The dashboard started with gauges and histories of equal weight, so a power failure and its effect were hard to find. The first rows now show the charge, the power source, the time remaining, and the errors of each device. Below them, one band shows each period on battery power, and one chart shows the charge of the three devices.
The charge history adds the pack current of the JBD and the River 2 on a right axis, as dashed lines in the color of each device. The charge limits apply to the charge lines only.
A time-remaining panel with no estimate now names the state. The JBD panel shows "On mains" while the pack does not discharge. The River 2 panel shows "Idle" while the device reports 5999 minutes for the two directions.
The River 2 errors panel names the module that reports a code and shows the code in hexadecimal. The solar controller reports 4096 with no solar input, and the panel counts that value as zero.
The server read each Bluetooth device every 60 seconds, and each read is one scan and one connection. An adapter that fails after many connections had no way to get fewer of them.
--bluetooth-interval sets the wait after a good Bluetooth read, and the default stays at 60 seconds. Clients receive the last read during the wait. A failed read still tries again after 60 seconds, or after the interval for a shorter one.
tools/generate_grafana_ups.py writes grafana-ups.json, which shows the JBD network UPS, the River 2 desktop UPS, and the River 3 NAS UPS in three columns. Git ignores the generated file.
Each EcoFlow column shows one time-remaining panel in hours, for the direction in use. The River 2 histories add pack current and remaining capacity. The River 3 reports neither, so its histories add battery power and charge.
The polling threads are daemon threads, so a stop ended the server in the middle of a Bluetooth read. The read got no disconnect, and the adapter kept an open connection or a scan in progress.
The server now sets the stop event at exit, closes its socket, and waits at most 20 seconds for the polling threads. A Bluetooth read that sees the stop event cancels its current step, disconnects, and returns. A read that waits for the adapter does not connect after a stop.
collect_data printed a failure only with -v, and gave no reason for a failed connection, a failed authentication, or a read with no telemetry. A device that was out of reach showed only as missing data.
Each failed read now gives one log line with the reason, at every debug level. The device object keeps the reason for a failed connection in last_error.
If the server had no data for the device, bmspy-ups slept and read again with no report. A battery that stopped answering during a power failure gave no email and no shutdown.
bmspy-ups now sends one email after 5 minutes with no data, and one more when the data returns. A new email for a loss needs another 5 minutes with no data after that return. The same rule applies to the wait at startup.
Set PYTHONUNBUFFERED=1 in the three service files. Python buffers its output if it does not write to a terminal, so log lines reached the journal late.
The charge tests and the power tests were one if/elif chain. If mains power returned while the charge was below the warning threshold, the loop did not report it, did not cancel the shutdown, and did not clear the flags for the next power failure.
The power tests now run first, on their own. The charge tests apply only while the battery discharges for two consecutive readings, so a low battery on mains power starts no shutdown. A device that reports no battery current is judged by its charge alone.
The loop compared the current reading with history[1], history[2], and history[3]. The history holds the oldest reading first, so these were three of the oldest readings. A plain power failure sent no email, and the shutdown started 4 minutes late. The loop now uses the three readings before the current one.
Four tests used reading sequences that matched the old comparison. They now use plain sequences.
socket_comms now raises DaemonError with the server message if the reply has the status ERROR. Before, the reply reached code that expects device data, and that code raised KeyError.
read_request called recv once for the length and once for the body. A stream socket can deliver a request in several pieces, and the server then rejected that request.
The server now uses the read loop of the client, which moves to utilities as recv_exact. The server refuses a request above 64 KiB.
The server handled one client at a time, so a client that connected and sent nothing blocked every other client. The request handling is now in handle_client, and the main loop starts one thread for each connection.
The server closes a connection that stays silent for 5 seconds, so its thread ends.
If a client disconnected before its reply, send_response raised OSError and the server ended. The request loop now logs the error and waits for the next client. An error from accept() still ends the server.
An unknown command ended the server loop, so one bad request stopped the service for every client. The server now sends {"status": "ERROR", "message": ...} and continues. A request with no command, or one that is not a JSON object, gets the same reply.
The server tests stopped the server with an unknown command. They now stop it through accept().
collect_data returns False if requestMessage returns False or None, for the first request and for the second one. Before the earlier fix, these values raised TypeError and ended the server.
Put each condition before its statement. Split long sentences. Use one term for each concept. Replace the description of possible checklist changes with a statement of the present behavior.
bmspy-ups selected its device from the devices that had data at startup, and exited with status 0 if the selection failed. One failed first read ended the monitor or made it select the wrong device.
