
    password: !secret marstek_password

See esphome.example.yaml for an annotated template.

Layout

Everything lives under esphome/components/ct002/:

• Emulator core — ct002.{h,cpp} (UDP server, consumer lifecycle, response builder, balancer wiring), protocol.{h,cpp} (CT002 wire format), balancer.{h,cpp} (load balancer with saturation tracking + efficiency rotation), sensor_backed.{h,cpp} + hampel/smoothing/pid.{h,cpp} + wrapper_base.h (filter pipeline).
• mqtt_insights: sub-block — mqtt_insights.{h,cpp} (per-consumer + device-level state publishing, command dispatch, Marstek broadcast), ha_discovery.{h,cpp} (HA Device Discovery payloads), marstek_responder.{h,cpp} (Marstek poll wire helpers — no ESPHome deps, host-gcc testable). Gated by #ifdef USE_MQTT.
• marstek_registration: sub-block — marstek_registration.{h,cpp} (cloud token → list → add → confirm state machine, ESPPreferences-persisted MAC). Gated by #ifdef USE_CT002_MARSTEK_REGISTRATION.
• Codegen — __init__.py owns the parent schema plus both nested sub-schemas; mqtt/http_request are enforced via cv.requires_component, json/md5 are auto-loaded.

Parity by discipline

Python under src/astrameter/ is canonical; the C++ is a mechanical mirror (same file/function/field names, same filter ordering). The rule is documented in CONTRIBUTING.md. Multiple parity audits were run against the Python source and its test suites; fixes landed for the balancer probe weights, consumer eviction, command-setter semantics, poll_interval EMA, smooth_target source, MQTT-insights field types / discovery re-publish, banker's rounding on wire values, parse_int strictness, and request dedup.

Intentional divergences (documented): the Shelly discovery path is absent (an ESPHome build is a single CT002), there's no ARP lookup (lwIP has no ARP table), origin.sw_version is a literal instead of a git SHA, and the MQTT layer uses ESPHome's client + synchronous publish instead of Python's asyncio queue/reconnect loop.

Testing

• Host-gcc gtest suites (built via CMake + FetchContent googletest, run from pytest): host_protocol_test (wire-format parity, golden-vector backed), host_balancer_test, host_wrappers_test (hampel/smoothing/pid), host_marstek_responder_test (Marstek poll/topic wire format).
• Shared golden vectors between the Python and C++ protocol suites.
• Tier-D end-to-end (test_host_e2e.py): drives the real BatterySimulator over UDP against the compiled host binary and asserts the response frame.
• ESPHome compile matrix in CI across 8 YAML configs (host, esp32-idf, esp32-arduino, 1-phase, no-filters, and the two sub-block targets).

Validated on real hardware

Flashed to an ESP32; fixed two bugs that only surface on-device: the HA consumer-discovery payload (~6.9 KB) exceeded ESPHome's json::build_json 5120-byte cap and was truncated to invalid JSON (now serialized via ArduinoJson directly), and the Marstek App-topic subscription missed the registration-applied MAC due to component setup ordering (now resolved lazily once the MAC is known).

Status & limitations

• Experimental — wider field testing welcome.
• ESP32 only (≥4 MB flash). ESP8266 is not supported in v1 — RAM/flash budgets are too tight with HTTPS+TLS, MQTT, and the balancer linked together.
• The discovery/state JSON builders have no host-gcc test yet (adding one needs ArduinoJson in the CMake harness); they're currently guarded by the ESPHome compile matrix and the parity audits.
• ver_v in Marstek replies is hardcoded to the default rather than derived from the cloud device-list version (not plumbed through on the ESPHome path; matches Python's fallback when the field is absent).

https://claude.ai/code/session_01TfTEE5DwAjqyassYprxaLp

Summary by CodeRabbit

• New Features

  • Experimental ESPHome external component "ct002" for ESP32 with optional MQTT insights, Home Assistant discovery, and Marstek registration.
  • Test-only control hooks and example ESPHome YAML for local emulation.

• Bug Fixes

  • Active-control logic updated so non-3-phase inputs run through the balancer.

• Tests

  • Extensive host/unit/e2e test suites, golden protocol vectors, and cross-backend E2E harness (python ↔ esphome).

• Chores

  • CI workflow: added ct002-focused jobs and new pytest esphome_e2e marker; release script improvements.

• Documentation

  • README, CONTRIBUTING, AGENTS, and CHANGELOG updates with usage and parity guidance.

[coderabbitai]

Walkthrough

Adds a full ESPHome CT002 external component with protocol, filters, balancer, MQTT insights, Marstek registration, host gtests, cross-backend E2E harness, CI jobs, YAML examples, docs, and a Python active-control tweak.

Changes

CT002 ESPHome port and cross-backend parity

Layer / File(s)

Summary

ESPHome CT002 feature set and tests esphome/components/ct002/*, tests/components/ct002/*, tests/*, .github/workflows/ci.yml, *.md, *.yaml, src/astrameter/ct002/ct002.py

Implements CT002 component, protocol, powermeter wrappers (Hampel, smoothing, deadband, PID), LoadBalancer (saturation/probe/efficiency), MQTT insights, Marstek registration/responder, HA discovery, test hooks, sensor-backed plumbing; adds host-gtests, generated protocol vectors, cross-backend E2E harness and tests, multiple ESPHome test YAMLs, CI jobs to compile/run these suites, documentation updates, and a Python active-control change allowing non-3-length balancing inputs.

Estimated code review effort

🎯 5 (Critical) | ⏱️ ~120 minutes

Possibly related PRs

• tomquist/AstraMeter#328 — Python-side Marstek MQTT responder added; this PR adds the ESPHome-side responder and integrates it into MQTT insights.
• tomquist/AstraMeter#321 — Saturation/efficiency partitioning changes; closely related to the C++ LoadBalancer saturation logic here.
• tomquist/AstraMeter#320 — Probe/lockup E2E harness and regression work; this PR extends that harness to dual backends (python/esphome).

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch claude/dazzling-cannon-vxuBk

[baal86]

Great idea! If this gets to happen I will definitely use it on my ESP32 based read-head, effectively making my own CT003.

