the BE9558 / BE18003 extension-byte case
The bytes at [7]/[11]/[15]/[19] are an annotation field (purpose still
unclear — empirically non-zero on intervals with sub-Hz or unmeasurable
freq), NOT the high byte of the peak count. The N844 fixture corpus
the original RE was done against had zero values in those bytes for
every block, so uint8 and uint16 LE were equivalent there — but on
real BE9558 Tran-drift events and BE18003 Histogram+Continuous events
the uint16 LE interpretation produced peaks up to 268 in/s and 35×
inflated PVS sums.
Cross-correlated against BW's per-interval ASCII export on:
- K558LKZU/LL1P/LL3K → 100% T/V/L/M peak match (1435 blocks each)
- T003LKZR/LL0O/LL1M → 100% T/V/L, 99.3% M (0.05 dB rounding only)
- N599LKZS/LL0L → 100% all channels
- N844 fixture corpus → 100% all channels (unchanged)
Annotations preserved on every record for future RE; the defensive
_MAX_PEAK_COUNT bound is no longer needed (uint8 maxes at 1.275 in/s,
well below any physical limit).
Synthetic regression test added using the verbatim K558LKZU.RE0H
interval-12 block.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
histogram_codec: drop _MAX_PEAK_COUNT 4096 → 2200. The old ceiling
let extension-byte blocks slip through at up to 20.48 in/s per
channel, producing 35× inflated PVS sums when first deployed to
prod. 2200 covers Normal-range full-scale (10 in/s = 2000 counts)
plus 10% headroom for quantization edge cases.
backfill_sidecars: also preserve the bw_report block alongside
review + extensions when regenerating sidecars. event_to_sidecar_dict
takes a BwAsciiReport dataclass not a dict, so for bw_report we
overlay the existing block after regen rather than passing as a kwarg.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Discovered while running the backfill on prod: certain histogram
blocks contain an undocumented extension byte format whose naive
uint16 LE interpretation yields physically impossible peak values
(150+ in/s when the device max is 10). Concrete example from
K558LKSG.3I0H block at body+7424:
bytes [6:10] = 05 79 69 00
current code: T_peak = uint16 LE = 0x7905 = 30981 → 154.9 in/s
reality: T_peak = byte[6] = 5 → 0.025 in/s (matches BW display)
The high byte (0x79 here) appears to be an extension field — possibly
"time of peak within interval" or a Histogram+Continuous sub-mode
marker. Observed across BE9558 and BE18003 units in prod data; never
appeared in the BE12844 fixture corpus the codec was originally
verified against.
Effect on prod: 26 out of 1433 blocks in this one event had inflated
peaks, plus dozens of similar events across the fleet → sum(PVS)
inflated from baseline 988 to 34501 (35x). Rolled back via the
pre-backfill snapshot before any UI exposure.
Defensive fix: bounds-check peak counts in `_decode_block`. Any
field exceeding `_MAX_PEAK_COUNT` (4096 = ~20 in/s, well past the
device's 10 in/s Normal-range FS) causes the block to be skipped
entirely. Other valid blocks in the same event still decode
correctly.
Trade-off: those skipped blocks lose their per-interval data
(peaks + frequencies). Acceptable until the extension format is
reverse-engineered — better than propagating bogus values into PVS
computations downstream.
The 24 existing tests all still pass — the fixtures used during the
original codec development don't exercise the extension-byte case.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Discovered while dry-running the backfill on prod: the waveform store
contains both BW (.AB0*/.N00) and Thor IDF (.IDFW/.IDFH) event files
side-by-side because both go through the same per-serial directory
layout. The script's `_looks_like_event_file` heuristic accepted any
3-4 char extension ending in W or H, which matched both BW and IDF.
The script then routes everything through
`event_file_io.read_blastware_file`, which rejects IDF files with
"not a Blastware file (bad header prefix)" — 3807 errors on prod
out of 7201 total events.
Thor IDF events have their own ingest path
(`WaveformStore.save_imported_idf`) and their sidecars are populated
at ingest from the paired `.IDFW.txt` ASCII report. The backfill
script has no value to add for them — there's no decoder to refresh,
and the sidecar metadata is already correct. Filter them out.
After this fix, the prod backfill should run clean: ~3392 BW events
get sidecar+h5 regen as expected; the ~3807 Thor IDF events are
silently skipped.
The proper "IDF backfill" (refresh tool_version stamp on IDF
sidecars by re-running event_to_sidecar_dict against the stored
DB row + sidecar extensions block) is a separate, narrower
follow-up — not blocking the BW backfill rollout.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The histogram-mode event body is now byte-exact decodable.
Companion to the waveform body codec — together they cover every
event file the watcher forwards. Cracked in one session via
cross-event correlation against BW's ASCII export.
The §7.6.2 spec in instantel_protocol_reference.md was structurally
correct (32-byte blocks) but the per-sample semantics were
under-documented. Cross-checking block 130 of N844L6Z8.ZR0H
against its TXT row revealed the layout perfectly:
slot[0] = 10 (constant marker)
slot[1] = T_peak_count (× 0.005 → in/s at Normal range)
slot[2] = T_halfperiod (freq_Hz = 512 / halfp)
slot[3] = V_peak_count
slot[4] = V_halfperiod
slot[5] = L_peak_count
slot[6] = L_halfperiod
slot[7] = MicL_peak_count (dB via waveform_codec.mic_count_to_db)
slot[8] = MicL_halfperiod
The `>100 Hz` sentinel is halfperiod ≤ 5 (since 512/5 = 100 Hz).
Mic dB uses the SAME formula as the waveform codec (sign × (81.94
+ 20·log10(|count|))) — they share the mic ADC calibration constant.
Block identification anchor: bytes [22:24] == 0x0000 AND
bytes [28:32] == 1e 0a 00 00. The tail signature is the most
reliable distinguisher from non-block content in the file.