The server has a new DEVICES command that lists the configured names, and client.list_devices calls it. bmspy-ups selects from that list and then waits for the first data of its device. A wrong or missing --device now exits with status 2. The service file does not restart on that status, because a restart cannot correct it.
An older server ends on the unknown DEVICES command, so update the server first.
A Bluetooth collection listened for a fixed 10 seconds and returned the data that arrived. If a required report was late, the snapshot missed shared measurements without notice.
Each decoder now tracks the reports behind the shared measurements. A collection listens for 10 seconds, then until those reports arrive, for at most 30 seconds in total. The new ef_missing_measurements field names the reports that did not arrive. The server log shows one line each time that list changes.
The diagnostic tool now uses the River 2 heartbeat list from the decoder.
The recipient check used a regular expression that never matched, so a full address got the host name appended. The main loop passed the device name with a parameter name that read_data does not have. A canceled shutdown did not clear its flag, so a second incident did not schedule a shutdown. The recorded shutdown time used the wrong unit. The --port option stayed text and never matched the STARTTLS ports.
A failed JBD read retried forever inside a client request and blocked every other device and client. The server now starts one polling thread for each device, and a request returns the last snapshot. Only the first request after startup waits, until each device completes one read attempt.
One lock makes sure that only one Bluetooth collection runs at a time. A serial BMS retries a failed read after 1 second and keeps its last snapshot for 30 seconds. An error in a collection does not end the polling of that device.
The server tests left main() in a thread after each test. These threads kept their patches active and delayed the suite. Each test now sends an invalid command, which ends the server loop, and then joins the thread. The suite time drops from about 8.5 seconds to about 2 seconds.
Remove one unused line from a test.
collect_data called len() on each response. A port that cannot open returns False or None, so the call raised TypeError and ended the server. Treat each empty or absent response as a failed read.
River 2 reported total power and temperature as integers. River 3 reported temperature as an integer. InfluxDB keeps one field type for each measurement, so it rejected these points when another device wrote floats to the same measurement.
Correct the spelling in the JBD and EcoFlow snapshots, the diagnostic tool, and the tests. UPS.from_dict maps the old name to the new name, so a client still reads data from an older server during an upgrade.
InfluxDB and Prometheus now receive this measurement under the new name.
The upstream River 3 schema spells field 355 `ups_alram`. Use `ups_alarm` in the decoder, the generated schema, and the field reference. The schema generator now applies a rename map, so a regenerated schema keeps the corrected name.
Decode River 3 separately and accumulate partial reports against a field-level checklist. Keep watch mode running after completion.
Label fields and nested records from the pinned pr705 schema. Preserve unknown values and document unverified meanings and routing.
Add synthetic regressions, the River 3 field reference, schema-generation tooling, and upstream attribution.
Use standard project metadata, dependency groups, and the uv build backend. The new lock preserves the previous dependency versions. Documentation and services use uv, and services leave their shared environment unchanged at startup.
Remove the entry point for the empty USB HID prototype. Exclude local prototypes, captures, and the external Type 7 lookup table from version control and distributions.
Serial paths and collector interfaces remain compatible. Bluetooth polling uses a 60-second cache and backoff. Offline Bluetooth devices no longer retry forever inside a client request. Invalid device specifications do not echo account IDs.
Decode River 2 charge, temperature, power flow, and time estimates. Preserve precise charge readings across packet reordering and omit unavailable time estimates.
Require valid PD, EMS, and BMS core heartbeats for diagnostic completion. Report optional packets separately and support raw watch output with reliable disconnect cleanup. Use synthetic regression tests and keep captures, credentials, and external keydata private.
Device details come from command-line arguments or discovery. This commit retains Poetry; the separate packaging commit migrates to uv.
server.py
- --device is now repeatable (-d ups1:/dev/ttyUSB0 -d ups2:/dev/ttyUSB1). Bare paths (/dev/ttyUSB0) auto-name from the last path component (ttyUSB0).
- Maintains {name: {ser, data, timestamp}} per UPS — each device has independent data freshness.
- GET response is now {ups_name: JBDUPS}. Accepts optional ups key in the request to return only one.
client.py
- read_data() gains ups=None parameter — pass a name to filter server-side, or omit for all.
- Always returns {ups_name: JBDUPS}.
influxdb.py
- influxdb_create_snapshot() iterates {name: JBDUPS} and tags every InfluxDB point with ups=name.
- influxdb_export() / bmspy-influxdb gain --ups to export only a specific UPS.
__init__.py
- bmspy CLI gains --ups to display only a named UPS.
- Displays each UPS under a === name === header.
tim