[tomquist]

@baal86 Feel free to test, it should mostly work. You can find an example config here: https://github.com/tomquist/AstraMeter/blob/claude/dazzling-cannon-vxuBk/esphome.example.yaml

Make sure to replace this line:

  - source: github://tomquist/astrameter@main

with this line:

  - source: github://tomquist/astrameter@claude/dazzling-cannon-vxuBk

[baal86]

Compiles fine. There seems to be a name resolution issue with the Marstel registration:

[20:36:00][I][marstek_registration:158]: Backoff elapsed, retrying registration
[20:36:00][I][esp-idf:000]: E (61587) esp-tls: couldn't get hostname for :eu.hamedata.com: getaddrinfo() returns 202, addrinfo=0x0
[20:36:00][I][esp-idf:000]: E (61588) esp-tls: Failed to open new connection
[20:36:00][I][esp-idf:000]: E (61589) transport_base: Failed to open a new connection
[20:36:00][I][esp-idf:000]: E (61590) HTTP_CLIENT: Connection failed, sock < 0
[20:36:00][E][component:405]: http_request set Error flag: unspecified
[20:36:00][E][http_request.idf:139]: HTTP Request failed: ESP_ERR_HTTP_CONNECT
[20:36:00][W][marstek_registration:377]: HTTP -1 from https://eu.hamedata.com/app/Solar/v2_get_device.php?mailbox=<SECRET>&pwd=<NOT_TELLING_YOU>
[20:36:01][E][component:431]: http_request cleared Error flag

[tomquist]

@baal86 This doesn't seem related to the ct002 component itself. What does your wifi config look like?

[baal86]

@baal86 This doesn't seem related to the ct002 component itself. What does your wifi config look like?

wifi:
  ssid: !secret bmt_wifi_ssid
  password: !secret bmt_wifi_password
  fast_connect: True
  power_save_mode: NONE
  manual_ip:
    static_ip: 192.168.178.202
    gateway: 192.168.178.1
    subnet: 255.255.255.0

[tomquist]

If you add a manual ip section you should also provide a dns1 value.

[baal86]

If you add a manual ip section you should also provide a dns1 value.

Very good point! Still not. Also tried just putting in the IP of eu.hamedata.com directly. Same effect.

[20:55:28][I][marstek_registration:158]: Backoff elapsed, retrying registration
[20:55:28][I][esp-idf:000]: E (61031) esp-tls-mbedtls: mbedtls_ssl_handshake returned -0x2700
[20:55:28][I][esp-idf:000]: E (61032) esp-tls: Failed to open new connection
[20:55:28][I][esp-idf:000]: E (61033) transport_base: Failed to open a new connection
[20:55:28][I][esp-idf:000]: E (61036) HTTP_CLIENT: Connection failed, sock < 0
[20:55:28][E][component:405]: http_request set Error flag: unspecified
[20:55:28][E][http_request.idf:139]: HTTP Request failed: ESP_ERR_HTTP_CONNECT

Sounds like a certificate problem?

[tomquist]

Oh, you should set dns1 to the IP of your gateway 192.168.178.1

[baal86]

Oh, you should set dns1 to the IP of your gateway 192.168.178.1

Had to clean build files. Now it registers! Thanks for the help. I'll continue testing.

[baal86]

OK. Does work in real life! Great work! In my particular case it is not useable right now since I have literally no delay in my samples. Getting them straight from OBIS and the control loop is extra spicy. Will need some tuning on my side to get it stable.

[tomquist]

The example config has all the possible parameters to tweak your values. I would start with throttle, deadband and dedupe_window

[coderabbitai]

Actionable comments posted: 10

🧹 Nitpick comments (2)

tests/components/ct002/_gen_protocol_test_vectors.py (1)

48-50: ⚡ Quick win

Add <utility> to the generated header includes.

The generated header uses std::move but does not explicitly include <utility>. Please avoid relying on transitive includes from other headers.

Suggested fix

         "`#include` <cstdint>",
         "`#include` <string>",
+        "`#include` <utility>",
         "`#include` <vector>",

Also applies to: 80-80

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@tests/components/ct002/_gen_protocol_test_vectors.py` around lines 48 - 50,
The generated header uses std::move but the includes list that builds the header
(the array of include strings in _gen_protocol_test_vectors.py) does not contain
"<utility>"; add "`#include` <utility>" into that includes list (and the same
includes set used at the other occurrence around the 80-range) so the generated
C++ header explicitly includes <utility> instead of relying on transitive
includes; ensure the include string matches the existing format ("`#include`
<utility>") and is placed with the other standard headers.

tests/components/ct002/host_marstek_responder_test.cpp (1)

181-204: 💤 Low value

Acknowledge the maintenance burden of the proxy parser.

The hand-rolled proxy_read_code_str duplicates logic from marstek_registration.cpp's anonymous namespace without compile-time enforcement. The comment correctly notes that "audit will catch it" if the real implementation diverges, but this approach requires manual synchronization.

This is acceptable given the namespace constraints, but consider adding a build-time integration test that exercises the actual marstek_registration.cpp::read_code function through a public wrapper if feasible in the future.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@tests/components/ct002/host_marstek_responder_test.cpp` around lines 181 -
204, The test duplicates parsing logic in proxy_read_code_str which drifts from
marstek_registration.cpp::read_code; replace the hand-rolled parser by invoking
the real implementation (or a small public test-only wrapper around the
anonymous read_code) so the test exercises the actual parsing logic, or
alternatively add a build-time integration test that links against
marstek_registration's read_code via a test-only public wrapper; locate
proxy_read_code_str in the test and change it to call the wrapper/read_code, or
add the wrapper in marstek_registration.cpp and use it from the test to
eliminate duplication.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In @.github/workflows/ci.yml:
- Around line 87-88: The workflow passes the generated matrix.yaml unquoted into
a shell command (used by the step named "Compile ${{ matrix.yaml }}") which
allows filenames with metacharacters to break or inject commands; change the
compile invocation so the interpolated value is treated as a single literal
argument by adding a conventional option terminator and quoting the parameter
(i.e., ensure the esphome compile invocation uses -- and a quoted ${ matrix.yaml
}), and update the "Compile ${{ matrix.yaml }}" run line to use that safely
quoted/terminated form so filenames cannot influence shell parsing.
- Around line 50-61: Pin third-party actions to immutable commit SHAs and
prevent credential persistence: replace tag refs for actions/checkout@v4,
astral-sh/setup-uv@v5, actions/setup-python@v5, and actions/upload-artifact@v4
with their corresponding commit SHAs, and add with: persist-credentials: false
to every actions/checkout step to avoid leaving the checkout token in the runner
git config; also harden the esphome compile run command (esphome compile ${{
matrix.yaml }}) by quoting the matrix variable (e.g., " ${{ matrix.yaml }} " or
single-quoted) to ensure proper argument parsing.