Files:
minimateplus/histogram_codec.py (new) — decoder + public API
matching the waveform codec's shape:
walk_body(body) -> records
decode_histogram_body(body) -> {Tran, Vert, Long, MicL}
decode_histogram_body_full(body) -> [per-interval dicts]
half_period_to_hz, geo_count_to_ins helpers
minimateplus/event_file_io.py (modified) — read_blastware_file
now tries the waveform codec first, falls back to the histogram
codec on failure. Same output shape, same downstream pipeline.
tests/test_histogram_codec.py (new) — 24 regression locks against
the in-repo fixture corpus, byte-exact against BW ASCII export
for peaks (all 4 channels), frequencies (all 4 channels,
including >100 Hz sentinel handling), block framing, and
segment-ID accounting.
scripts/backfill_sidecars.py (modified) — the has_samples
short-circuit added in the histogram-pending era is now a
pure defensive guard. Histograms in prod will regen .h5 files
correctly on the next backfill run.
docs/histogram_codec_re_status.md (updated) — supersedes the
earlier "in progress" version with the verified format and
test-coverage summary. Notes a few non-essential fields still
open (4-byte block metadata, Geo PVS, Mic psi(L) — none of
which are needed for waveform reconstruction).
Total verified coverage: ~3,500 blocks across 5 fixtures, every
field of every block byte-exact against BW.
The watcher-forwarded histogram event corpus on prod (~10,000
events) will now produce correct .h5 sidecars on the next backfill
run. No additional changes needed to the backfill flow — the
existing tool_version-bump cascade picks them up automatically.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Captures everything learned in the 2026-05-20 session before scope
forced a pause:
- Block framing is solved: 32-byte blocks, one per histogram
interval, signature byte pattern `[22:24]=0x0000` +
`[28:32]=0x1e 0x0a 0x00 0x00` reliably identifies data blocks.
- Block count = interval count (791 blocks in N844L20G.630H for
a TXT-reported 792 intervals).
- Sample[0] = Tran peak in 0.0005 in/s/count units (verified on
one event — needs cross-event confirmation).
- Samples 1-8 → channel/metric mapping is still open. None of
the obvious layouts (peak-then-freq alternating, all-peaks-
then-all-freqs, per-channel 3-tuples) match the TXT values
across multiple blocks. Likely needs a higher-activity
fixture (current N844 corpus is all noise-floor data) to
disambiguate.
- `>100 Hz` sentinel encoding in the binary is unknown.
- 4-byte variable metadata field at block[24:28] needs
correlation work against TXT columns.
Doc mirrors the structure of docs/waveform_codec_re_status.md so
a future RE session has a familiar entry point. Includes the
suggested attack plan + the code seam where the eventual decoder
will land (minimateplus/histogram_codec.py).
The §7.6.2 spec in instantel_protocol_reference.md is structurally
correct but doesn't pin down per-sample semantics — this doc
supersedes it where they conflict on confidence level.
No code shipped on this branch. When the codec is cracked, the
plan is to land minimateplus/histogram_codec.py + wire into
event_file_io.read_blastware_file() + remove the has_samples
short-circuit from scripts/backfill_sidecars.py.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Fixes a data-loss bug discovered while dry-running the backfill against
the prod store.
Symptom: every histogram event in the store has its body decoded by
read_blastware_file → codec returns None → samples = empty dict →
``ev.peak_values = _peaks_from_samples(empty)`` returns
``PeakValues(0, 0, 0, 0, 0)`` (NOT None). The backfill script's
existing "seed from DB row when peak_values is None" branch then
correctly *skips* the seeding, and the all-zeros PeakValues flows into
``db.insert_events()``'s UPSERT path, OVERWRITING the existing good DB
peak values for that event (which were populated from the paired BW
ASCII report at ingest).
Net effect: running the backfill on prod would have wiped the PPV /
mic / vector-sum columns for ~10,000 histogram events.
Fix: only compute peaks-from-samples when there are actually samples.
For events the codec couldn't decode (histogram-mode bodies, until
the §7.6.2 histogram codec is wired in), leave peak_values=None as
the "we don't know" signal. Downstream consumers:
- backfill_sidecars.py — its existing ``if ev.peak_values is None:``
branch (line 243) seeds from the DB row, preserving the real
BW-report peaks across the regen.
- WaveformStore.save_imported_bw — apply_report_to_event overlays
peaks from the paired BW ASCII report when one was uploaded.
Histogram imports without a paired report end up with NULL peaks
in the DB, which is correct (better than zeros — clearly says
"no peak data available" rather than "peaks are exactly zero").
Updated the existing synthetic-event round-trip test to expect
peak_values=None for the no-real-body case, which is the truth now.
The 7 fixture-corpus regression tests for real BW waveforms continue
to pass — those have decodable samples, so peak_values is still
populated from the codec output as before.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Discovered while dry-running the backfill on the prod store: ~10,000
of ~10,059 events are histogram-mode (filename extension `*H`), and
the waveform-body codec wired in via the previous commit doesn't
handle histogram-mode bodies — only the waveform-mode codec at
§7.6.1 is implemented; the histogram-mode codec at §7.6.2 of the
protocol reference is documented but no Python implementation
exists yet.
Without this guard, every histogram event's .h5 file would be
*replaced* with an empty one — strictly worse than today's
broken-int16-LE .h5 because any downstream viewer expecting
non-empty sample arrays would now error out instead of just
rendering wrong values.
Fix: after the decoder runs, check whether any channel has samples.
If not, skip the .h5 write entirely. The sidecar still regenerates
(refreshing the tool_version stamp and any peaks/project info from
the DB row), but the existing .h5 is left untouched.
This is a *temporary* gate. When the histogram codec lands (next
branch: `feat/wire-histogram-codec`), the has_samples check can be
removed and the backfill will then correctly regenerate all .h5
files, histogram and waveform alike.
Observed effect (dry-run on prod store, 10,059 events):
- waveform events (~5%): "[DRY ] would write … + .h5 (would (re)write)"
- histogram events (~95%): "[DRY ] would write … + .h5 (skipped-empty-samples)"
- sidecar tool_version bump succeeds for both
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two coupled changes that close the rollout gap left by the
read_blastware_file codec wiring:
1. minimateplus/event_file_io.py: bump TOOL_VERSION from 0.16.1 to
0.20.0. This is the version stamp the backfill script reads from
each sidecar's source.tool_version field to detect "this sidecar
was written before the current decoder shipped, regenerate it."
Bumping past every value baked into existing prod sidecars flags
them all as stale on the next backfill run — which is exactly what
we want, since every pre-codec-wiring sidecar was written by the
retracted int16-LE decoder.
2. scripts/backfill_sidecars.py: when the sidecar is being
regenerated this iteration (sha mismatch, tool_version too old,
or --force), also regenerate the .h5. Previously the .h5 logic
only rewrote when --force was passed or the file was missing —
so a tool_version-driven sidecar regen left the broken .h5 in
place forever. Added a `sidecar_stale` boolean to track the
"we're rewriting the sidecar this iteration" state and wired it
into the h5 need-rewrite check.
