You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
New SHA-256:390758b059370beff8f6f0a8e014333c4f40c26e7ef5551db61a057798875a7d
Detected: 2026-09-14T06:49:17+00:00
Review
Read the new document and note what changed (registers renumbered,
parameters added, scaling corrected).
Cross-check against the registers/parameters easee_cloud reads.
If the driver is affected, open a fix and bump its version;
otherwise close this issue.
Detection is a raw-byte hash: the file's bytes moved. That can be a
genuine content change or just rebuild churn in the PDF metadata —
this issue is a prompt to look, not proof the registers changed.
The driver polls a fixed set of observation IDs (OBS_IDS = "48,96,103,109,120,121,124,183,194,223", defined at lines ~272-283). Checking each against the current page:
ID
Driver name
Doc name (current)
Type
Match
48
OBS_DYN_CURRENT
Dynamic Charger Current
Double (A)
✅
96
OBS_REASON_NO_CUR
Reason For No Current
Integer (enum)
✅
103
OBS_CABLE_LOCKED
Cable Locked
Boolean
✅
109
OBS_OP_MODE
Charger Op Mode
Integer (enum)
✅
120
OBS_TOTAL_POWER
Total Power
Double (kW)
✅
121
OBS_SESSION_ENERGY
Session Energy
Double (kWh)
✅
124
OBS_LIFETIME_ENERGY
Lifetime Energy
Double (kWh)
✅
183
OBS_CURRENT
Int Current T3
Double
✅
194
OBS_VOLTAGE
In Volt T2T3
Double
✅
223
OBS_SESSION_START
Charge Session Start
String (JSON)
✅
All 10 IDs the driver reads are still present, with the same names, types and units the driver assumes (notably the kW→W and kWh→Wh scaling at driver_poll, and the boolean coercion for 103). The enum value tables for 96 (REASON_LABELS) and 109 (OP_MODE_LABELS) live on separate linked enumeration pages, not on this specific page — this hash change is on the observation-ID listing itself, which reads as unchanged content (very likely formatting/metadata churn, consistent with the note that this is a raw byte-hash diff).
For context: this driver's session/observation handling was actively reworked and merged just hours before this hash change was detected — PR #116 (merged 2026-09-14T03:51 UTC, this issue opened 2026-09-14T06:49 UTC) reworked observation freshness and session-proof logic against the live API. So this area is current, not stale.
No discrepancy found between what the driver expects and what the document currently states. Closing as not-planned.
Activity
frahlg commentedon Sep 14, 2026
Reviewed the live content of https://developer.easee.com/docs/charger-observation-ids against what
drivers/lua/easee_cloud.luaactually reads.The driver polls a fixed set of observation IDs (
OBS_IDS = "48,96,103,109,120,121,124,183,194,223", defined at lines ~272-283). Checking each against the current page:All 10 IDs the driver reads are still present, with the same names, types and units the driver assumes (notably the kW→W and kWh→Wh scaling at
driver_poll, and the boolean coercion for 103). The enum value tables for 96 (REASON_LABELS) and 109 (OP_MODE_LABELS) live on separate linked enumeration pages, not on this specific page — this hash change is on the observation-ID listing itself, which reads as unchanged content (very likely formatting/metadata churn, consistent with the note that this is a raw byte-hash diff).For context: this driver's session/observation handling was actively reworked and merged just hours before this hash change was detected — PR #116 (merged 2026-09-14T03:51 UTC, this issue opened 2026-09-14T06:49 UTC) reworked observation freshness and session-proof logic against the live API. So this area is current, not stale.
No discrepancy found between what the driver expects and what the document currently states. Closing as not-planned.
Generated by Claude Code