In `@CHANGELOG.md`:
- Around line 5-8: The CHANGELOG.md "## Next" section contains multiple bullets
but must be reduced to a single continuously-edited branch-summary bullet; edit
the "## Next" block to remove the extra bullets and replace them with one
concise summary sentence that captures the overall outcome of this branch (e.g.,
key additions/fixes like ct002, Modbus UDP, HA powermeter fix, multi-Venus fix),
leaving that single bullet as the continuing summary for future edits.

In `@esphome/components/ct002/marstek_responder.cpp`:
- Around line 109-112: The current parsing of numeric fields accepts strings
with trailing junk because it only checks that strtol consumed at least one
digit; update the parsing logic for cd (where end and cd_val are used) and the
analogous p1 parsing (where p1_val/end are used) to also verify that *end ==
'\0' (i.e., the entire string was consumed) and return std::nullopt if not;
locate the strtol calls and their end-pointer checks in marstek_responder.cpp
(the cd parsing block around PollContext ctx and the p1 parsing block) and add
the trailing junk rejection so partially numeric inputs like "4x" or "2foo" are
treated as invalid.

In `@esphome/components/ct002/mqtt_insights.cpp`:
- Around line 89-96: The subscription gate currently checks marstek_mac_.empty()
so ensure_marstek_subscription_ is skipped once a MAC exists, which leaves
subscriptions stale if ct_mac changes; update the logic in the connected branch
(where connected && this->marstek_mqtt_enabled_ is evaluated) to always call
ensure_marstek_subscription_ while connected and enabled (remove the
marstek_mac_.empty() condition) and ensure ensure_marstek_subscription_ itself
is idempotent/safe to call repeatedly (or add a more robust check inside
ensure_marstek_subscription_ against the current marstek_mac_ to re-subscribe
only when it changes).

In `@esphome/components/ct002/sensor_backed.cpp`:
- Around line 28-35: The current freshness check uses max_stamp and so a fresh
phase can hide a stale one; change the logic to iterate each phase (using
this->num_phases_ and (*this->raw_stamp_ms_)[i]) and for each phase treat a zero
stamp or a stamp where (now - stamp) > this->max_sensor_age_ms_ as stale and
return {} (unavailable). Replace the max_stamp calculation and the single
comparison with a per-phase loop that validates each timestamp against now and
this->max_sensor_age_ms_ so any stale or missing phase causes the function to
return unavailable.

In `@esphome/components/ct002/test_hooks.cpp`:
- Around line 62-70: The pump_control_() loop dereferences control_socket_
without checking for null (start_control_server_ may leave it null); add an
early guard in CT002Component::pump_control_() to return immediately if
control_socket_ is null (or not open) before calling control_socket_->recvfrom,
so the method safely exits when the control socket wasn't created or failed to
bind.

In `@tests/_ct002_e2e_backend.py`:
- Around line 58-64: ensure_e2e_binary currently only compiles when E2E_BINARY
is missing, allowing stale artifacts to be used; change it to rebuild when
sources have changed by comparing modification times: if E2E_BINARY does not
exist OR E2E_BINARY.stat().st_mtime is older than E2E_YAML.stat().st_mtime (and
ideally older than any files under the e2e source dir like _E2E_DIR or its
children), invoke subprocess.run(["esphome", "compile", str(E2E_YAML)],
check=True, cwd=_E2E_DIR.parent.parent) to recompile; use pathlib to walk source
files and take the newest mtime to compare against E2E_BINARY to decide whether
to rebuild.

In `@tests/components/ct002/test_host_e2e.py`:
- Around line 95-116: Guard POSIX-only process-group handling around the
subprocess creation and teardown: when spawning the host binary, only pass
preexec_fn=os.setsid on POSIX (e.g. if os.name == "posix"), and on Windows use
subprocess.CREATE_NEW_PROCESS_GROUP via creationflags instead; in teardown,
protect os.killpg/os.getpgid with the same POSIX check and ensure the child is
still running (proc.poll() is None) before attempting to kill it to avoid
exceptions if it already exited. Update references in this fixture to the POSIX
symbols preexec_fn=os.setsid, os.killpg, and os.getpgid and add the non-POSIX
alternative using subprocess.CREATE_NEW_PROCESS_GROUP and safe existence checks
around kills.

In `@tests/components/ct002/test_host_protocol.py`:
- Around line 25-36: _pytestmark_ currently skips based on cmake/C++ but
_regen_vectors() always invokes "uv run python ..." which fails if uv is not
installed; update the _regen_vectors function to prefer using "uv" when
available but fall back to the system Python: detect uv with shutil.which("uv")
and if present call subprocess.run(["uv","run","python", str(HERE /
"_gen_protocol_test_vectors.py")], check=True, cwd=REPO_ROOT) otherwise call
subprocess.run([sys.executable, str(HERE / "_gen_protocol_test_vectors.py")],
check=True, cwd=REPO_ROOT); add imports for shutil and sys if not already
present.

---

Nitpick comments:
In `@tests/components/ct002/_gen_protocol_test_vectors.py`:
- Around line 48-50: The generated header uses std::move but the includes list
that builds the header (the array of include strings in
_gen_protocol_test_vectors.py) does not contain "<utility>"; add "`#include`
<utility>" into that includes list (and the same includes set used at the other
occurrence around the 80-range) so the generated C++ header explicitly includes
<utility> instead of relying on transitive includes; ensure the include string
matches the existing format ("`#include` <utility>") and is placed with the other
standard headers.