Path coverage (verified by trace):
- sidecar missing → both regen
- --force → both regen
- sha mismatch → both regen
- tool_ver too old → both regen (THE post-codec-wiring case)
- everything OK → skip iteration entirely (h5 untouched)
Operator review state (review.false_trigger, reviewer, notes) and
the sidecar's extensions block are preserved across regen by the
existing read-existing-sidecar / pass-into-event_to_sidecar_dict
path — unchanged from prior behavior.
Deploy procedure (on prod):
1. Pull this change + the read_blastware_file codec wiring.
2. `python scripts/backfill_sidecars.py --dry-run` to preview.
Every sidecar with source.tool_version<0.20.0 will show as
"would (re)write".
3. Run for real (drop --dry-run). Expect every pre-fix event
to regen. Big stores may take a while.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
`read_blastware_file()` was still calling `_decode_samples_4ch_int16_le`
(the retracted int16-LE-interleaved hypothesis) on the body bytes,
producing ±32K noise on every channel of every BW file read from disk.
This was the path watcher-forwarded events take into the system
(via the import endpoint → save_imported_bw → read_blastware_file,
since the watcher doesn't ship A5 frames), so every .h5 sidecar
generated for a forwarded event has been wrong since the feature
shipped.
The fix is mechanical: pass the body bytes straight to
`waveform_codec.decode_waveform_v2()` and run the result through
`decoded_to_adc_counts()` for the 16x geo scaling. The body already
starts with the codec's exact 7-byte preamble `00 02 00 [Tran[0] BE]
[Tran[1] BE]` — confirmed by `body[:3].hex()` across all 9 fixture
events. No body-slice adjustment needed.
If the codec returns None (truncated/malformed file, synthetic test
input with no real waveform), fall back to empty channels with a log
warning. The rest of the event (timestamp, waveform_key, project
strings, sensor_location, peaks-from-samples=0) is still recoverable.
Verified against the bundled fixture corpus:
V70 Tran/Vert/Long 3328/3328 sample-sets match .TXT ground truth
within the 0.005 in/s display quantum, every row
6S0/RG0/AB0/470 (5-8-26) 3328/2304/1280/1280 samples; Vert PPVs
match BW's own report within 0.02 in/s
JQ0 3328 samples, Vert PPV 3.384 vs BW 3.465
SP0/SS0/SV0 (loud events) 3072–3328 samples; known walker
tail-truncation 1–7 samples per channel, samples reached are
byte-exact
Existing `test_read_blastware_file_round_trip` (synthetic empty event)
continues to pass thanks to the None-fallback. Codec verify scripts
(`analysis/verify_quiet_bundle.py`, `analysis/verify_full_decode.py`)
re-run unchanged.
Added two regression-lock tests in tests/test_event_file_io.py:
- test_read_blastware_file_decodes_via_codec[6 fixtures]
— verifies sample count + Vert PPV per fixture
- test_read_blastware_file_v70_samples_match_txt_truth
— verifies every one of V70's 3328 sample-sets across Tran/Vert/Long
matches the .TXT ground truth row-by-row within 0.003 in/s
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
## v0.19.0 — 2026-05-20
The "device-family separation" release. Tightens the boundary between Series III (MiniMate Plus / Blastware) and Series IV (Micromate / Thor) so the UI and storage layer dispatch deterministically by family instead of sniffing filename extensions or magnitude heuristics.
### Added — Phase 1: `device_family` column on `events`
- **`events.device_family TEXT`** — new column carrying `"series3"` or `"series4"`. Populated by every import path (`/db/import/blastware_file`, `/db/import/idf_file`, ACH server, BW CLI, sidecar backfill script). Returned through `/db/events` since `query_events` uses `SELECT *`.
- **Self-applying migration** — on startup, `ALTER TABLE ... ADD COLUMN` lands the new column; a follow-on `UPDATE` backfills existing rows from the binary filename extension (`.IDFH`/`.IDFW` → `series4`, everything else → `series3`). No manual SQL needed.
- **UPSERT preserves family** — re-imports without an explicit family don't blank existing rows (`COALESCE(?, device_family)`).
- **UI dispatches on the column** — `sfm_webapp.html` events-table mic formatter now branches on `ev.device_family === 'series4'` (Thor stores native dB(L); BW stores psi). Modal uses `source.kind === 'idf-import'` from the sidecar (sidecars don't carry the DB column). Source-files section labels changed from "BW filename / BW filesize / BW sha256" to format-neutral "Event file / File size / File sha256".
### Added — Phase 2: `micromate/` package alongside `minimateplus/`
- **`micromate/`** — new sibling package for the Thor / Micromate Series IV device. Currently scoped to offline-file ingest; live-device support (TCP transport, framing, protocol, client) will land here when reverse-engineering happens.
- `micromate/idf_ascii_report.py` — moved from `sfm/idf_ascii_report.py`. No behaviour change.
- `micromate/models.py` — typed `IdfReport`, `IdfEvent`, `IdfPeaks`, `IdfProjectInfo`, `IdfSensorCheck`. Stores mic in native `mic_pspl_dbl` (dB(L)) instead of the pseudo-psi shoehorn that the BW-shaped model uses. `IdfEvent.from_report()` constructs from a parsed dict + filename; `IdfEvent.to_minimateplus_event(waveform_key)` bridges to the existing sidecar / DB-insert machinery.
- `micromate/idf_file.py` — placeholder for the binary codec (`.IDFH` / `.IDFW`). Stubbed `read_idf_file()` raises `NotImplementedError`; documents the planned reverse-engineering path.
- **`WaveformStore.save_imported_idf`** refactored to use the native `IdfEvent` and bridge at the SQL-insert boundary. Cleaner separation of "parse a Thor event" (in `micromate/`) from "store it on disk + write a sidecar" (in `sfm/waveform_store.py`).
- **Tests** — `tests/test_idf_ascii_report.py` imports updated to `micromate.idf_ascii_report`. All 1,014 example-data sidecars round-trip through `IdfEvent.from_report()` without errors.
### Companion releases
- **thor-watcher** unaffected — it talks to the relay over HTTP only. No version bump needed.
- **terra-view** unaffected today; can use `device_family` in its event-detail rendering when convenient.
