mirror of https://github.com/pe1hvh/meshcore-gui.git synced 2026-07-05 17:31:32 +02:00

Files

T

pe1hvh cced48c10e patch: JSONL stream output for RX log alongside existing JSON archivef(#v1.22.1)

v1.22.1: JSONL stream output for RX log alongside existing JSON archive

Every received LoRa packet is now also written immediately as a single
JSON line to ~/.meshcore-gui/archive/<device>_rxlog.jsonl. This is an
append-only, unbuffered stream format that lets separate local services
(such as meshcore-watchlist) consume the RX feed in real time without
depending on the GUI's internal batched-JSON format.

- The existing <device>_rxlog.json is unchanged (60 s flush interval,
  atomic rewrite). The GUI, the public REST API and the domca.nl
  ingest continue to work without modification.
- Writes to the JSONL file are direct (no buffer), so end-to-end
  latency from radio reception to JSONL line is sub-second.
- A failure on the JSONL path is logged via debug_print and does not
  affect the buffered JSON archive — the two paths are independent.
- _cleanup_rxlog() now also rewrites the JSONL file to drop entries
  older than RXLOG_RETENTION_DAYS. Corrupt lines (e.g. a partial
  last line after a crash) are skipped during cleanup.

No BLE/worker changes, no public REST API changes; SharedData and the
BLE command pipeline are untouched. Disk usage increases modestly
(one additional file per device, same retention window).

PATCH bump 1.22.0 → 1.22.1: purely additive, fully backwards-compatible.

2026-04-27 09:13:47 +02:00

96 KiB

Raw Permalink Blame History

CHANGELOG

All notable changes to MeshCore GUI are documented in this file. Format follows Keep a Changelog and Semantic Versioning.

[1.22.1] - 2026-04-27

Added

JSONL stream output for the RX log (services/message_archive.py): Every received LoRa packet is now also written immediately to an append-only JSON Lines file at ~/.meshcore-gui/archive/<device>_rxlog.jsonl, one JSON object per line. This provides a real-time data source for separate local services (such as meshcore-watchlist) that want to consume the raw RX feed without depending on the GUI's internal batched JSON file format.
- The new file lives alongside the existing <device>_rxlog.json. The original batched archive (60 s flush interval, atomic rewrite) is unchanged so the GUI, the public REST API and domca.nl keep working without modification.
- Writes are direct (no buffer) so end-to-end latency from radio reception to JSONL line is sub-second, suitable for live monitoring use cases.
- A failure on the JSONL append path is logged via debug_print and does not affect the buffered JSON archive — the two paths are independent.

Changed

_cleanup_rxlog() now also rewrites the JSONL stream file to drop entries older than RXLOG_RETENTION_DAYS. Same retention policy as the existing JSON archive; corrupt lines (e.g. a partial last line after a crash) are skipped during cleanup.
VERSION bumped 1.22.0 → 1.22.1 (PATCH: additive feature, fully backwards-compatible).

Impact

No BLE/worker changes; SharedData and the BLE command pipeline are untouched.
No public REST API changes; domca.nl ingest is unaffected.
Existing consumers of <device>_rxlog.json see no difference.
Disk usage increases modestly (one additional file per device, same retention window). On a Raspberry Pi 5 with SSD the extra per-packet I/O is negligible.

[1.22.0] - 2026-04-21

Added

Drawer channel-list sort toggle (services/channel_sort_store.py, services/channel_service.py, gui/dashboard.py): The drawer Messages and Archive submenus can now be sorted either by channel index (the native MeshCore order, default) or alphabetically by channel name. The choice is exposed as a single sub-button per submenu labelled ↕ Sort: index / ↕ Sort: name and is shared between both submenus so a single click reorders both lists at once.
- New store ChannelSortStore persists the preference to ~/.meshcore-gui/channel_sort.json so it survives application restarts. The store is global (not per-device) because the sort mode is a pure UI concern.
- New pure helper sort_channels(channels, mode) in services/channel_service.py. The Public channel (idx == 0) is always pinned to the top regardless of mode; moving it during an alphabetical sort would be confusing as it is the default broadcast slot. Sort-by-name is case-insensitive (str.casefold).
- The dashboard submenu fingerprint now includes the sort mode so a user-initiated toggle forces a rebuild even when the channel list itself has not changed. After toggling, the dashboard calls _update_submenus immediately with a fresh SharedData snapshot so the reorder is visible without waiting for the next 500 ms tick.

Changed

CHANNEL_SORT_MODE_DEFAULT added to config.py as the factory default for the sort mode ("index"). Used by ChannelSortStore on first run and when the stored file is missing or contains an invalid value.
VERSION bumped 1.21.0 → 1.22.0 (MINOR: new backwards-compatible feature).

Impact

No BLE/worker changes; SharedData and the BLE command pipeline are untouched.
The ChannelPanel Move/Reindex dropdown is intentionally NOT affected — that list is an admin tool for operators who know which slot they want, not a navigation aid.
The (unused but still present) FilterPanel is not touched.
Existing drawer submenu functionality — delete/move per-channel buttons, Add/Backup/Restore buttons, DM and ALL entries — is preserved unchanged.

Rationale

Operators with a large number of channels reported needing a way to find a specific channel quickly. Index order is fine for small deployments but becomes unwieldy above ~10 channels. Alphabetical order provides a predictable scan path; keeping Public pinned at the top preserves the mental model of "slot 0 is the broadcast channel". Persistence across restarts avoids forcing the user to re-select the preferred order every session.

Added

Local channel backup & restore (services/channel_backup_store.py, gui/panels/channel_backup_panel.py): Channels can now be snapshotted to a local JSON file and restored after a firmware reflash, NVS erase or device replacement. The operator no longer needs to externally note down each slot's PSK before flashing.
- New store ChannelBackupStore writes to ~/.meshcore-gui/channel_backups/_<safe_dev_id>_channels.json, mirroring the BotConfigStore / PinStore per-device filename convention so multiple nodes coexist without conflict.
- Data is sourced exclusively from DeviceCache (channel_keys + channel_names) plus the live channel snapshot from SharedData. No new BLE/serial protocol calls are introduced — restore reuses the existing add_channel command handler via the worker command queue.
- On-disk schema: {schema_version, device_id, firmware_version, exported_at, channels: [{slot_idx, name, psk_hex}]}. Schema version is pinned at 1; ChannelBackupStore._parse() rejects unknown versions so older clients fail fast when the format changes.
- The schema is intentionally distinct from the public REST API's channel payload: backups include every slot (public, hashtag, and private) because local recovery is the whole point, while public_api_service continues to filter out private channels as before. Backup files never leave ~/.meshcore-gui/ and are not exposed by the API route layer.
- New panel ChannelBackupPanel adds two dialogs accessed from the Messages submenu:
  - 💾 Backup Channels — one-click export with pre/post summary showing how many slots were written and, critically, how many had no cached PSK (those cannot be restored without manual input — the dialog surfaces the count so the operator is not caught out).
  - 📥 Restore Channels — loads the device's own backup file by default, or an uploaded .json file for cross-device restores (e.g. migrating from an old Heltec V3 to a new one after firmware flashing). Entries are classified into restorable / conflict / identical / skipped buckets before any device write happens, so the operator sees in advance which slots will be overwritten and which have no PSK. Confirm dispatches one add_channel command per entry through the normal worker queue.

Changed

gui/dashboard.py: DashboardPage.__init__ now accepts an optional device_id parameter (default ""). This is threaded through to ChannelBackupPanel so the backup store can namespace its file to the correct node. All existing call sites were updated; the parameter is keyword-compatible so external callers that did not supply it continue to work unchanged.
__main__.py: added module-level _device_id so the @ui.page('/') handler (which recreates DashboardPage per client session) can pass the device identifier into the dashboard constructor.
config.py: version bump 1.20.2 → 1.21.0.

Not changed (deliberately out of scope)

No automatic or scheduled backups — the operator triggers a backup explicitly before risky operations. Auto-backup on channel changes was considered and deferred; the cache already persists channel keys on every write, so an on-demand export is sufficient.
No cloud/remote storage and no integration with domca.nl. Backup files contain private channel PSKs and are therefore strictly local.
No new BLE/serial commands. Restore is purely a replay of existing add_channel operations; the firmware is unaware that a bulk restore is happening.

[1.20.2] - 2026-04-19

Fixed

Public API omitted sender_pubkey and path_names from the message payload (services/public_api_service.py): get_messages_payload() built each item dict with sender, text, timestamp, hops and path_hashes, but silently dropped the already- resolved sender_pubkey and path_names fields that BleEventHandler writes to every archived message (see _resolve_path_names() in ble/events.py and the archive schema in services/message_archive.py lines 135–137). Downstream consumers were therefore forced to re-resolve path hashes themselves using only the 1-byte path-hash prefix — a lookup that collides heavily in networks with more than ~256 nodes and yields the wrong repeater name, type and coordinates on nearly every hop. The sender_pubkey omission had a similar effect on the sender column: clients could only match on display-name, which is ambiguous when two nodes share a name stem (e.g. NL-OV-ZWO-LGH-PD5WB vs the mobile variant NL-OV-ZWO-LGH-PD5WB-MOB). Fix: added "sender_pubkey" and "path_names" to the item dict. Both fields are read straight from the archive — no new resolution logic is introduced, so there is no additional cost on the hot path. The response schema change is additive: existing clients that ignore unknown keys continue to work unchanged.

Changed

config.py: version bump 1.20.1 → 1.20.2.