In `@tests/components/ct002/host_marstek_responder_test.cpp`:
- Around line 181-204: The test duplicates parsing logic in proxy_read_code_str
which drifts from marstek_registration.cpp::read_code; replace the hand-rolled
parser by invoking the real implementation (or a small public test-only wrapper
around the anonymous read_code) so the test exercises the actual parsing logic,
or alternatively add a build-time integration test that links against
marstek_registration's read_code via a test-only public wrapper; locate
proxy_read_code_str in the test and change it to call the wrapper/read_code, or
add the wrapper in marstek_registration.cpp and use it from the test to
eliminate duplication.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 617ba5de-dabd-4843-a4c6-667b1368ff50

📥 Commits

Reviewing files that changed from the base of the PR and between 1fb073a and fa9cc39.

📒 Files selected for processing (62)

• .github/workflows/ci.yml
• AGENTS.md
• CHANGELOG.md
• CONTRIBUTING.md
• README.md
• esphome.example.yaml
• esphome/components/ct002/__init__.py
• esphome/components/ct002/balancer.cpp
• esphome/components/ct002/balancer.h
• esphome/components/ct002/ct002.cpp
• esphome/components/ct002/ct002.h
• esphome/components/ct002/ha_discovery.cpp
• esphome/components/ct002/ha_discovery.h
• esphome/components/ct002/hampel.cpp
• esphome/components/ct002/hampel.h
• esphome/components/ct002/marstek_registration.cpp
• esphome/components/ct002/marstek_registration.h
• esphome/components/ct002/marstek_responder.cpp
• esphome/components/ct002/marstek_responder.h
• esphome/components/ct002/mqtt_insights.cpp
• esphome/components/ct002/mqtt_insights.h
• esphome/components/ct002/pid.cpp
• esphome/components/ct002/pid.h
• esphome/components/ct002/protocol.cpp
• esphome/components/ct002/protocol.h
• esphome/components/ct002/sensor_backed.cpp
• esphome/components/ct002/sensor_backed.h
• esphome/components/ct002/smoothing.cpp
• esphome/components/ct002/smoothing.h
• esphome/components/ct002/test_hooks.cpp
• esphome/components/ct002/wrapper_base.h
• pyproject.toml
• src/astrameter/ct002/ct002.py
• tests/_ct002_e2e_backend.py
• tests/components/ct002/.gitignore
• tests/components/ct002/CMakeLists.txt
• tests/components/ct002/_gen_protocol_test_vectors.py
• tests/components/ct002/fixtures/_populate_wire_hex.py
• tests/components/ct002/fixtures/protocol_golden_vectors.json
• tests/components/ct002/host_balancer_test.cpp
• tests/components/ct002/host_marstek_responder_test.cpp
• tests/components/ct002/host_protocol_test.cpp
• tests/components/ct002/host_protocol_test_vectors.h
• tests/components/ct002/host_wrappers_test.cpp
• tests/components/ct002/test.1phase.yaml
• tests/components/ct002/test.e2e.host.yaml
• tests/components/ct002/test.esp32-ard.yaml
• tests/components/ct002/test.esp32-idf.yaml
• tests/components/ct002/test.host.yaml
• tests/components/ct002/test.insights.esp32-ard.yaml
• tests/components/ct002/test.no-filters.yaml
• tests/components/ct002/test.registration.esp32-ard.yaml
• tests/components/ct002/test.registration.host.yaml
• tests/components/ct002/test_codegen.py
• tests/components/ct002/test_host_e2e.py
• tests/components/ct002/test_host_protocol.py
• tests/components/ct002/test_shared_e2e.py
• tests/test_ct002_active_control.py
• tests/test_ct002_protocol_parity.py
• tests/test_e2e_probe_lockup.py
• tests/test_efficiency_e2e.py
• tests/test_issue376_pv_passthrough.py

[coderabbitai]

⚠️ Potential issue | 🟠 Major | ⚡ Quick win

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
# Verify unpinned action refs and missing persist-credentials in this workflow
rg -n 'uses:\s*[^@]+@v[0-9]+' .github/workflows/ci.yml
rg -n 'uses:\s*actions/checkout@' .github/workflows/ci.yml -A3

Repository: tomquist/AstraMeter

Length of output: 1416

---

🏁 Script executed:

#!/bin/bash
set -euo pipefail

FILE=".github/workflows/ci.yml"

echo "== persist-credentials occurrences =="
rg -n "persist-credentials\s*:" "$FILE" || true

echo
echo "== checkout blocks with context (first few) =="
rg -n "uses:\s*actions/checkout@v4" "$FILE" -n -C 6 || true

echo
echo "== checkout blocks around the originally cited ranges =="
sed -n '40,140p' "$FILE" | nl -ba | sed -n '1,120p'

Repository: tomquist/AstraMeter

Length of output: 2347

---

🏁 Script executed:

#!/bin/bash
set -euo pipefail

FILE=".github/workflows/ci.yml"

echo "== persist-credentials occurrences =="
rg -n "persist-credentials\s*:" "$FILE" || true

echo
echo "== esphome compile occurrences =="
rg -n "esphome\s+compile" "$FILE" || true

echo
echo "== matrix.yaml usage occurrences =="
rg -n "\bmatrix\.yaml\b" "$FILE" || true

echo
echo "== checkout steps (with nearby context) =="
rg -n "uses:\s*actions/checkout@v4" "$FILE" -C 4 || true

echo
echo "== ranges around where new jobs likely compile =="
sed -n '1,230p' "$FILE" | nl -ba | sed -n '40,220p'

Repository: tomquist/AstraMeter

Length of output: 2101

---

🏁 Script executed:

#!/bin/bash
set -euo pipefail

FILE=".github/workflows/ci.yml"

echo "== matrix generation + compile step =="
awk 'NR>=70 && NR<=160 {printf "%d:%s\n", NR, $0}' "$FILE" | sed -n '1,120p'

echo
echo "== all checkout@v4 blocks (full with nearby) =="
awk '
  $0 ~ /uses:\s*actions\/checkout@v4/ { start=NR-2; if(start<1) start=1 }
  { if(start && NR>=start && NR<=NR+10) print }
' "$FILE"

Repository: tomquist/AstraMeter

Length of output: 10937

---

🌐 Web query:

actions/checkout v4 persist-credentials default value

💡 Result:

The default value for the persist-credentials input in actions/checkout@v4 is true [1][2]. When set to true (the default), the action automatically configures the local git repository with the authentication token or SSH key used to perform the checkout [1][2]. This allows subsequent steps in the workflow to execute authenticated git commands directly [1][2]. For security-sensitive environments, it is often recommended to set persist-credentials: false to prevent the credentials from being stored in the local git configuration, which could otherwise be accessible to other steps or accidentally included in artifacts [3][4]. Example usage to disable this behavior: - uses: actions/checkout@v4 with: persist-credentials: false

Citations:

• 1: https://github.com/actions/checkout/tree/v4
• 2: https://github.com/actions/checkout
• 3: Remove persist-credentials or change the default to false actions/checkout#485
• 4: https://yossarian.net/til/post/actions-checkout-can-leak-github-credentials/

---

🏁 Script executed:

#!/bin/bash
# Verify unpinned action refs and missing persist-credentials in this workflow
FILE=".github/workflows/ci.yml"

rg -n 'uses:\s*[^@]+@v[0-9]+' "$FILE"
rg -n 'uses:\s*actions/checkout@' "$FILE" -A3
rg -n 'persist-credentials\s*:' "$FILE" || true

Repository: tomquist/AstraMeter

Length of output: 1416

---

🏁 Script executed:

#!/bin/bash
# Verify unpinned action refs and missing persist-credentials in this workflow
rg -n 'uses:\s*[^@]+@v[0-9]+' .github/workflows/ci.yml
rg -n 'uses:\s*actions/checkout@' .github/workflows/ci.yml -A3

Repository: tomquist/AstraMeter

Length of output: 1416

---

🏁 Script executed:

#!/bin/bash
FILE=".github/workflows/ci.yml"
rg -n 'uses:\s*[^@]+@v[0-9]+' "$FILE"
rg -n 'uses:\s*actions/checkout@' "$FILE" -C 3
rg -n 'persist-credentials\s*:' "$FILE" || true
rg -n 'run:\s*esphome\s+compile\s+\$\{\{\s*matrix\.yaml\s*\}\}' "$FILE"

Repository: tomquist/AstraMeter

Length of output: 2056

---

Pin third-party actions to commit SHAs and disable checkout credential persistence in CI.

.github/workflows/ci.yml uses tag-only action refs (actions/checkout@v4, astral-sh/setup-uv@v5, actions/setup-python@v5, actions/upload-artifact@v4) and every actions/checkout@v4 step omits with: persist-credentials: false (default is true, so the checkout token is persisted in the runner git config). Also harden ct002-esphome-compile: run: esphome compile ${{ matrix.yaml }} should quote the value.

🔧 Suggested hardening pattern

-      - name: Checkout repository
-        uses: actions/checkout@v4
+      - name: Checkout repository
+        uses: actions/checkout@<full_commit_sha>
+        with:
+          persist-credentials: false

-      - name: Install uv
-        uses: astral-sh/setup-uv@v5
+      - name: Install uv
+        uses: astral-sh/setup-uv@<full_commit_sha>
         with:
           enable-cache: true

-      - name: Set up Python
-        uses: actions/setup-python@v5
+      - name: Set up Python
+        uses: actions/setup-python@<full_commit_sha>
         with:
           python-version: "3.13"

🧰 Tools

🪛 zizmor (1.25.2)

[warning] 50-51: credential persistence through GitHub Actions artifacts (artipacked): does not set persist-credentials: false

(artipacked)

---

[error] 51-51: unpinned action reference (unpinned-uses): action is not pinned to a hash (required by blanket policy)

(unpinned-uses)

---

[error] 54-54: unpinned action reference (unpinned-uses): action is not pinned to a hash (required by blanket policy)

(unpinned-uses)

---

[error] 59-59: unpinned action reference (unpinned-uses): action is not pinned to a hash (required by blanket policy)

(unpinned-uses)

---

[error] 54-54: runtime artifacts potentially vulnerable to a cache poisoning attack (cache-poisoning): this step

(cache-poisoning)

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In @.github/workflows/ci.yml around lines 50 - 61, Pin third-party actions to
immutable commit SHAs and prevent credential persistence: replace tag refs for
actions/checkout@v4, astral-sh/setup-uv@v5, actions/setup-python@v5, and
actions/upload-artifact@v4 with their corresponding commit SHAs, and add with:
persist-credentials: false to every actions/checkout step to avoid leaving the
checkout token in the runner git config; also harden the esphome compile run
command (esphome compile ${{ matrix.yaml }}) by quoting the matrix variable
(e.g., " ${{ matrix.yaml }} " or single-quoted) to ensure proper argument
parsing.

[coderabbitai]

⚠️ Potential issue | 🟠 Major | ⚡ Quick win

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
# Confirm direct matrix interpolation into shell execution points
rg -n 'esphome compile \${{\s*matrix\.yaml\s*}}' .github/workflows/ci.yml

Repository: tomquist/AstraMeter

Length of output: 205

---

🏁 Script executed:

#!/bin/bash
set -euo pipefail

# Show the referenced lines for exact context
nl -ba .github/workflows/ci.yml | sed -n '70,130p'

# Find how matrix.yaml is used in esphome compile invocations
rg -n "matrix\.yaml" .github/workflows/ci.yml
rg -n "esphome compile" .github/workflows/ci.yml
rg -n "compile \\$\\{\\{\\s*matrix\\.yaml\\s*\\}\\}" .github/workflows/ci.yml || true

Repository: tomquist/AstraMeter

Length of output: 106

---

🏁 Script executed:

#!/bin/bash
set -euo pipefail

# Print the referenced region with line numbers via awk
awk 'NR>=70 && NR<=130 {printf "%d:%s\n", NR, $0}' .github/workflows/ci.yml

# Find all occurrences of matrix.yaml and esphome compile usage
rg -n "matrix\.yaml" .github/workflows/ci.yml
rg -n "esphome compile" .github/workflows/ci.yml
rg -n "esphome compile.*\\$\\{\\{\\s*matrix\\.yaml\\s*\\}\\}" .github/workflows/ci.yml || true

Repository: tomquist/AstraMeter

Length of output: 2290

---

🏁 Script executed:

#!/bin/bash
set -euo pipefail

rg -n "\\$\\{\\{\\s*matrix\\." .github/workflows/ci.yml || true
rg -n "esphome compile" .github/workflows/ci.yml

Repository: tomquist/AstraMeter

Length of output: 792

---

Quote and pass matrix.yaml safely to esphome compile to prevent filename-based command injection (CI).

matrix.yaml is built from repo filenames (lines 87-88) and then used unquoted in run: esphome compile ${{ matrix.yaml }} (lines 109-110); filenames containing shell metacharacters (e.g., ;, $(), backticks, spaces) can alter command parsing/execution.