---
## v0.18.0 — 2026-05-19
The "Thor / Series IV ingest adapter" release. Seismo-relay can now accept event files from Instantel Micromate Series IV (Thor) units alongside the existing MiniMate Plus (Series III) Blastware pipeline.
### Added — Thor (Series IV) IDF ingest
- **`POST /db/import/idf_file`** (`sfm/server.py`) — multipart upload endpoint for `.IDFH` (histogram) and `.IDFW` (waveform) event files plus their `.IDFH.txt` / `.IDFW.txt` ASCII sidecars. Mirrors the shape of `/db/import/blastware_file`: pairing by filename, optional `serial` query hint, per-file outcome reporting.
- **`sfm/idf_ascii_report.py`** — parser for Thor's TXT sidecars (verified against 1,014 real-world samples). Extracts device-authoritative PPV, ZC Freq, Peak Vector Sum, Mic PSPL, calibration date, firmware version, sensor self-check results, and project/client/operator strings.
- **`WaveformStore.save_imported_idf()`** (`sfm/waveform_store.py`) — stores Thor binaries verbatim in `<root>/<serial>/<filename>`, writes a `.sfm.json` sidecar with `source.kind = "idf-import"` and the full parsed report under `extensions.idf_report`. Reuses the existing `events` table — Thor events dedupe on (serial, timestamp) and surface in `/db/events` alongside BW events.
- **`tests/test_idf_ascii_report.py`** — parser tests against the `thor-watcher/example-data/` corpus.
### Changed
- `event_to_sidecar_dict()` (`minimateplus/event_file_io.py`) allow-list for `source_kind` now includes `"idf-import"` so the existing sidecar machinery can carry Thor imports.
- Bumped `pyproject.toml` version to `0.18.0`.
### Companion release
This release ships alongside **thor-watcher v0.3.0**, which adds the SFM forwarder that targets the new `/db/import/idf_file` endpoint. Operators flip the switch in thor-watcher's new "SFM Forward" Settings tab; events POST to seismo-relay just like the series3-watcher BW forwarder does today.
Tighten the Series III / Series IV boundary so UI and storage dispatch
on a clean signal instead of sniffing filenames or applying magnitude
heuristics.
Phase 1 — events.device_family column ("series3" | "series4"):
self-applying migration with filename-based backfill of existing rows
(1,132 backfilled on prod 2026-05-20); plumbed through every import
path (BW endpoint, IDF endpoint, ACH server, BW CLI, sidecar
backfill); UPSERT preserves via COALESCE; UI dispatches on it.
Phase 2 — extract micromate/ package alongside minimateplus/:
native IdfEvent / IdfReport / IdfPeaks / IdfProjectInfo /
IdfSensorCheck (mic in dB(L), not pseudo-psi); moved
idf_ascii_report.py from sfm/ to micromate/; refactored
save_imported_idf to use IdfEvent and bridge to minimateplus.Event at
the SQL-insert boundary; idf_file.py stub for the future binary codec.
Phase 3 prep — docs/idf_protocol_reference.md captures the two
observed Thor binary header signatures (1,012 newer-firmware files vs
2 old files whose layout is byte-for-byte BW-STRT-compatible), file-size
hints suggesting int8 sample encoding, open questions in dependency
order, and a concrete first-session plan for cracking the codec.
Also rolled in the v0.18.1 hotfixes that motivated this work:
- idf_ascii_report parser now handles "<0.005 in/s" (below-threshold)
and "N/A" markers without leaving raw strings in numeric DB columns.
- sfm_webapp.html: defensive _ppvFmt / mic formatter so future
data-shape drift can't kill the whole events table render.
All 1,014 example-data sidecars round-trip through the new package.
See CHANGELOG.md for full notes.
Reviewed-on: #21
## v0.17.0 — 2026-05-17
The "field rescue + DB management" release. Hardened against units that are stuck in a runaway call-home loop, and added an operator-facing path for purging bogus events that those same units dump into the DB before recovery. All work in this release was driven by the BE9558H incident (full incident log + recovery procedure at `docs/runbooks/wedged_unit_recovery.md`).
### Added — wedged-unit recovery toolkit
A toolkit for breaking the call-home loop on a misbehaving unit whose firmware is too busy to keep up with normal request/response handshakes. Tested in production against BE9558H (16 May 2026) — a unit with a stuck-triggered Long-axis geophone that had been call-homing the office BW ACH server every 30 seconds for hours. Endpoints layered from "single attempt" to "siege mode" to suit different contention levels:
- **`GET /device/events/storage_range`** — SUB 0x06 probe. POLL + one read; ~2s. Returns first/last event keys and an `is_empty` flag. Use to triage whether a unit has stored events without invoking the slow `count_events()` 1E/1F chain (which choked on BE9558H's corrupted event chain).
- **`GET /device/events/index`** — SUB 0x08 probe. POLL + one read; ~2s. Returns the lifetime event counter (does NOT decrement on erase — use `storage_range` for "right now" state).
- **`POST /device/events/erase`** — full erase sequence `0xA3 → 0x1C → 0x06 → 0xA2` (confirmed 2026-04-11, see the protocol reference). Resets event keys to `0x01110000`. Caller's responsibility to disable ACH first if the underlying trigger condition will re-fill the buffer.
- **`POST /device/rescue`** — one TCP session, short connect+recv timeouts: POLL → disable ACH (compliance config write) → erase events → close. Designed for race-loop usage when the device is busy in another session. 503 on connect-refused, 502 on protocol failure, 200 on full sequence success.
- **`POST /device/stop_monitoring_blind`** — fire-and-forget Stop Monitoring (SUB 0x97), TCP-only. Dumps `SESSION_RESET + POLL_PROBE + SESSION_RESET + POLL_DATA + 0x97 × repeat` and closes without reading any S3 response. The full POLL preamble is required — write commands without it are silently ignored by the device's protocol parser (false-positive surface area that bit the first version of this endpoint). Use when the device's firmware can't keep up with full request/response but might process inbound bytes at its own pace.
- **`POST /device/stop_monitoring_spam`** — server-side hammer loop, duration-bounded. Open TCP → write the same blind payload → close → repeat as fast as possible until `duration_s` elapses. Configurable `connect_timeout` (default 500ms) and `repeat` (frames per session). Reports `sent_ok`, `connect_failed`, `write_failed`, `rate_attempts_per_s`. Clamped to 5min duration.