[1.20.1] - 2026-04-15

Fixed

Bot global cooldown blocked all senders (services/bot.py): _last_reply was a single float, so a reply to any node would start a 5-second silence window for all other nodes. During testing with multiple contacts, only the first #test message received a bot reply; subsequent messages from different senders were silently dropped by Guard 5. Fix: replaced _last_reply: float with _last_reply_per_sender: Dict[str, float]. Each sender now has an independent cooldown window; a reply to one node does not affect any other node. LRU eviction caps the dict at 200 entries to prevent unbounded memory growth in long-running sessions.
Bot deaf on first run when no channels saved (services/bot.py): _get_active_channels() returned an empty frozenset when BotConfigStore had no saved channel selection (i.e. the user had never clicked "💾 Save channels"). The bot was therefore silent on all channels despite the BOT panel showing all channels pre-checked. The BotSettings.channels docstring already documented this as the intended fallback case, but the code did not implement it. Fix: _get_active_channels() now falls back to BotConfig.channels ({1, 4} — #test and #bot) when the stored set is empty, matching the documented intent.

[1.20.0] - 2026-04-10

Changed

services/device_identity.py: device_identity.json upgraded from v1 (single flat object) to v2 (dict keyed by source_device). Multiple GUI instances running on different serial ports (e.g. /dev/ttyUSB0 and /dev/ttyUSB1) each write their own entry without overwriting each other.
write_device_identity() now reads the existing file before writing, updating only the entry for the current source_device.
read_device_identity() accepts an optional source_device parameter: returns a single entry dict when specified, or the full multi-device dict when called without arguments.
_load_raw() (internal) handles v1 → v2 migration transparently on first write: the old flat object is re-keyed under its source_device value.
Console output now includes the device path: 📝 Device identity saved → ~/.meshcore-gui/device_identity.json [/dev/ttyUSB1].
No changes to ble/worker.py or any other module — API is fully backward compatible.

[1.19.0] - 2026-04-06

FIXED

Wrong channel attribution for hashtag channels (ble/packet_decoder.py): ChannelCrypto.calculate_channel_hash() produces a different channel identifier than what the MeshCore firmware embeds in packets, causing _hash_to_idx to return None for all hashtag-channel messages. As a result, on_channel_msg fell back to the channel_idx from the CHANNEL_MSG_RECV event — which was itself incorrect — and messages sent on #mc-radar appeared in #weather and vice versa. Fix: brute-force channel resolution. When the hash lookup returns None, every registered key is tried individually via a single-key keystore. The first key that produces a valid decryption determines the channel index. The result is cached in _hash_to_idx for O(1) resolution on all subsequent packets for that channel.
Cross-channel message duplication (ble/events.py): A second copy of the same physical packet was stored under a different channel index because two independent dedup guards both failed simultaneously: (1) the message_hash from meshcoredecoder differed from the one in the CHANNEL_MSG_RECV event payload (two separate libraries), breaking hash-based dedup; (2) the channel_idx resolved by PacketDecoder differed from the channel_idx reported by the event, breaking content-based dedup. Fix: when on_rx_log successfully stores a message, it now also marks a channel-agnostic sentinel key ('*') in DualDeduplicator. on_channel_msg checks this sentinel before storing and suppresses the message if set, regardless of hash or channel_idx differences between the two systems.
Empty path_hashes on hashtag channels (ble/events.py): _path_cache is keyed by the message_hash from meshcoredecoder, but on_channel_msg performed the lookup with the message_hash from meshcore (different value), causing _path_cache.pop() to always return []. Fix: secondary _path_cache_by_content keyed by "sender:text[:100]" as fallback. Path hashes are now recovered even when the two hash values disagree.
Stale channel key in cache after del_channel reindex (ble/commands.py, services/cache.py): After moving a channel slot from old_idx to new_idx, the cache entry for old_idx was never removed. On the next startup, both the old and the new index had the same secret, causing the same channel to appear twice in the cache and the last channel not to be displayed. Fix: new DeviceCache.remove_channel_key(idx) method; called after each slot move in _cmd_del_channel. Added asyncio.sleep(0.5) before re-discovery to let the device commit all slot changes before _discover_channels reads them.

ADDED

Channel Move / Reindex (gui/panels/channel_panel.py, ble/commands.py, gui/dashboard.py): New mode ↕️ Move / Reindex in the Channel Manager dialog. The user selects a source channel from a dropdown and a target index from the number field. A ↕ button appears inline next to 🗑 for each channel in both the Messages and Archive submenus. _cmd_move_channel reads the channel secret from DeviceCache (or fetches it directly from the device as fallback), writes to the new slot, clears the old slot, and updates both cache entries atomically before triggering re-discovery.

CHANGED

gui/panels/channel_panel.py: Dialog title changed from 📡 Add Channel to 📡 Channel Manager; submit button renamed from Add Channel to Confirm.
gui/dashboard.py: _make_channel_sub_item() extended with on_move callback and inline ↕ button; both Messages and Archive submenus pass the callback.
config.py: version bump 1.18.1 → 1.19.0.

IMPACT

ble/packet_decoder.py: _secret_to_idx dict added; _resolve_channel_by_brute_force() helper added; fallback invoked in decode() when hash lookup fails. O(n_channels) cost on first packet per channel; O(1) thereafter.
ble/events.py: _path_cache_by_content dict added; sentinel mark added in on_rx_log; sentinel check and content-key path fallback added in on_channel_msg.
ble/commands.py: move_channel registered in handler dict; _cmd_move_channel() added; _cmd_del_channel() calls remove_channel_key() and includes settle delay.
services/cache.py: remove_channel_key(idx) added — no-op when key absent.
gui/panels/channel_panel.py: _move_section, _move_select widgets added; open() accepts mode and preselect_idx parameters; _submit_move() added.
gui/dashboard.py: _make_channel_sub_item() signature extended with on_move.

[1.18.1] - 2026-04-05

FIXED

Drawer width (gui/dashboard.py): increased from 300 px to 360 px so that longer channel names such as [19] #radio-zend-amateurs fit without truncation.
del_channel library fallback (ble/commands.py): _cmd_del_channel now catches AttributeError when mc.commands.del_channel() is not available in the installed pymeshcore version and falls back to overwriting the slot with an empty name via set_channel(idx, '', None), which the firmware treats as removal.
Delete confirmation dialog (gui/dashboard.py): clicking 🗑 now opens an "Are you sure?" dialog (Cancel / Delete) before dispatching the del_channel command, preventing accidental removals.

CHANGED

config.py: version bump 1.18.0 → 1.18.1.

[1.18.0] - 2026-04-05

ADDED

Channel delete button (gui/dashboard.py): each channel entry in the MESSAGES and ARCHIVE submenus now shows an inline 🗑 delete button next to the channel name. Clicking it queues a del_channel command for the BLE worker without requiring any additional confirmation dialog.
del_channel command handler (ble/commands.py): new _cmd_del_channel() async method that deletes the target channel slot via mc.commands.del_channel(idx) and then re-indexes all higher-numbered channels by one position using set_channel + del_channel. Secrets for private channels are read from the DeviceCache so no key material is lost during renumbering. A full channel re-discovery is triggered afterwards via _load_data_callback().

CHANGED

Drawer width (gui/dashboard.py): left navigation panel widened from 260 px to 300 px (min-width 200 px → 220 px) for better readability of channel names.
config.py: version bump 1.17.1 → 1.18.0.

IMPACT

gui/dashboard.py: new static method _make_channel_sub_item(); _update_submenus() uses it instead of _make_sub_btn() for channel rows. All existing submenu logic (ALL, DM, ＋ Add Channel, rooms) is unchanged.
ble/commands.py: one new handler registered; no existing handlers touched.

RATIONALE

Users need a quick way to remove channels directly from the navigation menu without opening a separate dialog.
Re-indexing keeps the channel list compact (no sparse gaps) which matches MeshCore firmware expectations and the visual convention of sequential indices.

[1.17.1] - 2026-04-04

FIXED

Multibyte path hash support (ble/events.py, core/shared_data.py): corrected docstrings in both _resolve_path_names methods that incorrectly described path hashes as "2-char hex strings". The actual contact lookup uses startswith matching, which is hash-size agnostic and correctly handles 1-byte (2 hex chars), 2-byte (4 hex chars) and 3-byte (6 hex chars) path hashes as introduced in MeshCore firmware v1.14.0. No functional code was changed — only the documentation was incorrect.
MariaDB schema (meshcore_schema.sql): meshcore_messages.path_hashes column widened from VARCHAR(128) to VARCHAR(255). The old limit caused silent truncation for paths longer than ~40 hops in 1-byte mode or ~25 hops in 2-byte mode. Migration is backward-compatible; existing data is unchanged.

CHANGED

config.py: version bump 1.17.0 → 1.17.1.

RATIONALE

MeshCore firmware v1.14.0 (2026-03-06) introduced configurable path hash sizes (1-, 2- or 3-byte per repeater). Verification confirmed that meshcoredecoder 0.3.2 already returns correctly sized hex strings via _decode_path_len_byte. The GUI path-resolution logic was already forward-compatible; only the docstrings and the MariaDB column width required correction.

IMPACT

No BLE handler, GUI panel, service or API endpoint modified.
meshcoredecoder library unchanged; no pip update required.
MariaDB migration: single ALTER TABLE statement, no downtime, no data loss.