🔒 Safer compile invocation

-      - name: Compile ${{ matrix.yaml }}
-        run: esphome compile ${{ matrix.yaml }}
+      - name: Compile ${{ matrix.yaml }}
+        run: |
+          set -euo pipefail
+          esphome compile -- "${{ matrix.yaml }}"

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In @.github/workflows/ci.yml around lines 87 - 88, The workflow passes the
generated matrix.yaml unquoted into a shell command (used by the step named
"Compile ${{ matrix.yaml }}") which allows filenames with metacharacters to
break or inject commands; change the compile invocation so the interpolated
value is treated as a single literal argument by adding a conventional option
terminator and quoting the parameter (i.e., ensure the esphome compile
invocation uses -- and a quoted ${ matrix.yaml }), and update the "Compile ${{
matrix.yaml }}" run line to use that safely quoted/terminated form so filenames
cannot influence shell parsing.

[coderabbitai]

⚠️ Potential issue | 🟡 Minor | ⚡ Quick win

Reject partially numeric cd/p1 values.

Line 110 and Line 121 accept values with trailing junk ("4x", "2foo"). That makes malformed control payloads look valid.

Suggested fix

+#include <cerrno>
+#include <climits>
@@
   const std::string *cd = find(kv, "cd");
   if (cd == nullptr) return std::nullopt;
+  errno = 0;
   char *end = nullptr;
   long cd_val = std::strtol(cd->c_str(), &end, 10);
-  if (end == cd->c_str()) return std::nullopt;
+  if (end == cd->c_str() || *end != '\0' || errno == ERANGE || cd_val < INT_MIN || cd_val > INT_MAX)
+    return std::nullopt;
@@
     const std::string *p1 = find(kv, "p1");
     if (p1 == nullptr) return std::nullopt;
+    errno = 0;
     char *e2 = nullptr;
     long sid = std::strtol(p1->c_str(), &e2, 10);
-    if (e2 == p1->c_str()) return std::nullopt;
+    if (e2 == p1->c_str() || *e2 != '\0' || errno == ERANGE || sid < INT_MIN || sid > INT_MAX)
+      return std::nullopt;

Also applies to: 121-123

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@esphome/components/ct002/marstek_responder.cpp` around lines 109 - 112, The
current parsing of numeric fields accepts strings with trailing junk because it
only checks that strtol consumed at least one digit; update the parsing logic
for cd (where end and cd_val are used) and the analogous p1 parsing (where
p1_val/end are used) to also verify that *end == '\0' (i.e., the entire string
was consumed) and return std::nullopt if not; locate the strtol calls and their
end-pointer checks in marstek_responder.cpp (the cd parsing block around
PollContext ctx and the p1 parsing block) and add the trailing junk rejection so
partially numeric inputs like "4x" or "2foo" are treated as invalid.

[coderabbitai]

⚠️ Potential issue | 🟠 Major | ⚡ Quick win

Always re-check Marstek identity while connected.

Line 94 gates ensure_marstek_subscription_() on marstek_mac_.empty(). If ct_mac changes later, subscriptions can stay on stale topics.

Suggested fix

-  if (connected && this->marstek_mqtt_enabled_ && this->marstek_mac_.empty()) {
+  if (connected && this->marstek_mqtt_enabled_) {
     this->ensure_marstek_subscription_();
   }

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	// Lazily subscribe to Marstek App topics once the MAC is known. Only
	// poll while not yet subscribed (marstek_mac_ empty) so there's no
	// steady-state per-loop cost; the MAC is applied once at boot by
	// marstek_registration and cleared here only on disconnect, so a
	// re-subscribe is driven by the reconnect path, not by polling.
	if (connected && this->marstek_mqtt_enabled_ && this->marstek_mac_.empty()) {
	this->ensure_marstek_subscription_();
	}
	// Lazily subscribe to Marstek App topics once the MAC is known. Only
	// poll while not yet subscribed (marstek_mac_ empty) so there's no
	// steady-state per-loop cost; the MAC is applied once at boot by
	// marstek_registration and cleared here only on disconnect, so a
	// re-subscribe is driven by the reconnect path, not by polling.
	if (connected && this->marstek_mqtt_enabled_) {
	this->ensure_marstek_subscription_();
	}

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@esphome/components/ct002/mqtt_insights.cpp` around lines 89 - 96, The
subscription gate currently checks marstek_mac_.empty() so
ensure_marstek_subscription_ is skipped once a MAC exists, which leaves
subscriptions stale if ct_mac changes; update the logic in the connected branch
(where connected && this->marstek_mqtt_enabled_ is evaluated) to always call
ensure_marstek_subscription_ while connected and enabled (remove the
marstek_mac_.empty() condition) and ensure ensure_marstek_subscription_ itself
is idempotent/safe to call repeatedly (or add a more robust check inside
ensure_marstek_subscription_ against the current marstek_mac_ to re-subscribe
only when it changes).

[coderabbitai]

⚠️ Potential issue | 🟠 Major | ⚡ Quick win

Per-phase staleness is not enforced in freshness check.

Line 30 tracks only the newest timestamp; one fresh phase can mask another stale phase, and stale data is still emitted. Validate age per phase instead of using max_stamp.