- **`POST /device/stop_monitoring_slow_drip`** — opposite of spam. Open ONE TCP session, drip the wake handshake + stop frames at `interval_s` (default 3s) for `duration_s` (default 120s, max 10min). Each drip is ~23 bytes — well under any UART FIFO size. Opportunistically drains any inbound bytes the device sends back; `bytes_received > 0` in the response strongly suggests the device has started talking and the session is healthy. **This is the endpoint that saved BE9558H.** Spam mode had been overrunning the device's UART FIFO; slow drip stayed under it.
- **Six rescue scripts** under `scripts/` — thin bash wrappers around the endpoints, default `SFM_BASE_URL=http://localhost:8200` (direct, not via Terra-View proxy whose 60s timeout would cut off the longer endpoints):
- `rescue_device.sh` — race-loop wrapper for `/device/rescue`
- `blind_stop.sh` — race-loop wrapper for `/device/stop_monitoring_blind`
- `spam_stop.sh` — single-call burst hammer
- `slow_drip.sh` — single-call held-session drip
- `watch_unit.sh` — passive periodic reachability check (every N min, logs to file), useful for unattended overnight monitoring of a wedged unit
- **`docs/runbooks/wedged_unit_recovery.md`** — symptoms, quick-reference recovery procedure, the modem-layer mechanism (Sierra Wireless serial-port mode-flipping is the real failure mode — not the device firmware), and a table of "why simpler approaches don't work" so the next incident skips the dead ends.
### Added — operator event DB management
Endpoints powering Terra-View's new `/admin/events` page (v0.12.0). Designed for purging bogus events from a unit that's been forwarding them in bulk (e.g. a stuck-triggered seismograph dumping hundreds of junk events before it's recovered).
- **`DELETE /db/events/{event_id}`** — hard-delete one event row. Also unlinks the associated blastware binary (`.AB0*`), `.a5.pkl`, `.sfm.json` sidecar, and `.h5` clean-waveform files via the WaveformStore. Returns the per-file removal status. 404 if the event doesn't exist.
- **`POST /db/events/delete_bulk`** — filter-based or id-list-based bulk delete with safety rails:
- Filters (`serial`, `from_dt`, `to_dt`, `false_trigger`) combine with AND; same semantics as `GET /db/events`. `ids` is an additional inclusion list. Refuses to run with no filters (would wipe the whole table — raises 422).
- `confirm` must be `true` to actually delete. Otherwise returns a dry-run summary (`status: "dry_run"`, `matched: N`, `sample_serials: [...]`).
- `max_rows` (default 10,000) caps how many rows can be deleted by-filter in one call. If exceeded, returns `status: "too_many"` with a hint to narrow or raise the cap. Bypassed when only `ids` is supplied.
- **`_cleanup_event_files(row)`** helper in `sfm/server.py` — best-effort `unlink()` of all four sidecar paths derived from the row's `blastware_filename`. Logged at WARN if a path exists but unlink fails; the DB row deletion still proceeds.
- **`SeismoDb.delete_event(id)` and `SeismoDb.delete_events_bulk(...)`** in `sfm/database.py` — both return the deleted row dict(s) so callers can do file cleanup. `delete_events_bulk` raises `ValueError` if no filters are supplied.
### Changed
- **Default protocol recv timeout dropped from 30s → 10s** in `_build_client()`. The unit usually responds in well under a second over cellular; 10s leaves comfortable headroom for retransmits while failing reasonably fast when a unit is wedged. The two endpoints that perform full 5A waveform downloads still pass `timeout=120.0` explicitly so multi-minute event transfers are unaffected.
- **`_build_client()` now accepts an optional `connect_timeout`** (TCP-only) so rescue / race-loop endpoints can fail fast on busy modems without affecting the protocol-level recv timeout.
### Fixed
- **`GET /device/monitor/status` returned HTTP 500 + uncaught traceback when the device was unresponsive**. The retry-on-`Exception` inner block let the second `client.poll()`'s `ProtocolError` propagate out of the handler. Now wrapped in proper try/except — returns 502 with `{"detail": "Protocol error: No S3 frame received within 10.0s ..."}` on timeout, 502 on connection errors, 500 only for genuinely unexpected exceptions.
### Migration
No schema changes. No data migration required.
If you've been running a previous version against a wedged unit and accumulated bogus events, the new `/admin/events` page in Terra-View v0.12.0 (or direct `POST /db/events/delete_bulk` with `confirm: true`) is the cleanup tool. Watcher state on the upstream DL2 PC does NOT need separate cleaning — the watcher's `sfm_forwarded.json` keys on file sha256 and won't re-forward the same files.
### Pairing
This release pairs with **Terra-View v0.12.0**, which adds the `/admin/events` UI that consumes the new bulk-delete endpoints, the bulk false-trigger flagging on `/unit/{id}`, and the field-deployment workflow that uses the same `series3-watcher` → SFM ingest path as before.
---
## v0.16.1 — 2026-05-14
### Fixed
- **`record_type` always "Waveform" for forwarded events.** `read_blastware_file()` hardcoded `ev.record_type = "Waveform"` regardless of the file's actual type. The watcher-forward pipeline (the main BW ACH ingest path) compounds this by parsing files from a tmp path with a `.bw` suffix, so even a filename-based fallback inside the parser still wouldn't see the original extension. Now:
1. New `derive_record_type_from_filename(filename)` helper in `minimateplus/event_file_io.py` derives the type from the LAST character of the filename's extension (V10.72+ AB0T scheme: `H`=Histogram, `W`=Waveform, `M`=Manual, `E`=Event, `C`=Combo). Falls back to `"Waveform"` for old S338 firmware (3-char extensions ending in `0`) and any unrecognized suffix.
2. `read_blastware_file()` now calls the helper with its `path.name` so direct callers (the `--dry-run` path in `scripts/import_bw.py`, tests, ad-hoc scripts) get the right value automatically.
3. `WaveformStore.save_imported_bw()` overrides `ev.record_type` with the **original** filename's derived type after parsing (the tmp file inside the parser doesn't carry the original extension). This is the path the live watcher-forwarder hits, so the DB column now reflects the actual event type going forward.
Events ingested before this fix are stuck with `record_type="Waveform"` in the DB; a one-off backfill (`UPDATE events SET record_type = ... WHERE blastware_filename LIKE '%H'`) would fix them retroactively if desired. Terra-view's event modal also derives client-side from the filename, so the UI already shows the correct type for old events even without the backfill.