[1.17.0] - 2026-04-04

ADDED

Public REST API (api/routes.py, api/__init__.py): four read-only GET endpoints registered on the NiceGUI/FastAPI application instance. Enabled via API_ENABLED = True in config.py (default: on).
- GET /api/v1/stats — aggregate statistics for the last 72 hours: total messages, unique senders, active nodes per type, average hops and peak hour. Only public and hashtag channel messages are counted.
- GET /api/v1/nodes — full contact list with node type, GPS coordinates and (when available) last-seen timestamp and battery voltage.
- GET /api/v1/messages?limit=N&offset=N — paginated message list restricted to public (idx 0) and hashtag (name.startswith('#')) channels. Private channel messages are unconditionally excluded.
- GET /api/v1/channels — channel list with is_private flag per entry.
PublicApiService (services/public_api_service.py): pure-Python business logic for all four endpoints. Contains the single source of truth for channel-type classification (is_public_channel, is_private_channel) used throughout the API layer.
API_ENABLED and API_CORS_ORIGINS constants (config.py): toggle the API on/off and configure allowed CORS origins.

CHANGED

config.py: version bump 1.16.0 → 1.17.0; PUBLIC API section added with API_ENABLED and API_CORS_ORIGINS.
__main__.py: conditional register_routes(_shared) call after SharedData construction; prints API URL or disabled notice at startup.

RATIONALE

Enables the domca.nl PHP collector to pull live mesh data over HTTP without direct access to the SQLite archive or SharedData internals.
Filtering is enforced server-side: what is not public can never leak, even without authentication.

IMPACT

No existing route, panel or BLE handler modified.
SharedData and MessageArchive accessed read-only.
Zero breaking changes to v1.16.0 behaviour when API_ENABLED = False.

[1.16.0] - 2026-04-04

ADDED

Add Channel dialog (gui/panels/channel_panel.py): new ChannelPanel class that renders a ui.dialog with three modes — Hashtag, Private New, and Private Existing — accessible via a ＋ Add Channel button at the bottom of the Messages submenu.
ChannelService (services/channel_service.py): pure-Python business logic for channel key management. Provides generate_secret(), derive_hashtag_key(), build_qr_url() and generate_qr_base64(). No GUI or BLE dependencies.
_cmd_add_channel (ble/commands.py): new BLE command handler that calls mc.commands.set_channel(idx, name, secret) and triggers a full channel re-discovery on success so the GUI immediately reflects the new channel.
＋ Add Channel submenu button (gui/dashboard.py): added to the Messages submenu — present on initial render and preserved through all dynamic rebuilds.
QR code sharing: after adding a new private channel the dialog shows a scannable QR code (meshcore://channel/add?name=…&secret=…) and a copy-to-clipboard button for the hex key.

CHANGED

gui/panels/__init__.py: ChannelPanel added to re-exports.
gui/dashboard.py: ChannelPanel imported, instantiated in render(), updated in _update_ui() and opened from the ＋ Add Channel submenu button.
ble/commands.py: add_channel registered in the handler dict.

RATIONALE

Channels could previously only be added by flashing or using a separate tool. The dialog covers all three user scenarios (hashtag join, private create, private join) without requiring any changes to the BLE worker's discovery logic.

IMPACT

No existing command handlers modified. No existing panel logic altered. Pure addition — all existing functionality unaffected.

[1.15.0] - 2026-03-16

ADDED

BOT panel (gui/panels/bot_panel.py): new dedicated panel in the main menu (between RX LOG and BBS) with enable toggle, private mode toggle and interactive channel assignment via checkboxes built from the live device channel list.
BotConfigStore (services/bot_config_store.py): persistent bot configuration per device stored at ~/.meshcore-gui/bot/_<dev_id>_bot.json. Saves enabled flag, private mode state and selected channel set across restarts.
Private mode: when enabled the bot only replies to pinned contacts. Guard 1.5 added to MeshBot.check_and_reply — reads live from BotConfigStore so changes take effect immediately without restart.
Private mode constraint: private mode can only be activated when at least one contact is pinned. The toggle is disabled (greyed out) with an explanation label when no pinned contacts exist; auto-disables if all pins are removed.
Interactive channel assignment: BOT panel shows a checkbox per discovered channel; selection persisted via BotConfigStore.set_channels() on Save.
BOT_DIR config constant (~/.meshcore-gui/bot/) centralising the storage root for bot configuration files.

CHANGED

BOT toggle removed from ActionsPanel: actions_panel.py no longer contains the BOT checkbox or set_bot_enabled wiring; the panel is now solely for Refresh, Advertise and Set device name.
MeshBot gains two optional constructor arguments: config_store (BotConfigStore) for live channel/private-mode reads, and pinned_check (Callable[[str], bool]) for pin lookups. Fully backwards-compatible — both default to None and existing behaviour is preserved when absent.
MeshBot.check_and_reply gains optional sender_pubkey kwarg used by Guard 1.5.
_BaseWorker now accepts optional pin_store kwarg; wires pinned_check and config_store into MeshBot at construction time.
create_worker forwards optional pin_store kwarg to subworkers.
DashboardPage receives BotConfigStore instance; ActionsPanel call no longer passes set_bot_enabled.

IMPACT

ble/events.py: both check_and_reply call sites now pass sender_pubkey=.
ble/worker.py: _BaseWorker, SerialWorker, BLEWorker, create_worker updated.
gui/dashboard.py: BotPanel registered as panel 'bot'; menu item 🤖 BOT added.
gui/panels/actions_panel.py: BOT toggle removed; ActionsPanel.__init__ signature simplified to (put_command).
config.py: VERSION bumped to 1.15.0; BOT_DIR constant added.

RATIONALE

Bot functionality was embedded in the Actions panel and had no persistence. Extracting it to a dedicated panel and a config store aligns with the existing modularity of the codebase (cf. BBS panel / BbsConfigStore) and enables future extension. Private mode fulfils the requirement to restrict bot replies to trusted contacts only.

📈 Performance note — v1.13.1 through v1.13.4 Although versions 1.13.1–1.13.4 were released as targeted bugfix releases, the cumulative effect of the fixes delivered a significant performance improvement:

v1.13.1 — Bot non-response fix eliminated a silent failure path that caused repeated dedup-marked command re-evaluation on every message tick.

v1.13.2 — Map display fixes prevented Leaflet from being initialized on hidden zero-size containers, removing a source of repeated failed bootstrap retries and associated DOM churn.

v1.13.3 — Active panel timer gating reduced the 500 ms dashboard update work to only the currently visible panel, cutting unnecessary UI updates and background redraw load substantially — especially noticeable over VPN or on slower hardware.

v1.13.4 — Room Server event classification fix and sender name resolution removed redundant fallback processing paths and reduced per-tick contact lookup overhead.

Users upgrading from v1.12.x or earlier will notice noticeably faster panel switching, lower CPU usage during idle operation, and more stable map rendering.

[1.14.3] - 2026-03-16 — BBS !h / !help NameError fix

Fixed

🐛 services/bbs_service.py — !h en !help DM-commando's gooiden een NameError: name 'cu' is not defined in _abbrev_table().
- Root cause: cu werd gedefinieerd in een inner list comprehension [cu.upper() for cu in categories], maar Python 3 list comprehensions hebben een eigen scope. De if cu.upper() in inv in de buitenste generator expression kon cu daardoor niet bereiken.
- Fix: list comprehension extracted naar een aparte variabele cats_upper; de generator itereert nu over die lijst.

[1.14.2] - 2026-03-16 — BBS whitelist fix: !bbs channel hook in on_rx_log

Fixed

🐛 ble/events.py — !bbs op een geconfigureerd BBS-channel deed nooit een whitelist-add, waardoor !h en andere DM-BBS-commando's daarna silently werden gedropped.
- Root cause: de BBS channel hook stond uitsluitend in on_channel_msg, maar on_channel_msg wordt in het normale pad onderdrukt door de content-dedup early-return (het bericht is dan al door on_rx_log verwerkt en gemarkeerd).
- Fix: BBS channel hook (handle_channel_msg) ook aangeroepen in on_rx_log, direct ná de bot-aanroep, binnen de GroupText + channel_idx resolved-branch. sender_pubkey is daar al opgelost via get_contact_by_name.
- De hook in on_channel_msg blijft intact als fallback voor het deferred-path (channel_idx onopgelost in on_rx_log).

[1.14.1] - 2026-03-16 — BBS test corrections

Changed

Testing package flattened to a single canonical meshcore_gui/... tree so runtime and validation target one code path.
!bbs channel bootstrap, DM-only !h / !help, and chunked BBS reply work were applied as in-progress fixes under version 1.14.1 while testing continues.
No release bump: version numbering is kept at 1.14.1 for this test set.

[1.14.0] - 2026-03-14 — BBS (Bulletin Board System)

Added

🆕 BBS — Bulletin Board System — offline berichtenbord voor mesh-netwerken.
- Toegangsmodel: de beheerder selecteert één of meer channels in de settings. Iedereen die op een van die channels een bericht stuurt, wordt automatisch gewhitelist en kan daarna commando's sturen via Direct Message aan de node. Het channel blijft schoon; alleen de eerste interactie verloopt via het channel.
- Korte syntax: !p <cat> <tekst> (post) en !r [cat] (lezen). Categorie-afkortingen automatisch berekend als kortste unieke prefix (bijv. U=URGENT M=MEDICAL).
- Volledige syntax behouden: !bbs post, !bbs read, !bbs help.
- Optioneel regio-filter en handmatige allowed-keys override in Advanced.
- Settings-pagina (/bbs-settings): checkboxes per channel, categorieën, retentie, Advanced voor regio's en handmatige keys.
- Berichten opgeslagen in SQLite (~/.meshcore-gui/bbs/bbs_messages.db, WAL-mode).