💡 Suggested fix

-    uint32_t max_stamp = 0;
-    for (uint8_t i = 0; i < this->num_phases_; ++i) {
-      if ((*this->raw_stamp_ms_)[i] > max_stamp) max_stamp = (*this->raw_stamp_ms_)[i];
-    }
-    // Unsigned subtraction wraps correctly across the 49.7-day boundary
-    // (matches the rollover-safe pattern documented in the plan).
-    if (max_stamp != 0 && (now - max_stamp) > this->max_sensor_age_ms_) {
-      return {};  // unavailable
-    }
+    // Unsigned subtraction wraps correctly across the 49.7-day boundary.
+    for (uint8_t i = 0; i < this->num_phases_; ++i) {
+      const uint32_t stamp = (*this->raw_stamp_ms_)[i];
+      if (stamp == 0 || (now - stamp) > this->max_sensor_age_ms_) {
+        return {};  // unavailable
+      }
+    }

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	uint32_t max_stamp = 0;
	for (uint8_t i = 0; i < this->num_phases_; ++i) {
	if ((*this->raw_stamp_ms_)[i] > max_stamp) max_stamp = (*this->raw_stamp_ms_)[i];
	}
	// Unsigned subtraction wraps correctly across the 49.7-day boundary
	// (matches the rollover-safe pattern documented in the plan).
	if (max_stamp != 0 && (now - max_stamp) > this->max_sensor_age_ms_) {
	return {}; // unavailable
	// Unsigned subtraction wraps correctly across the 49.7-day boundary.
	for (uint8_t i = 0; i < this->num_phases_; ++i) {
	const uint32_t stamp = (*this->raw_stamp_ms_)[i];
	if (stamp == 0 || (now - stamp) > this->max_sensor_age_ms_) {
	return {}; // unavailable
	}
	}

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@esphome/components/ct002/sensor_backed.cpp` around lines 28 - 35, The current
freshness check uses max_stamp and so a fresh phase can hide a stale one; change
the logic to iterate each phase (using this->num_phases_ and
(*this->raw_stamp_ms_)[i]) and for each phase treat a zero stamp or a stamp
where (now - stamp) > this->max_sensor_age_ms_ as stale and return {}
(unavailable). Replace the max_stamp calculation and the single comparison with
a per-phase loop that validates each timestamp against now and
this->max_sensor_age_ms_ so any stale or missing phase causes the function to
return unavailable.

[coderabbitai]

⚠️ Potential issue | 🟡 Minor | ⚡ Quick win

Guard pump_control_() against missing socket before recvfrom.

Line 67 dereferences control_socket_ unconditionally, but start_control_server_() can leave it null (disabled or bind failure). Add an early null check.

💡 Suggested fix

 void CT002Component::pump_control_() {
+  if (this->control_socket_ == nullptr) return;
   for (int i = 0; i < 8; ++i) {

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@esphome/components/ct002/test_hooks.cpp` around lines 62 - 70, The
pump_control_() loop dereferences control_socket_ without checking for null
(start_control_server_ may leave it null); add an early guard in
CT002Component::pump_control_() to return immediately if control_socket_ is null
(or not open) before calling control_socket_->recvfrom, so the method safely
exits when the control socket wasn't created or failed to bind.

[coderabbitai]

⚠️ Potential issue | 🟠 Major | ⚡ Quick win

Rebuild the ESPHome e2e binary when sources change.

ensure_e2e_binary() only compiles when the binary is missing, so local runs can silently use stale artifacts after C++/YAML edits. That can produce false green/red e2e results.

Suggested fix

 def ensure_e2e_binary() -> None:
-    if not E2E_BINARY.exists():
-        subprocess.run(
-            ["esphome", "compile", str(E2E_YAML)],
-            check=True,
-            cwd=_E2E_DIR.parent.parent,
-        )
+    subprocess.run(
+        ["esphome", "compile", str(E2E_YAML)],
+        check=True,
+        cwd=_E2E_DIR.parent.parent,
+    )

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@tests/_ct002_e2e_backend.py` around lines 58 - 64, ensure_e2e_binary
currently only compiles when E2E_BINARY is missing, allowing stale artifacts to
be used; change it to rebuild when sources have changed by comparing
modification times: if E2E_BINARY does not exist OR E2E_BINARY.stat().st_mtime
is older than E2E_YAML.stat().st_mtime (and ideally older than any files under
the e2e source dir like _E2E_DIR or its children), invoke
subprocess.run(["esphome", "compile", str(E2E_YAML)], check=True,
cwd=_E2E_DIR.parent.parent) to recompile; use pathlib to walk source files and
take the newest mtime to compare against E2E_BINARY to decide whether to
rebuild.

[coderabbitai]

⚠️ Potential issue | 🟠 Major | ⚡ Quick win

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
# Verify POSIX-only APIs are used and whether there is a platform guard.
rg -n --type=py 'preexec_fn=os\.setsid|os\.killpg|pytest\.mark\.skipif|os\.name|sys\.platform' tests/components/ct002/test_host_e2e.py

Repository: tomquist/AstraMeter

Length of output: 239

---

Guard POSIX-only process-group handling in this fixture.
tests/components/ct002/test_host_e2e.py calls POSIX-only APIs (preexec_fn=os.setsid and unguarded os.killpg(os.getpgid(proc.pid), ...)) with no non-POSIX/platform skip, so it can fail off POSIX and teardown can raise if the child exits before the cleanup.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@tests/components/ct002/test_host_e2e.py` around lines 95 - 116, Guard
POSIX-only process-group handling around the subprocess creation and teardown:
when spawning the host binary, only pass preexec_fn=os.setsid on POSIX (e.g. if
os.name == "posix"), and on Windows use subprocess.CREATE_NEW_PROCESS_GROUP via
creationflags instead; in teardown, protect os.killpg/os.getpgid with the same
POSIX check and ensure the child is still running (proc.poll() is None) before
attempting to kill it to avoid exceptions if it already exited. Update
references in this fixture to the POSIX symbols preexec_fn=os.setsid, os.killpg,
and os.getpgid and add the non-POSIX alternative using
subprocess.CREATE_NEW_PROCESS_GROUP and safe existence checks around kills.