---
- Created a comprehensive runbook (`wedged_unit_recovery.md`) detailing the recovery process for units stuck in a call-home loop, including symptoms, recovery steps, and explanations of the failure mode.
- Added `blind_stop.sh` script to send stop-monitoring commands in a tight loop for unresponsive devices.
- Introduced `rescue_device.sh` script to disable Auto Call Home and erase events from a busy device.
- Implemented `slow_drip.sh` script to send stop-monitoring frames at a slow rate to prevent UART overrun.
- Developed `spam_stop.sh` script to rapidly send stop-monitoring commands to a device.
- Created `watch_unit.sh` script for passive monitoring of device reachability, logging results over time.
Pre-v0.16.1 (commit aac1c8e), every event ingested through
read_blastware_file got record_type="Waveform" regardless of actual
type because the field was hardcoded. New ingests derive correctly
from the AB0T filename scheme (H/W/M/E/C). Existing rows still hold
the wrong value.
This script walks the events table, derives the correct record_type
from each row's blastware_filename, and bulk-updates rows that differ.
Idempotent + dry-run by default.
Usage:
python -m scripts.backfill_record_type --db bridges/captures/seismo_relay.db
python -m scripts.backfill_record_type --db bridges/captures/seismo_relay.db --apply
Terra-view's event-detail modal already derives the record_type
client-side from the filename for display, so operators see the
correct type in the UI even before this backfill runs. This script
brings the DB column in line with what the UI is already showing —
matters for reporting and any downstream consumer that reads the
column directly.
The BW ACH ingest path was inserting every event with
record_type="Waveform" regardless of the actual type because
read_blastware_file() had `ev.record_type = "Waveform"` hardcoded, and
the live watcher-forward path parses files from a tmp path (suffix
".bw") that doesn't carry the original extension.
V10.72+ MiniMate Plus firmware encodes the event type as the last
character of the AB0T extension scheme (H=Histogram, W=Waveform,
M=Manual, E=Event, C=Combo). This change:
1. Adds derive_record_type_from_filename() public helper in
minimateplus/event_file_io.py
2. Uses it inside read_blastware_file() so direct callers (the
--dry-run path of scripts/import_bw.py, tests, ad-hoc scripts)
get correct types automatically
3. Overrides ev.record_type in WaveformStore.save_imported_bw()
using the ORIGINAL filename (source_path.name) — required
because the parser sees only the tmp file
Old S338 firmware (3-char extensions ending in `0`) and any
unrecognized suffix fall back to "Waveform".
Existing DB rows ingested before this fix are stuck with
record_type="Waveform" — a one-off SQL backfill would fix them
retroactively if desired. Terra-view's event modal also derives
client-side from the filename, so the UI already shows the correct
type for old events even without the backfill.
Version bumped to 0.16.1 in pyproject.toml, event_file_io.py
TOOL_VERSION, sfm/server.py FastAPI version, and CHANGELOG.md.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The "BW ACH ingestion" release. Paired with series3-watcher v1.5.0,
every Blastware ACH event (binary + _ASCII.TXT report) lands in
SeismoDb with device-authoritative peaks, project metadata, sensor
self-check, and ZC/Time-of-Peak data — without depending on the
still-undecoded waveform body codec.
Bumps pyproject.toml + minimateplus/event_file_io.py TOOL_VERSION
to 0.16.0. README banner + CHANGELOG entry summarise the work
that landed across commits cdfe4ad..f83993a on this branch.
The series3-watcher v1.5.0 fix taught the WATCHER to look for BW
ACH's _ASCII.TXT report alongside each binary. But the SFM
SERVER's import endpoint only knew about the legacy <binary>.TXT
naming when building its TXT lookup table.
Effect: even though the watcher correctly shipped both files in
the multipart POST (and logged "+ <name>_ASCII.TXT attached"),
the server's reports dict was keyed on the wrong name, so
report_bytes resolved to None for every event. Without the
report, save_imported_bw fell back to broken-codec peak values
and no project info — exactly the same symptom as before the
watcher fix landed, just for a different reason.
Fix: when stripping the ".TXT" suffix, also recognise the
"_ASCII" trailer and reconstruct the binary's filename by
converting the last "_" back to ".". Register the report under
BOTH possible binary names so the subsequent lookup matches
whichever convention the operator's BW installation uses.
ACH convention (Blastware ACH):
binary T003L2G6.0E0H + report T003L2G6_0E0H_ASCII.TXT ✅
Manual export (operator clicks Save As Text in BW):
binary M529LK44.AB0 + report M529LK44.AB0.TXT ✅
Both for same event (e.g. ACH + operator manual save):
register under both names; binary lookup wins ✅
Smoke-tested against the four real fixture filenames in the
project archive. Full SFM suite still 62 pass.
For the user's situation: pull, restart, and the NEXT re-forward
pass (after deleting watcher state file again if needed) will
hit this code path, parse the report correctly, apply the
overlay onto the Event, and the upsert path will land
authoritative peak values + project info in the DB.
Two compounding bugs caused forwarded events to land in the DB with
broken-codec peak values (~10 in/s saturation on every channel) and
no project info, even when the watcher correctly paired a BW ASCII
report with the binary.
Bug 1: save_imported_bw built the sidecar JSON with the report's
authoritative peak / project values via event_to_sidecar_dict(
bw_report=...), but never overlaid those onto the in-memory Event
that flows to db.insert_events(). So the DB row got peak_values
from read_blastware_file()._peaks_from_samples() — which runs the
still-undecoded waveform body codec assuming raw int16 LE and
produces ±32K-shaped noise (= ±10 in/s at Normal range) regardless
of the actual signal. The sidecar JSON had the truth but the DB
columns (which the webapp queries for fast filter/sort) lied.
Bug 2: insert_events' IntegrityError handler only refreshed the
filename/filesize/a5_pickle/sidecar columns when a duplicate
(serial, timestamp) was seen. Peak values, project info,
sample_rate, record_type stayed locked in at whatever the FIRST
insert wrote. So even after Bug 1 was fixed, the historical
events in the DB (already inserted with broken-codec peaks) would
never get their values corrected, because a re-forward would just
hit IntegrityError and skip the field refresh.