Changed

🔄 ble/events.py — on_channel_msg roept BbsCommandHandler.handle_channel_msg() aan op geconfigureerde BBS-channels: auto-whitelist + bootstrap reply. on_contact_msg stuurt !-DMs direct naar handle_dm(). Beide paden volledig los van MeshBot.
🔄 services/bot.py — MeshBot is weer een pure keyword/channel responder; BBS-routing verwijderd.
🔄 services/bbs_config_store.py — configure_board() (multi-channel), add_allowed_key() (auto-whitelist), clear_board().
🔄 gui/dashboard.py — BbsPanel geregistreerd, 📋 BBS drawer-item toegevoegd.

Storage

~/.meshcore-gui/bbs/bbs_config.json   — board configuratie
~/.meshcore-gui/bbs/bbs_messages.db   — SQLite berichtenopslag

Fixed

🛠 Route page back-button navigated to main menu regardless of origin — the two fixed navigation buttons (/ and /archive) are replaced by a single arrow_back button that calls window.history.back(), so the user is always returned to the screen that opened the route page.
🛠 Map marker popup flickered on every 500 ms update tick — the periodic applyContacts / applyDevice calls in leaflet_map_panel.js invoked setIcon() and setPopupContent() on all existing markers unconditionally. setIcon() rebuilds the marker DOM element; when a popup was open this caused the popup anchor to detach and reattach, producing visible flickering. Both functions now check marker.isPopupOpen() and skip icon/content updates while the popup is visible.
🛠 Map marker popup appeared with a flicker/flash on first click (main map and route map) — Leaflet's default fadeAnimation: true caused popups to fade in from opacity 0, which on the Raspberry Pi rendered as a visible flicker. Both L.map() initialisations (ensureMap and MeshCoreRouteMapBoot) now set fadeAnimation: false and markerZoomAnimation: false so popups appear immediately without animation artefacts.

Changed

🔄 meshcore_gui/gui/route_page.py — Replaced two fixed-destination header buttons with a single arrow_back button using window.history.back().
🔄 meshcore_gui/static/leaflet_map_panel.js — applyDevice and applyContacts guard setIcon / setPopupContent behind isPopupOpen(). Both L.map() calls add fadeAnimation: false, markerZoomAnimation: false.
🔄 meshcore_gui/config.py — Version bumped to 1.13.5.

Impact

Back navigation from the route page now always returns to the correct origin screen.
Open marker popups are stable during map update ticks; content refreshes on next tick after the popup is closed.
Popup opening is instant on both maps; no animation artefacts on low-power hardware.

[1.13.4] - 2026-03-12 — Room Server message classification fix

Fixed

🛠 Incoming room messages from other participants could be misclassified as normal DMs — CONTACT_MSG_RECV room detection now keys on txt_type == 2 instead of requiring signature.
🛠 Incoming room traffic could be attached to the wrong key — room message handling now prefers room_pubkey / receiver-style payload keys before falling back to pubkey_prefix.
🛠 Room login UI could stay out of sync with the actual server-confirmed state — LOGIN_SUCCESS now updates room_login_states and refreshes room history using the resolved room key.
🛠 Room Server panel showed hex codes instead of sender names — when a contact was not yet known at the time a room message was archived, msg.sender was stored as a raw hex prefix. The panel now performs a live lookup against the current contacts snapshot on every render tick, so names are shown as soon as the contact is known.

Changed

🔄 meshcore_gui/ble/events.py — Broadened room payload parsing and added payload-key debug logging for incoming room traffic.
🔄 meshcore_gui/ble/worker.py — LOGIN_SUCCESS handler now updates per-room login state and refreshes cached room history.
🔄 meshcore_gui/config.py — Version kept at 1.13.4.

Impact

Keeps the existing Room Server panel logic intact.
Fix is limited to room event classification and room login confirmation handling.
No intended behavioural change for ordinary DMs or channel messages.

[1.13.3] - 2026-03-12 — Active Panel Timer Gating

Changed

🔄 meshcore_gui/gui/dashboard.py — The 500 ms dashboard timer now keeps only lightweight global state updates running continuously (status label, channel filters/options, drawer submenu consistency). Expensive panel refreshes are now gated to the currently active panel only
🔄 meshcore_gui/gui/dashboard.py — Added immediate active-panel refresh on panel switch so newly opened panels populate at once instead of waiting for the next timer tick
🔄 meshcore_gui/gui/panels/map_panel.py — Removed eager hidden ensure_map bootstrap from render(); the browser map now starts only when real snapshot work exists or when a live map already exists
🔄 meshcore_gui/static/leaflet_map_panel.js — Theme-only calls without snapshot work no longer start hidden host retry processing before a real map exists
🔄 meshcore_gui/config.py — Version bumped to 1.13.3

Fixed

🛠 Hidden panels still refreshed every 500 ms — Device, actions, contacts, messages, rooms and RX log are no longer needlessly updated while another panel is active
🛠 Map bootstrap activity while panel is not visible — Removed one source of MeshCoreLeafletBoot timeout waiting for visible map host caused by eager hidden startup traffic
🛠 Slow navigation over VPN — Reduced unnecessary dashboard-side UI churn by limiting timer-driven work to the active panel

Impact

Faster panel switching because the selected panel gets one direct refresh immediately
Lower background UI/update load on dashboard level, especially when the map panel is not active
Smaller chance of Leaflet hidden-host retries and related console noise outside active map usage
No intended functional regression for route maps or visible panel behaviour

[1.13.2] - 2026-03-11 — Map Display Bugfix

Fixed

🛠 MAP panel blank when contacts list is empty at startup — dashboard update loop had two separate conditional map-update blocks that both silently stopped firing after tick 1 when data['contacts'] was empty. Map panel received no further snapshots and remained blank indefinitely.
🛠 Leaflet map initialized on hidden (zero-size) container — processPending in the browser runtime called L.map() on the host element while it was still display:none (Vue v-show, panel not yet visible). This produced a broken 0×0 map that never recovered because ensureMap returned the cached broken state on all subsequent calls. Fixed by adding a clientWidth/clientHeight guard in ensureMap: initialization is deferred until the host has real dimensions.
🛠 Route map container had no height — route_page.py used the Tailwind class h-96 for the Leaflet host <div>. NiceGUI/Quasar does not include Tailwind CSS, so h-96 had no effect and the container rendered at height 0. Leaflet initialized on a zero-height element and produced a blank map.
🛠 Route map not rendered when no node has GPS coordinates — _render_map returned early before creating the Leaflet container when payload['nodes'] was empty. Fixed: container is always created; a notice label is shown instead.

Changed

🔄 meshcore_gui/static/leaflet_map_panel.js — Added size guard in ensureMap: returns null when host has clientWidth === 0 && clientHeight === 0 and no map state exists yet. processPending retries on the next tick once the panel is visible.
🔄 meshcore_gui/gui/dashboard.py — Consolidated two conditional map-update blocks into a single unconditional update while the MAP panel is active. Added h-96 to the DOMCA CSS height overrides for consistency with the route page map container.
🔄 meshcore_gui/gui/route_page.py — Replaced h-96 Tailwind class on the route map host <div> with an explicit inline style (height: 24rem). Removed early return guard so the Leaflet container is always created.

Impact

MAP panel now renders reliably on first open regardless of contact/GPS availability
Route map now always shows with correct height even when route nodes have no GPS
No breaking changes outside the three files listed above

[1.13.0] - 2026-03-09 — Leaflet Map Runtime Stabilization

Added

✅ meshcore_gui/static/leaflet_map_panel.js — Dedicated browser-side Leaflet runtime responsible for map lifecycle, marker registry, clustering and theme handling independent from NiceGUI redraw cycles
✅ meshcore_gui/static/leaflet_map_panel.css — Styling for browser-side node markers, cluster icons and map container
✅ meshcore_gui/services/map_snapshot_service.py — Snapshot service that normalizes device/contact map data into a compact payload for the browser runtime
✅ Browser-side map state management for center, zoom and theme
✅ Theme persistence across reconnect events via browser storage fallback
✅ Browser-side contact clustering via Leaflet.markercluster
✅ Separate non-clustered device marker layer so the own device remains individually visible

Changed

🔄 meshcore_gui/gui/panels/map_panel.py — Replaced NiceGUI Leaflet wrapper usage with a pure browser-managed Leaflet container while preserving the existing card layout, theme toggle and center-on-device control
🔄 Leaflet bootstrap moved out of inline Python into a dedicated browser runtime loaded from /static
🔄 Asset loading order is now explicit: Leaflet first, then Leaflet.markercluster, then the MeshCore panel runtime
🔄 Map initialization now occurs only once per container; NiceGUI refresh cycles no longer recreate the map
🔄 Dashboard update loop now sends compact map snapshots instead of triggering redraws
🔄 Snapshot processing in the browser is coalesced so only the newest payload is applied
🔄 Map markers are managed in separate device/contact layers and updated incrementally by stable node id
🔄 Contact markers are rendered inside a persistent cluster layer while the device marker remains outside clustering
🔄 Theme switching moved to a dedicated theme channel instead of being embedded in snapshot data

Fixed

🛠 Map disappearing during dashboard refresh cycles — prevented repeated map reinitialization caused by the 500 ms NiceGUI update loop
🛠 Markers disappearing between refreshes — marker updates are now incremental and keyed by node id
🛠 Blank map container on load — browser bootstrap now waits for DOM host, Leaflet runtime and panel runtime before initialization
🛠 Leaflet clustering bootstrap failure (L is not defined) — resolved by enforcing correct script dependency order before the panel runtime starts
🛠 MarkerClusterGroup failure (Map has no maxZoom specified) — the map now defines maxZoom during initial creation before the cluster layer is attached
🛠 Half-initialized map retry cascade (Map container is already initialized) — map state is now registered safely during initialization so a failed attempt cannot trigger a second L.map(...) on the same container
🛠 Race condition between queued snapshot and theme selection — explicit theme changes can no longer be overwritten by stale snapshot payloads
🛠 Viewport jumping back to default center/zoom — stored viewport is no longer reapplied on each snapshot update
🛠 Theme reverting to default during reconnect — effective map theme is restored before snapshot processing resumes

Impact

Leaflet map is now managed entirely in the browser and is no longer recreated on each dashboard refresh
Node markers remain stable and no longer flicker or disappear during the 500 ms update cycle
Dense contact sets can now be rendered with clustering without violating the browser-owned map lifecycle
Theme switching and viewport state persist reliably across reconnect events
No breaking changes outside the map subsystem

[1.12.1] - 2026-03-08 — Minor change bot

Changed

🔄 meshcore_gui/services/bot.py: remove path id's

Impact

No breaking changes — all existing functionality preserved serial.