[coderabbitai]

⚠️ Potential issue | 🟡 Minor | ⚡ Quick win

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
# Verify current skip gate and uv invocation.
rg -n --type=py '_have\("cmake"\)|_have\("g\+\+"\)|_have\("clang\+\+"\)|_have\("uv"\)|\["uv",' tests/components/ct002/test_host_protocol.py

Repository: tomquist/AstraMeter

Length of output: 212

---

🏁 Script executed:

#!/bin/bash
set -euo pipefail

FILE="tests/components/ct002/test_host_protocol.py"

echo "=== File header + skip gate + _regen_vectors ==="
sed -n '1,120p' "$FILE" | cat -n

echo
echo "=== All _regen_vectors references ==="
rg -n "_regen_vectors\(" "$FILE"

echo
echo "=== Any uv-related gating elsewhere in this file ==="
rg -n --type=py "_have\(\"uv\"\)|uv run|\\buv\\b" "$FILE"

Repository: tomquist/AstraMeter

Length of output: 2878

---

🏁 Script executed:

#!/bin/bash
set -euo pipefail

FILE="tests/components/ct002/_gen_protocol_test_vectors.py"
echo "=== $FILE (first 120 lines) ==="
sed -n '1,120p' "$FILE" | cat -n

echo
echo "=== External imports heuristic (non-stdlib) ==="
# crude heuristic: show import lines; will manually judge from them
rg -n "^import |^from " "$FILE" | head -n 80

Repository: tomquist/AstraMeter

Length of output: 4048

---

Handle missing uv when regenerating protocol test vectors.

pytestmark skips only on cmake/C++ toolchain, but _regen_vectors() unconditionally runs uv run ..., so environments with cmake/compiler but no uv fail. _gen_protocol_test_vectors.py is stdlib-only, so falling back to sys.executable is sufficient.

💡 Proposed patch

@@
 import shutil
 import subprocess
+import sys
 from pathlib import Path
@@
 def _regen_vectors() -> None:
+    cmd = (
+        ["uv", "run", "python", str(HERE / "_gen_protocol_test_vectors.py")]
+        if _have("uv")
+        else [sys.executable, str(HERE / "_gen_protocol_test_vectors.py")]
+    )
     subprocess.run(
-        ["uv", "run", "python", str(HERE / "_gen_protocol_test_vectors.py")],
+        cmd,
         check=True,
         cwd=REPO_ROOT,
     )

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	pytestmark = pytest.mark.skipif(
	not (_have("cmake") and (_have("g++") or _have("clang++"))),
	reason="cmake and a C++ compiler are required for host-gcc protocol tests",
	)
	def _regen_vectors() -> None:
	subprocess.run(
	["uv", "run", "python", str(HERE / "_gen_protocol_test_vectors.py")],
	check=True,
	cwd=REPO_ROOT,
	)
	pytestmark = pytest.mark.skipif(
	not (_have("cmake") and (_have("g++") or _have("clang++"))),
	reason="cmake and a C++ compiler are required for host-gcc protocol tests",
	)
	def _regen_vectors() -> None:
	cmd = (
	["uv", "run", "python", str(HERE / "_gen_protocol_test_vectors.py")]
	if _have("uv")
	else [sys.executable, str(HERE / "_gen_protocol_test_vectors.py")]
	)
	subprocess.run(
	cmd,
	check=True,
	cwd=REPO_ROOT,
	)

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@tests/components/ct002/test_host_protocol.py` around lines 25 - 36,
_pytestmark_ currently skips based on cmake/C++ but _regen_vectors() always
invokes "uv run python ..." which fails if uv is not installed; update the
_regen_vectors function to prefer using "uv" when available but fall back to the
system Python: detect uv with shutil.which("uv") and if present call
subprocess.run(["uv","run","python", str(HERE /
"_gen_protocol_test_vectors.py")], check=True, cwd=REPO_ROOT) otherwise call
subprocess.run([sys.executable, str(HERE / "_gen_protocol_test_vectors.py")],
check=True, cwd=REPO_ROOT); add imports for shutil and sys if not already
present.

[coderabbitai]

♻️ Duplicate comments (1)

CHANGELOG.md (1)

5-8: ⚠️ Potential issue | 🟡 Minor | ⚡ Quick win

Consolidate ## Next to a single branch-summary bullet.

## Next still has multiple bullets (Lines 5-8). Keep one continuously edited bullet and fold these points into that single summary line.

✏️ Suggested shape

- - **Added** ...
- - **Added** ...
- - **Fixed** ...
- - **Fixed** ...
+ - **Added/Fixed** <single concise summary of this branch’s user-visible outcome>.

As per coding guidelines: "For user-facing work, keep one bullet under ## Next in CHANGELOG.md that summarizes the overall outcome of the branch; edit that bullet on later iterations rather than appending new bullets."

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@CHANGELOG.md` around lines 5 - 8, CHANGELOG.md's "## Next" currently contains
multiple bullets (the ESPHome ct002 addition, Modbus UDP TRANSPORT option, Home
Assistant powermeter fix, and multi-Venus CT002 cross-talk fix). Replace those
separate bullets with a single continuously edited branch-summary bullet under
"## Next" that briefly summarizes the branch outcome and fold the individual
points into that one line (mentioning the ct002 ESPHome external component,
Modbus UDP/TRANSPORT option, Home Assistant powermeter startup fix, and
multi‑Venus CT002 cross-talk / *_chrg_power *_dchrg_power fix) so the changelog
has only one editable bullet going forward.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Duplicate comments:
In `@CHANGELOG.md`:
- Around line 5-8: CHANGELOG.md's "## Next" currently contains multiple bullets
(the ESPHome ct002 addition, Modbus UDP TRANSPORT option, Home Assistant
powermeter fix, and multi-Venus CT002 cross-talk fix). Replace those separate
bullets with a single continuously edited branch-summary bullet under "## Next"
that briefly summarizes the branch outcome and fold the individual points into
that one line (mentioning the ct002 ESPHome external component, Modbus
UDP/TRANSPORT option, Home Assistant powermeter startup fix, and multi‑Venus
CT002 cross-talk / *_chrg_power *_dchrg_power fix) so the changelog has only one
editable bullet going forward.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 0c8ac0d5-e7bf-4004-a931-cd48e0529ac2

📥 Commits

Reviewing files that changed from the base of the PR and between fa9cc39 and e473b41.

📒 Files selected for processing (5)

• CHANGELOG.md
• README.md
• esphome.example.yaml
• release.sh
• tests/test_release.py