Fix 1 (minimateplus/event_file_io.py + sfm/waveform_store.py):
- New apply_report_to_event(event, report) helper folds the BW
report's device-authoritative fields onto the Event in-place:
per-channel PPV, peak vector sum, mic PSPL→psi, project /
client / operator / sensor_location, sample_rate, record_time.
- save_imported_bw() calls the helper right after parsing the
report. The Event that flows to insert_events() now carries
correct values.
Fix 2 (sfm/database.py):
- insert_events()'s IntegrityError UPDATE now refreshes every
device-authoritative column from the new data: tran_ppv,
vert_ppv, long_ppv, peak_vector_sum, mic_ppv, project, client,
operator, sensor_location, sample_rate, record_type, plus
the existing filename/filesize/a5_pickle/sidecar fields.
- Preserves: id, waveform_key, session_id, created_at (immutable
/ FK fields), and false_trigger (operator review state).
End-to-end simulation verified:
- Step 1: import without report → DB has ±10 in/s peaks, no project
- Step 2: re-import WITH report → upsert path fires, DB now has
device-authoritative 0.005 in/s peaks + sensor_location
- Step 3: operator sets false_trigger=1, re-import again → flag
preserved, peaks remain correct
For the user's situation: deleting the watcher state file forces a
re-forward of all events. Each re-forward now pairs with its
_ASCII.TXT, applies the report onto the Event, and the upsert
refreshes the DB row. No DB nuke needed.
Full SFM suite: 62 passed, 44 skipped.
Previous query_units() only joined on ach_sessions, which is created
exclusively by the live ACH server. The BW-importer path
(/db/import/blastware_file → WaveformStore.save_imported_bw →
SeismoDb.insert_events) populates `events` but never creates an
ach_sessions row. Consequence: every serial whose events flowed in
through the series3-watcher forwarder was invisible to
/db/units (and therefore to the SFM webapp's fleet overview / units
list), even though the events were correctly populated in the
events table with proper serial attribution.
Rewrite query_units() to aggregate from BOTH tables and union the
serials:
- total_events / last_event_at come from `events` (every ingest path)
- last_session_at / total_monitor_entries / total_sessions
come from `ach_sessions` (ACH-only),
0 when no sessions exist for the serial
- last_seen = max(last_event_at, last_session_at)
Verified on the user's actual prod DB after the
repair_unknown_serials run: /db/units now returns 24 serials instead
of 2. All 3,257 watcher-forwarded events become visible in the
fleet overview without any further DB surgery.
The /db/import/blastware_file endpoint was bucketing every
forwarded event into serial='UNKNOWN' in the DB. WaveformStore
correctly decoded the serial from the BW filename and saved
files to <store>/<serial>/<filename> (e.g.
.../BE17353/S353L5KC.DR0H.h5), but the endpoint code called
db.insert_events(serial=_serial_from_event(ev)) — and
_serial_from_event was a stub that always returned None,
falling back to "UNKNOWN".
Effect on the user's prod server: 3,039 events forwarded across
24 distinct units, ALL inserted under serial='UNKNOWN'. The
on-disk waveform store + sidecars + HDF5s were fine, but the
SFM webapp's /db/units only showed the two original manually-
uploaded serials because every forwarded row had its serial
column zeroed to UNKNOWN.
Fix:
- WaveformStore.save_imported_bw() now surfaces the decoded
serial on the returned `rec` dict (rec["serial"]).
- The import endpoint uses rec["serial"] as the authoritative
fallback when the operator hasn't supplied a serial_hint query
parameter. Order of precedence:
query string `serial` → rec["serial"] → _serial_from_event(ev) → "UNKNOWN"
- Response payload now includes `serial` per file so the watcher
log lines (or any future caller) can see which unit each event
was attributed to.
Recovery for existing DB rows:
scripts/repair_unknown_serials.py walks the events table looking
for rows with serial='UNKNOWN' and re-attributes each one to the
serial decoded from blastware_filename. Updates the row in place
unless the target (serial, timestamp) already has a row, in which
case the UNKNOWN duplicate is deleted. Idempotent. Default
dry-run; pass --apply to commit.
Verified on the user's actual DB (dry-run):
UNKNOWN rows scanned: 3039
Updated to real serial: 2602
Deleted (duplicate of an
already-correct row): 437
Unresolved (bad filename): 0
After running the repair, /db/units will show all 24 units
correctly populated.
The four operator-supplied note fields in BW's Compliance Setup →
Notes tab (Project / Client / User Name / Seis Loc) have
USER-EDITABLE LABELS — an operator can rename them in BW's UI to
"Building:", "Site Address:", "Inspector:", or anything else, and
the ASCII export writes those literal labels verbatim. The
previous label-normalisation map approach (just added in commit
6a7e8c6) was fragile: it could only match label spellings we'd
enumerated in advance. An operator using "Site:" instead of
"Seis Loc:" would have their sensor location silently dropped.
What IS reliable: BW always writes the 4 user-notes lines
contiguously, in the same order, between the "Units :" line and
the "Geo Range :" line of the export. So parse them by POSITION:
position 1 → project
position 2 → client
position 3 → operator
position 4 → sensor_location
The original labels BW wrote are preserved in a new
`BwAsciiReport.user_note_labels` dict (canonical slot → literal
label string) so terra-view can render them as the operator named
them.
Removes the `_OPERATOR_LABEL_MAP` / `_normalise_label_for_lookup`
helpers and the elif-by-normalised-label branch in `parse_report`.
Replaces with a small state machine that flips on the "Units" line
and flips off on the "Geo Range" line.
Tests:
- Default-label fixtures (waveform + histogram) still populate
correctly, with operator's labels captured.
- Synthetic custom-labelled exports ("Building:" / "Site Address:" /
etc.) populate the right slots by position.
- Histogram-specific "Seis. Location:" works.
- Lines outside the Units→Geo Range range are ignored even if
they look like user notes (defensive against malformed exports).
- Partial blocks (fewer than 4 lines) leave later slots None.
- Extra lines beyond 4 are dropped (5th slot doesn't exist).
26 tests in test_bw_ascii_report.py (was 33; net drop reflects
parametrised label tests collapsed into 6 focused position tests).
Full SFM suite: 62 passed, 44 skipped.
Pairs with series3-watcher v1.5.0 which fixes the filename pairing
so the report reaches this parser in the first place.
Blastware writes the operator-supplied fields with different label
spellings across firmware versions and recording modes — most
notably "Seis. Location" on histogram exports vs "Seis Loc:" on
waveform exports. Previous parser only matched the latter, so
every histogram event silently lost its sensor_location field.