[1.12.0] - 2026-02-26 — MeshCore Observer Fase 1

Added

✅ MeshCore Observer daemon — New standalone read-only daemon (meshcore_observer.py) that reads archive JSON files produced by meshcore_gui and meshcore_bridge, aggregates them, and presents a unified NiceGUI monitoring dashboard on port 9093.
✅ ArchiveWatcher — Core component that polls ~/.meshcore-gui/archive/ for *_messages.json and *_rxlog.json files, tracks mtime changes, and returns only new entries since previous poll. Thread-safe, zero writes, graceful on corrupt JSON.
✅ Observer dashboard panels — Sources overview, aggregated messages feed (sorted by timestamp), aggregated RX log table, and statistics panel with uptime/counters/per-source breakdown. Full DOMCA theme (dark + light mode).
✅ Source filter — Dropdown to filter messages and RX log by archive source.
✅ Channel filter — Dropdown to filter messages by channel name.
✅ ObserverConfig — YAML-based configuration with from_yaml() classmethod, defaults work without config file.
✅ observer_config.yaml — Documented config template with all options.
✅ install_observer.sh — systemd installer (/opt/meshcore-observer/, /etc/meshcore/observer_config.yaml), with --uninstall option.
✅ RxLogEntry raw packet fields — 5 new fields on RxLogEntry dataclass: raw_payload, packet_len, payload_len, route_type, packet_type_num (all with defaults, backward compatible).
✅ EventHandler.on_rx_log() metadata — Raw payload hex and packet metadata now passed through to RxLogEntry and archived (preparation for Fase 2 LetsMesh uplink).

Changed

🔄 meshcore_gui/core/models.py: RxLogEntry +5 fields with defaults (backward compatible).
🔄 meshcore_gui/ble/events.py: on_rx_log() fills raw_payload and metadata (~10 lines added).
🔄 meshcore_gui/services/message_archive.py: add_rx_log() serializes the 5 new RxLogEntry fields.
🔄 meshcore_gui/config.py: Version bumped to 1.12.0.

Impact

No breaking changes — All new RxLogEntry fields have defaults; existing archives and code work identically.
New daemon — meshcore_observer is fully standalone; no imports from meshcore_gui (reads only JSON files).

Added

✅ Serial CLI flags — --baud=BAUD and --serial-cx-dly=SECONDS for serial configuration at startup.

Changed

🔄 Connection layer — Switched from BLE to serial (MeshCore.create_serial) with serial reconnect handling.
🔄 config.py: Added SERIAL_BAUDRATE, SERIAL_CX_DELAY, DEFAULT_TIMEOUT, MESHCORE_LIB_DEBUG; removed BLE PIN settings; version bumped to 1.10.0.
🔄 meshcore_gui.py / meshcore_gui/__main__.py: Updated usage, banners and defaults for serial ports.
🔄 Docs: Updated README and core docs for serial usage; BLE documents marked as legacy.

Impact

No breaking changes — all existing functionality preserved serial.

[1.9.11] - 2026-02-19 — Message Dedup Hotfix

Fixed

🛠 Duplicate messages after (re)connect — load_recent_from_archive() appended archived messages on every connect attempt without clearing existing entries; after N failed connects, each message appeared N times. Method is now idempotent: clears the in-memory list before loading.
🛠 Persistent duplicate messages — Live BLE events for messages already loaded from archive were not suppressed because the DualDeduplicator was never seeded with archived content. Added _seed_dedup_from_messages() in BLEWorker after cache/archive load and after reconnect.
🛠 Last-line-of-defence dedup in SharedData — add_message() now maintains a fingerprint set (message_hash or channel:sender:text) and silently skips messages whose fingerprint is already tracked. This guards against duplicates regardless of their source.
🛠 Messages panel empty on first click — _show_panel() made the container visible but relied on the next 500 ms timer tick to populate it. Added an immediate _messages.update() call so content is rendered the moment the panel becomes visible.

Changed

🔄 core/shared_data.py: Added _message_fingerprints set and _message_fingerprint() static method; add_message() checks fingerprint before insert and evicts fingerprints when messages are rotated out; load_recent_from_archive() clears messages and fingerprints before loading (idempotent)
🔄 ble/worker.py: Added _seed_dedup_from_messages() helper; called after _apply_cache() and after reconnect _load_data() to seed DualDeduplicator with existing messages
🔄 gui/dashboard.py: _show_panel() now forces an immediate _messages.update() when the messages panel is shown, eliminating the stale-content flash
🔄 config.py: Version bumped to 1.9.11

Impact

Eliminates all duplicate message display scenarios: initial connect, failed retries, reconnect, and BLE event replay
No breaking changes — all existing functionality preserved
Fingerprint set is bounded to the same 100-message cap as the message list

[1.9.10] - 2026-02-19 — Map Tooltips & Separate Own-Position Marker

Added

✅ Map marker tooltips — All markers on the Leaflet map now show a tooltip on hover with the node name and type icon (📱, 📡, 🏠) from TYPE_ICONS
✅ Separate own-position marker — The device's own position is now tracked as a dedicated _own_marker, independent from contact markers. This prevents the own marker from being removed/recreated on every contact update cycle

Changed

🔄 gui/panels/map_panel.py: Renamed _markers to _contacts_markers; added _own_marker attribute; own position marker is only updated when device_updated flag is set (not every timer tick); contact markers are only rebuilt when contacts_updated is set; added TYPE_ICONS import for tooltip icons
🔄 gui/dashboard.py: Added self._map.update(data) call in the device_updated block so the own-position marker updates when device info changes (e.g. GPS position update)
🔄 config.py: Version bumped to 1.9.10

Impact

Map centering on own device now works correctly and updates only when position actually changes
Contact markers are no longer needlessly destroyed and recreated on every UI timer tick — only on actual contact data changes
Tooltips make it easy to identify nodes on the map without clicking
No breaking changes — all existing map functionality preserved

Credits

Based on PR #16 by @rich257

[1.9.9] - 2026-02-18 — Variable Landing Page & Operator Callsign

Added

✅ Configurable operator callsign — New OPERATOR_CALLSIGN constant in config.py (default: "PE1HVH"). Used in the landing page SVG and the drawer footer copyright label. Change this single value to personalize the entire GUI for a different operator
✅ External landing page SVG — The DOMCA splash screen is now loaded from a standalone file (static/landing_default.svg) instead of being hardcoded in dashboard.py. New LANDING_SVG_PATH constant in config.py points to the SVG file. The placeholder {callsign} in the SVG is replaced at runtime with OPERATOR_CALLSIGN
✅ Landing page customization — To use a custom landing page: copy landing_default.svg (or create your own SVG), use {callsign} wherever the operator callsign should appear, and point LANDING_SVG_PATH to your file. The default SVG includes an instructive comment block explaining the placeholder mechanism

Changed

🔄 config.py: Added OPERATOR_CALLSIGN and LANDING_SVG_PATH constants in new OPERATOR / LANDING PAGE section; version bumped to 1.9.9
🔄 gui/dashboard.py: Removed hardcoded _DOMCA_SVG string (~70 lines); added _load_landing_svg() helper that reads SVG from disk and replaces {callsign} placeholder; CSS variable --pe1hvh renamed to --callsign; drawer footer copyright label now uses config.OPERATOR_CALLSIGN

Added (files)

✅ static/landing_default.svg — The original DOMCA splash SVG extracted as a standalone file, with {callsign} placeholder and --callsign CSS variable. Serves as both the default landing page and a reference template for custom SVGs

Impact

Out-of-the-box behavior is identical to v1.9.8 (same DOMCA branding, same PE1HVH callsign)
Operators personalize by changing 1–2 lines in config.py — no code modifications needed
Fallback: if the SVG file is missing, a minimal placeholder text is shown instead of a crash
No breaking changes — all existing dashboard functionality (panels, menus, timer, theming) unchanged

[1.9.8] - 2026-02-17 — Bugfix: Route Page Sender ID, Type & Location Not Populated

Fixed

🛠 Sender ID, Type and Location empty in Route Page — After the v4.1 refactoring to RouteBuilder/RouteNode, the sender contact lookup relied solely on SharedData.get_contact_by_prefix() (live lock-based) and get_contact_by_name(). When both failed (e.g. empty sender_pubkey from RX_LOG decode, or name mismatch), route['sender'] remained None and the route table fell through to a hardcoded fallback with type: '-', location: '-'. The contact data was available in the snapshot data['contacts'] but was never searched
🛠 Route table fallback row ignored available contact data — When route['sender'] was None, the _render_route_table method used a static fallback row without attempting to find the contact in the data snapshot. Even when the contact was present in data['contacts'] with valid type and location, these fields showed as '-'