Replace the four hardcoded `key.rstrip(":") == "X"` branches with
a single `_OPERATOR_LABEL_MAP` dispatch table keyed by normalised
label (lowercase, trailing colon/period stripped, internal
whitespace collapsed). Adds these variants on day 1:
project: "Project:" / "Project"
client: "Client:" / "Client"
operator: "User Name:" / "User Name"
sensor_location: "Seis Loc:" / "Seis. Location" / "Seis Location"
/ "Sensor Location" / "Seis Loc"
To absorb future BW label drift, add a one-line dict entry — no
new elif branch.
14 new tests cover:
- Each label variant routes to the correct field (parametrised)
- Case-insensitive matching ("seis loc" / "SEIS LOC" / "SeIs LoC")
- Whitespace-collapse ("Seis Loc" with double-space)
- End-to-end parse of a real histogram fixture from
example-events/histogram/ — sensor_location ('Loc #1 - 2652 Hepner...')
populates correctly even though the file uses "Seis. Location"
Total bw_ascii_report tests: 19 → 33. Full SFM suite still green
(69 passed, 44 skipped — pre-existing skips for h5py-dep tests).
Pairs with series3-watcher v1.5.4 (which fixes the filename pairing
so histograms actually reach this parser in the first place).
Blastware's ACH writes a per-event ASCII report (.TXT) alongside each
event binary, containing the rich derived per-channel fields BW
computes (PPV, ZC Freq, Time of Peak, Peak Acceleration, Peak
Displacement, Peak Vector Sum + time, sensor self-check Pass/Fail,
monitor-log timestamps). None of this lives in the BW binary itself.
When the watcher daemon forwards both files to /db/import/blastware_file
in one multipart POST, we now:
- Pair binaries with their .TXT partners by filename match
- Parse the report into a structured BwAsciiReport
- Land the rich fields in a new top-level `bw_report` block of the
sidecar JSON
- Overlay the report's peaks/project_info/timestamp/sample_rate/
record_time/total_samples/pretrig_samples onto the canonical
sidecar fields (the report values are device-authoritative; the
BW-binary STRT-derived values had bugs like reading the 0x46
record-type marker as rectime)
This unblocks the monthly-summary review workflow — events become
sortable/filterable by peak, location, project, etc. — without
depending on the still-undecoded waveform body codec.
test: add regression tests for v0.14.x SUB 5A protocol fixes
refactor(logging): change warning logs to debug for less verbosity in write_blastware_file
### Added
- **Layered event storage architecture.** Each event now lands as four
files in the per-serial waveform store, each with a clear role:
- `<filename>` — the Blastware-readable binary (BW file). Untouched.
- `<filename>.a5.pkl` — the raw 5A frames (regenerative source).
- `<filename>.h5` — clean per-channel waveform arrays in physical
units (in/s for geo, psi for mic) plus event metadata (HDF5 with
gzip compression). This is the canonical format for downstream
analysis tools.
- `<filename>.sfm.json` — the modern review/metadata sidecar (peaks,
project, source provenance, review state, extensions).
SQLite (`seismo_relay.db`) is the searchable index over all four.
- **Plot-ready waveform JSON (`sfm.plot.v1`).** The `/device/event/{idx}/waveform`
and `/db/events/{id}/waveform.json` endpoints now return samples in
physical units with explicit time-axis metadata, peak markers, and
per-channel unit hints — no more guessing the ADC-to-velocity scale
client-side. The webapp waveform viewer was rewritten to consume
this shape.
- **In-app waveform viewer accuracy fix.** The standalone SFM webapp
viewer was scaling geophone amplitudes by `geoAdcScale / 32767`
(≈ 6.206 / 32767), where `geoAdcScale = 6.206053` is the device's
*in/s per V* hardware constant — not the ADC-counts-to-velocity
factor. This silently scaled every plot ~38% too low for Normal-range
geophones (the correct full-scale is 10.0 in/s, or 1.25 in/s for
Sensitive). Conversion is now done server-side using the geo_range
from compliance config; the client just plots.
- New `sfm/event_hdf5.py` module: `write_event_hdf5()`,
`read_event_hdf5()`, plus a plot-JSON helper.
- Backfill script extended to also emit `.h5` for existing events.
### Dependencies
- Added `h5py>=3.10` and `numpy>=1.24` for the HDF5 storage layer.
- Added `python-multipart>=0.0.7` (required by FastAPI for the
`/db/import/blastware_file` endpoint introduced in this release).
- Added `waveform_key` and `event_timestamp` columns to `CachedEvent` and `CachedWaveform` for integrity verification.
- Implemented logic to flush the cache when a mismatch in (waveform_key, event_timestamp) is detected during event and waveform updates.
- Enhanced `set_events` and `set_waveform` methods to check for mismatches and trigger cache eviction as necessary.
- Introduced a new `LiveCache` class to manage in-memory caching of live device data, separating it from the server logic for better testability.
- Added tests to verify the correctness of cache invalidation logic, particularly for post-erase key reuse scenarios.
- Updated web application to include a "Force refresh" toggle, allowing users to bypass the cache and re-fetch data from the device.
test: add regression tests for v0.14.x SUB 5A protocol fixes
refactor(logging): change warning logs to debug for less verbosity in write_blastware_file
Introduces sfm/cache.py — a SQLite-backed cache (via SQLAlchemy) that
sits between the SFM REST endpoints and the device, eliminating redundant
cellular downloads for data that doesn't change.
Cache behaviour by data type:
- Device info / compliance config: cached until a config write occurs;
POST /device/config now calls mark_config_dirty() to force a fresh read
on the next /device/info call.
- Event headers + peak values: cached permanently (append-only). On
subsequent calls to /device/events, the server does a fast count_events()
(~2s) instead of a full download (~10-30s); only new events are fetched
from the device and merged into the cache.
- Full waveforms (raw ADC samples): cached permanently — immutable once
recorded. Repeated requests for the same waveform return instantly with
zero device contact.
- Monitor status (battery, memory, is_monitoring): 30-second TTL; auto-
invalidated on start/stop monitoring commands.
All endpoints gain a ?force=true param to bypass the cache when needed.
New endpoints: GET /cache/stats, DELETE /cache/device.
Adds requirements.txt listing fastapi, uvicorn, sqlalchemy, pyserial.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
serversdownandClaude Opus 4.7