Changed

🔄 services/route_builder.py: Added two additional fallback strategies in build() after the existing SharedData lookups: (3) bidirectional pubkey prefix match against data['contacts'] snapshot, (4) case-insensitive adv_name match against data['contacts'] snapshot. Added helper methods _find_contact_by_pubkey() and _find_contact_by_adv_name() for snapshot-based lookups
🔄 gui/route_page.py: Added defensive fallback in _render_route_table() sender section — when route['sender'] is None, attempts to find the contact in the snapshot via _find_sender_contact() before falling back to the static '-' row. Added _find_sender_contact() helper method

Impact

Sender ID (hash), Type and Location are now populated correctly in the route table when the contact is known
Four-layer lookup chain ensures maximum resolution: (1) SharedData pubkey lookup, (2) SharedData name lookup, (3) snapshot pubkey lookup, (4) snapshot name lookup
Defensive fallback in route_page guarantees data is shown even if RouteBuilder misses it
No breaking changes — all existing route page behavior, styling and data flows unchanged

[1.9.7] - 2026-02-17 — Layout Fix: Archive Filter Toggle & Route Page Styling

Changed

🔄 gui/archive_page.py: Archive filter card now hidden by default; toggle visibility via a filter_list icon button placed right-aligned on the same row as the "📚 Archive" title. Header restructured from single label to ui.row() with justify-between layout
🔄 gui/route_page.py: Route page now uses DOMCA theme (imported from dashboard.py) with dark mode as default, consistent with the main dashboard. Header restyled from bg-blue-600 to Quasar-themed header with JetBrains Mono font. Content container changed from w-full max-w-4xl mx-auto to domca-panel class for consistent responsive sizing
🔄 gui/dashboard.py: Added domca-header-text CSS class with @media (max-width: 599px) rule to hide header text on narrow viewports; applied to version label and status label
🔄 gui/route_page.py: Header label also uses domca-header-text class for consistent responsive behaviour

Added

✅ Archive filter toggle — filter_list icon button in archive header row toggles the filter card visibility on click
✅ Route page close button — X (close) icon button added right-aligned in the route page header; calls window.close() to close the browser tab
✅ Responsive header — On viewports < 600px, header text labels are hidden; only icon buttons (menu, dark mode toggle, close) remain visible

Impact

Archive page is cleaner by default — filters only shown when needed
Route page visually consistent with the main dashboard (DOMCA theme, dark mode, responsive panel width)
Headers degrade gracefully on mobile (< 600px): only icon buttons visible, no text overflow
No functional changes — all event handlers, callbacks, data bindings, logic and imports are identical to the input

[1.9.6] - 2026-02-17 — Bugfix: Channel Discovery Reliability

Fixed

🛠 Channels not appearing (especially on mobile) — Channel discovery aborted too early on slow BLE connections. The _discover_channels() probe used a single attempt per channel slot and stopped after just 2 consecutive empty responses. On mobile BLE stacks (WebBluetooth via NiceGUI) where GATT responses are slower, this caused discovery to abort before finding any channels, falling back to only [0] Public
🛠 Race condition: channel update flag lost between threads — get_snapshot() and clear_update_flags() were two separate calls, each acquiring the lock independently. If the BLE worker set channels_updated = True between these two calls, the GUI consumed the flag via get_snapshot() but then clear_update_flags() reset it — causing the channel submenu and dropdown to never populate
🛠 Channels disappear on browser reconnect — When a browser tab is closed and reopened, render() creates new (empty) NiceGUI containers for the drawer submenus, but did not reset _last_channel_fingerprint. The _update_submenus() method compared the new fingerprint against the stale one, found them equal, and skipped the rebuild — leaving the new containers permanently empty. Fixed by resetting both _last_channel_fingerprint and _last_rooms_fingerprint in render()

Changed

🔄 core/shared_data.py: New atomic method get_snapshot_and_clear_flags() that reads the snapshot and resets all update flags in a single lock acquisition. Internally refactored to _build_snapshot_unlocked() helper. Existing get_snapshot() and clear_update_flags() retained for backward compatibility
🔄 ble/worker.py: _discover_channels() — max_attempts increased from 1 to 2 per channel slot; inter-attempt delay increased from 0.5s to 1.0s; consecutive error threshold raised from 2 to 3; inter-channel pause increased from 0.15s to 0.3s for mobile BLE stack breathing room
🔄 gui/dashboard.py: _update_ui() now uses get_snapshot_and_clear_flags() instead of separate get_snapshot() + clear_update_flags(); render() now resets _last_channel_fingerprint and _last_rooms_fingerprint to None so that _update_submenus() rebuilds into the freshly created containers; channel-dependent updates (update_filters, update_channel_options, _update_submenus) now run unconditionally when channel data exists — safe because each method has internal idempotency checks
🔄 gui/panels/messages_panel.py: update_channel_options() now includes an equality check on options dict to skip redundant .update() calls to the NiceGUI client on every 500ms timer tick

Impact

Channel discovery now survives transient BLE timeouts that are common on mobile connections
Atomic snapshot eliminates the threading race condition that caused channels to silently never appear
Browser close+reopen no longer loses channels — the single-instance timer race on the shared DashboardPage is fully mitigated
No breaking changes — all existing API methods retained, all other functionality unchanged

[1.9.5] - 2026-02-16 — Layout Fix: RX Log Table Responsive Sizing

Fixed

🛠 RX Log table did not adapt to panel/card size — The table used max-h-48 (a maximum height cap) instead of a responsive fixed height, causing it to remain small regardless of available space. Changed to h-40 which is overridden by the existing dashboard CSS to calc(100vh - 20rem) — the same responsive pattern used by the Messages panel
🛠 RX Log table did not fill card width — Added w-full class to the table element so it stretches to the full width of the parent card
🛠 RX Log card did not fill panel height — Added flex-grow class to the card container so it expands to fill the available panel space

Changed

🔄 gui/panels/rxlog_panel.py: Card classes 'w-full' → 'w-full flex-grow' (line 45); table classes 'text-xs max-h-48 overflow-y-auto' → 'w-full text-xs h-40 overflow-y-auto' (line 65)

Impact

RX Log table now fills the panel consistently on both desktop and mobile viewports
Layout is consistent with other panels (Messages, Contacts) that use the same h-40 responsive height pattern
No functional changes — all event handlers, callbacks, data bindings, logica and imports are identical to the input

[1.9.4] - 2026-02-16 — BLE Address Log Prefix & Entry Point Cleanup

Added

✅ BLE address prefix in log filename — Log file is now named <BLE_ADDRESS>_meshcore_gui.log (e.g. AA_BB_CC_DD_EE_FF_meshcore_gui.log) instead of the generic meshcore_gui.log. Makes it easy to identify which device produced which log file when running multiple instances
- New helper _sanitize_ble_address() strips literal: prefix and replaces colons with underscores
- New function configure_log_file(ble_address) updates LOG_FILE at runtime before the logger is initialised
- Rotated backups follow the same naming pattern automatically

Removed

❌ meshcore_gui/meshcore_gui.py — Redundant copy of main() that was never imported. All three entry points (meshcore_gui.py root, __main__.py, and meshcore_gui/meshcore_gui.py) contained near-identical copies of the same logic, causing changes to be missed (as demonstrated by this fix). __main__.py is now the single source of truth; root meshcore_gui.py is a thin wrapper that imports from it

Changed

🔄 config.py: Added _sanitize_ble_address() and configure_log_file(); version bumped to 1.9.4
🔄 __main__.py: Added config.configure_log_file(ble_address) call before any debug output
🔄 meshcore_gui.py (root): Reduced to 4-line wrapper importing main from __main__

Impact

Log files are now identifiable per BLE device
Single source of truth for main() eliminates future sync issues between entry points
Both startup methods (python meshcore_gui.py and python -m meshcore_gui) remain functional
No breaking changes — defaults and all existing behaviour unchanged

[1.9.3] - 2026-02-16 — Bugfix: Map Default Location & Payload Type Decoding

Fixed

🛠 Map centred on hardcoded Zwolle instead of device location — All Leaflet maps used magic-number coordinates (52.5, 6.0) as initial centre and fallback. These are now replaced by a single configurable constant DEFAULT_MAP_CENTER in config.py. Once the device reports a valid adv_lat/adv_lon, maps re-centre on the actual device position (existing behaviour, unchanged)
🛠 Payload type shown as raw integer — Payload type is now retrieved from the decoded payload and translated to human-readable text using MeshCoreDecoder functions, instead of displaying the raw numeric type value

Changed

🔄 config.py: Added DEFAULT_MAP_CENTER (default: (52.5168, 6.0830)) and DEFAULT_MAP_ZOOM (default: 9) constants in new MAP DEFAULTS section. Version bumped to 1.9.2
🔄 gui/panels/map_panel.py: Imports DEFAULT_MAP_CENTER and DEFAULT_MAP_ZOOM from config; ui.leaflet(center=...) uses config constants instead of hardcoded values
🔄 gui/route_page.py: Imports DEFAULT_MAP_CENTER and DEFAULT_MAP_ZOOM from config; fallback coordinates (or 52.5 / or 6.0) replaced by DEFAULT_MAP_CENTER[0] / [1]; zoom uses DEFAULT_MAP_ZOOM

Impact

Map default location is now a single-point-of-change in config.py
Payload type is displayed as readable text instead of a raw number
No breaking changes — all existing map behaviour (re-centre on device position, contact markers) unchanged

[1.9.2] - 2026-02-15 — CLI Parameters & Cleanup

Added

✅ --port=PORT CLI parameter — Web server port is now configurable at startup (default: 8081). Allows running multiple instances simultaneously on different ports
✅ --ble-pin=PIN CLI parameter — BLE pairing PIN is now configurable at startup (default: 123456). Eliminates the need to edit config.py for devices with a non-default PIN, and works in systemd service files
✅ Per-device log file — Debug log file now includes the BLE address in its filename (e.g. F0_9E_9E_75_A3_01_meshcore_gui.log), so multiple instances log to separate files

Fixed

🛠 BLE PIN not applied from CLI — ble/worker.py imported BLE_PIN as a constant at module load time (from config import BLE_PIN), capturing the default value "123456" before CLI parsing could override config.BLE_PIN. Changed to runtime access via config.BLE_PIN so the --ble-pin parameter is correctly passed to the BLE agent

Removed

❌ Redundant meshcore_gui/meshcore_gui.py — This file was a near-identical copy of both meshcore_gui.py (top-level) and meshcore_gui/__main__.py, but was never imported or referenced. Removed to eliminate maintenance risk. The two remaining entry points cover all startup methods: python meshcore_gui.py and python -m meshcore_gui

Impact

Multiple instances can run side-by-side with different ports, PINs and log files
Service deployments no longer require editing config.py — all runtime settings via CLI
No breaking changes — all defaults are unchanged

[1.9.1] - 2026-02-14 — Bugfix: Dual Reconnect Conflict

Fixed

🛠 Library reconnect interfered with application reconnect — The meshcore library's internal auto_reconnect (visible in logs as "Attempting reconnection 1/3") ran a fast 3-attempt reconnect cycle without bond cleanup. This prevented the application's own reconnect_loop (which does remove_bond() + backoff) from succeeding, because BlueZ retained a stale bond → "failed to discover service"

Changed

🔄 ble/worker.py: Set auto_reconnect=False in both MeshCore.create_ble() call sites (_connect() and _create_fresh_connection()), so only the application's bond-aware reconnect_loop handles reconnection
🔄 ble/worker.py: Added "failed to discover" and "service discovery" to disconnect detection keywords for defensive coverage

Impact

Eliminates the ~9 second wasted library reconnect cycle after every BLE disconnect
Application's reconnect_loop (with bond cleanup) now runs immediately after disconnect detection
No breaking changes — the application reconnect logic was already fully functional

[1.9.0] - 2026-02-14 — BLE Connection Stability

Added

✅ Built-in BLE PIN agent — New ble/ble_agent.py registers a D-Bus agent with BlueZ to handle PIN pairing requests automatically. Eliminates the need for external bt-agent.service and bluez-tools package
- Uses dbus_fast (already a dependency of bleak, no new packages)
- Supports RequestPinCode, RequestPasskey, DisplayPasskey, RequestConfirmation, AuthorizeService callbacks
- Configurable PIN via BLE_PIN in config.py (default: 123456)
✅ Automatic bond cleanup — New ble/ble_reconnect.py provides remove_bond() function that removes stale BLE bonds via D-Bus, equivalent to bluetoothctl remove <address>. Called automatically on startup and before each reconnect attempt
✅ Automatic reconnect after disconnect — BLEWorker main loop now detects BLE disconnects (via connection error exceptions) and automatically triggers a reconnect sequence: bond removal → linear backoff wait → fresh connection → re-wire handlers → reload device data
- Configurable via RECONNECT_MAX_RETRIES (default: 5) and RECONNECT_BASE_DELAY (default: 5.0s)
- After all retries exhausted: waits 60s then starts a new retry cycle (infinite recovery)
✅ Generic install script — install_ble_stable.sh auto-detects user, project directory, venv path and entry point to generate systemd service and D-Bus policy. Supports --uninstall flag

Changed

🔄 ble/worker.py — _async_main() rewritten with three phases: (1) start PIN agent, (2) remove stale bond, (3) connect + main loop with disconnect detection. Reconnect logic re-wires all event handlers and reloads device data after successful reconnection
🔄 config.py — Added BLE_PIN, RECONNECT_MAX_RETRIES, RECONNECT_BASE_DELAY constants

Removed

❌ bt-agent.service dependency — No longer needed; PIN pairing is handled by the built-in agent
❌ bluez-tools system package — No longer needed
❌ ~/.meshcore-ble-pin file — No longer needed
❌ Manual bluetoothctl remove before startup — Handled automatically
❌ ExecStartPre in systemd service — Bond cleanup is internal

Impact

Zero external dependencies for BLE pairing on Linux
Automatic recovery from the T1000e ~2 hour BLE disconnect issue
No manual intervention needed after BLE connection loss
Single systemd service (meshcore-gui.service) manages everything
No breaking changes to existing functionality

[1.8.0] - 2026-02-14 — DRY Message Construction & Archive Layout Unification

Fixed

🛠 Case-sensitive prefix matching — get_contact_name_by_prefix() and get_contact_by_prefix() in shared_data.py failed to match path hashes (uppercase, e.g. 'B8') against contact pubkeys (lowercase, e.g. 'b8a3f2...'). Added .lower() to both sides of the comparison, consistent with _resolve_path_names() which already had it
🛠 Route page 404 from archive — Archive page linked to /route/{hash} but route was registered as /route/{msg_index:int}, causing a JSON parse error for hex hash strings. Route parameter changed to str with 3-strategy lookup (index → memory hash → archive fallback)
🛠 Three entry points out of sync — meshcore_gui.py (root), meshcore_gui/meshcore_gui.py (inner) and meshcore_gui/__main__.py had diverging route registrations. All three now use identical /route/{msg_key} with str parameter

Changed

🔄 core/models.py — DRY factory methods and formatting
- Message.now_timestamp(): static method replacing 7× hardcoded datetime.now().strftime('%H:%M:%S') across events.py and commands.py
- Message.incoming(): classmethod factory for received messages (direction='in', auto-timestamp)
- Message.outgoing(): classmethod factory for sent messages (sender='Me', direction='out', auto-timestamp)
- Message.format_line(channel_names): single-line display formatting ("12:34:56 ← [Public] [2h✓] PE1ABC: Hello mesh!"), replacing duplicate inline formatting in messages_panel.py and archive_page.py
🔄 ble/events.py — 4× Message(...) constructors replaced by Message.incoming(); datetime import removed
🔄 ble/commands.py — 3× Message(...) constructors replaced by Message.outgoing(); datetime import removed
🔄 gui/panels/messages_panel.py — 15 lines inline formatting replaced by single msg.format_line(channel_names) call
🔄 gui/archive_page.py — Layout unified with main page
- Multi-row card layout replaced by single-line msg.format_line() in monospace container (same style as main page)
- DM added to channel filter dropdown (post-filter on channel is None)
- Message click opens /route/{message_hash} in new tab (was: no click handler on archive messages)
- Removed _render_message_card() (98 lines) and _render_archive_route() (75 lines)
- Removed RouteBuilder dependency and TYPE_LABELS import
- File reduced from 445 to 267 lines
🔄 gui/route_page.py — render(msg_index: int) → render(msg_key: str) with 3-strategy message lookup: (1) numeric index from in-memory list, (2) hash match in memory, (3) archive.get_message_by_hash() fallback
🔄 services/message_archive.py — New method get_message_by_hash(hash) for single-message lookup by packet hash
🔄 __main__.py + meshcore_gui.py (both) — Route changed from /route/{msg_index} (int) to /route/{msg_key} (str)

Impact

DRY: timestamp formatting 7→1 definition, message construction 7→2 factories, line formatting 2→1 method
Archive page visually consistent with main messages panel (single-line, monospace)
Archive messages now clickable to open route visualization (was: only in-memory messages)
Case-insensitive prefix matching fixes path name resolution for contacts with uppercase path hashes
No breaking changes to BLE protocol handling, dedup, bot, or data storage

Known Limitations

DM filter in archive uses post-filtering (query without channel filter + filter on channel is None); becomes exact when query_messages() gets native DM support

Parked for later

Multi-path tracking (enrich RxLogEntry with multiple path observations)
Events correlation improvements (only if proven data loss after .lower() fix)

[1.7.0] - 2026-02-13 — Archive Channel Name Persistence

Added

✅ Channel name stored in archive — Messages now persist channel_name alongside the numeric channel index in <ADDRESS>_messages.json, so archived messages retain their human-readable channel name even when the device is not connected
- Message dataclass: new field channel_name: str (default "", backward compatible)
- SharedData.add_message(): automatically resolves channel_name from the live channels list when not already set (new helper _resolve_channel_name())
- MessageArchive.add_message(): writes channel_name to the JSON dict
✅ Archive channel selector built from archived data — Channel filter dropdown on /archive now populated via SELECT DISTINCT channel_name on the archive instead of the live BLE channels list
- New method MessageArchive.get_distinct_channel_names() returns sorted unique channel names from stored messages
- Selector shows only channels that actually have archived messages
✅ Archive filter on channel name — MessageArchive.query_messages() parameter changed from channel: Optional[int] to channel_name: Optional[str] (exact match on name string)

Changed

🔄 core/models.py: Added channel_name field to Message dataclass and from_dict()
🔄 core/shared_data.py: add_message() resolves channel name; added _resolve_channel_name() helper
🔄 services/message_archive.py: channel_name persisted in JSON; query_messages() filters by name; new get_distinct_channel_names() method
🔄 gui/archive_page.py: Channel selector built from archive.get_distinct_channel_names(); filter state changed from _channel_filter (int) to _channel_name_filter (str); message cards show channel_name directly from archive

Fixed

🛠 Main page empty after startup — After a restart the messages panel showed no messages until new live BLE traffic arrived. SharedData.load_recent_from_archive() now loads up to 100 recent archived messages during the cache-first startup phase, so historical messages are immediately visible
- New method SharedData.load_recent_from_archive(limit) — reads from MessageArchive.query_messages() and populates the in-memory list without re-archiving
- BLEWorker._apply_cache() calls load_recent_from_archive() at the end of cache loading

Impact

Archived messages now self-contained — channel name visible without live BLE connection
Main page immediately shows historical messages after startup (no waiting for live BLE traffic)
Backward compatible — old archive entries without channel_name fall back to "Ch <idx>"
No breaking changes to existing functionality

[1.6.0] - 2026-02-13 — Dashboard Layout Consolidation

Changed

🔄 Messages panel consolidated — Filter checkboxes (DM + channels) and message input (text field, channel selector, Send button) are now integrated into the Messages panel, replacing the separate Filter and Input panels
- DM + channel checkboxes displayed centered in the Messages header row, between the "💬 Messages" label and the "📚 Archive" button
- Message input row (text field, channel selector, Send button) placed below the message list within the same card
- messages_panel.py: Constructor now accepts put_command callable; added update_filters(data), update_channel_options(channels) methods and channel_filters, last_channels properties (all logic 1:1 from FilterPanel/InputPanel); update() signature unchanged
🔄 Actions panel expanded — BOT toggle checkbox moved from Filter panel to Actions panel, below the Refresh/Advert buttons
- actions_panel.py: Constructor now accepts set_bot_enabled callable; added update(data) method for BOT state sync; _on_bot_toggle() logic 1:1 from FilterPanel
🔄 Dashboard layout simplified — Centre column reduced from 4 panels (Map → Input → Filter → Messages) to 2 panels (Map → Messages)
- dashboard.py: FilterPanel and InputPanel no longer rendered; all dependencies rerouted to MessagesPanel and ActionsPanel; _update_ui() call-sites updated accordingly

Removed (from layout, files retained)

❌ Filter panel no longer rendered as separate panel — filter_panel.py retained in codebase but not instantiated in dashboard
❌ Input panel no longer rendered as separate panel — input_panel.py retained in codebase but not instantiated in dashboard

Impact

Cleaner, more compact dashboard: 2 fewer panels in the centre column
All functionality preserved — message filtering, send, BOT toggle, archive all work identically
No breaking changes to BLE, services, core or other panels

[1.5.0] - 2026-02-11 — Room Server Support, Dynamic Channel Discovery & Contact Management

Added

✅ Room Server panel — Dedicated per-room-server message panel in the centre column below Messages. Each Room Server (type=3 contact) gets its own ui.card() with login/logout controls and message display
- Click a Room Server contact to open an add/login dialog with password field
- After login: messages are displayed in the room card; send messages directly from the room panel
- Password row + login button automatically replaced by Logout button after successful login
- Room Server author attribution via signature field (txt_type=2) — real message author is resolved from the 4-byte pubkey prefix, not the room server pubkey
- New panel: gui/panels/room_server_panel.py — per-room card management with login state tracking
✅ Room Server password store — Passwords stored outside the repository in ~/.meshcore-gui/room_passwords/<ADDRESS>.json
- New service: services/room_password_store.py — JSON-backed persistent password storage per BLE device, analogous to PinStore
- Room panels are restored from stored passwords on app restart
✅ Dynamic channel discovery — Channels are now auto-discovered from the device at startup via get_channel() BLE probing, replacing the hardcoded CHANNELS_CONFIG
- Single-attempt probe per channel slot with early stop after 2 consecutive empty slots
- Channel name and encryption key extracted in a single pass (combined discovery + key loading)
- Configurable channel caching via CHANNEL_CACHE_ENABLED (default: False — always fresh from device)
- MAX_CHANNELS setting (default: 8) controls how many slots are probed
✅ Individual contact deletion — 🗑️ delete button per unpinned contact in the contacts list, with confirmation dialog
- New command: remove_single_contact in BLE command handler
- Pinned contacts are protected (no delete button shown)
✅ "Also delete from history" option — Checkbox in the Clean up confirmation dialog to also remove locally cached contact data

✅ Room Server protocol research — RoomServer_Companion_App_Onderzoek.md documents the full companion app message flow (login, push protocol, signature mechanism, auto_message_fetching)

Changed

🔄 config.py: Removed CHANNELS_CONFIG constant; added MAX_CHANNELS (default: 8) and CHANNEL_CACHE_ENABLED (default: False)
🔄 ble/worker.py: Replaced hardcoded channel loading with _discover_channels() method; added _try_get_channel_info() helper; _apply_cache() respects CHANNEL_CACHE_ENABLED setting; removed _load_channel_keys() (integrated into discovery pass)
🔄 ble/commands.py: Added login_room, send_room_msg and remove_single_contact command handlers
🔄 gui/panels/contacts_panel.py: Contact click now dispatches by type — type=3 (Room Server) opens room dialog, others open DM dialog; added on_add_room callback parameter; added 🗑️ delete button per unpinned contact
🔄 gui/panels/messages_panel.py: Room Server messages filtered from general message view via _is_room_message() with prefix matching; update() accepts room_pubkeys parameter
🔄 gui/dashboard.py: Added RoomServerPanel in centre column; _update_ui() passes room_pubkeys to Messages panel; added _on_add_room_server callback
🔄 gui/panels/filter_panel.py: Channel filter checkboxes now built dynamically from discovered channels (no hardcoded references)
🔄 services/bot.py: Removed stale comment referencing hardcoded channels

Fixed

🛠 Room Server messages appeared as DM — Messages from Room Servers (txt_type=2) were displayed in the general Messages panel as direct messages. They are now filtered out and shown exclusively in the Room Server panel
🛠 Historical room messages not shown after login — Post-login fetch loop was polling get_msg() before room server had time to push messages over LoRa RF (10–75s per message). Removed redundant fetch loop; the library's auto_message_fetching handles MESSAGES_WAITING events correctly and event-driven
🛠 Author attribution incorrect for room messages — Room server messages showed the room server name as sender instead of the actual message author. Now correctly resolved from the signature field (4-byte pubkey prefix) via contact lookup

Impact

Room Servers are now first-class citizens in the GUI with dedicated panels
Channel configuration no longer requires manual editing of config.py
Contact list management is more granular with per-contact deletion
No breaking changes to existing functionality (messages, DM, map, archive, bot, etc.)

[1.4.0] - 2026-02-09 — SDK Event Race Condition Fix

Fixed

🛠 BLE startup delay of ~2 minutes eliminated — The meshcore Python SDK (commands/base.py) dispatched device response events before wait_for_events() registered its subscription. On busy networks with frequent RX_LOG_DATA events, this caused send_device_query() and get_channel() to fail repeatedly with no_event_received, wasting 110+ seconds in timeouts

Changed

📄 meshcore SDK commands/base.py: Rewritten send() method to subscribe to expected events before transmitting the BLE command (subscribe-before-send pattern), matching the approach used by the companion apps (meshcore.js, iOS, Android). Submitted upstream as meshcore_py PR #52

Impact

Startup time reduced from ~2+ minutes to ~10 seconds on busy networks
All BLE commands (send_device_query, get_channel, get_bat, send_appstart, etc.) now succeed on first attempt instead of requiring multiple retries
No changes to meshcore_gui code required — the fix is entirely in the meshcore SDK

Temporary Installation

Until the fix is merged upstream, install the patched meshcore SDK:

pip install --force-reinstall git+https://github.com/PE1HVH/meshcore_py.git@fix/event-race-condition

[1.3.2] - 2026-02-09 — Bugfix: Bot Device Name Restoration After Restart

Fixed

🛠 Bot device name not properly restored after restart/crash — After a restart or crash with bot mode previously active, the original device name was incorrectly stored as the bot name (e.g. NL-OV-ZWL-STDSHGN-WKC Bot) instead of the real device name (e.g. PE1HVH T1000e). The original device name is now correctly preserved and restored when bot mode is disabled

Changed

🔄 commands.py: set_bot_name handler now verifies that the stored original name is not already the bot name before saving
🔄 shared_data.py: original_device_name is only written when it differs from BOT_DEVICE_NAME to prevent overwriting with the bot name on restart

[1.3.1] - 2026-02-09 — Bugfix: Auto-add AttributeError

Fixed

🛠 Auto-add error on first toggle — Setting auto-add for the first time raised AttributeError: 'telemetry_mode_base'. The set_manual_add_contacts() SDK call now handles missing telemetry_mode_base attribute gracefully

Changed

🔄 commands.py: set_auto_add handler wraps set_manual_add_contacts() call with attribute check and error handling for missing telemetry_mode_base

[1.3.0] - 2026-02-08 — Bot Device Name Management

Added

✅ Bot device name switching — When the BOT checkbox is enabled, the device name is automatically changed to a configurable bot name; when disabled, the original name is restored
- Original device name is saved before renaming so it can be restored on BOT disable
- Device name written to device via BLE set_name() SDK call
- Graceful handling of BLE failures during name change
✅ BOT_DEVICE_NAME constant in config.py — Configurable fixed device name used when bot mode is active (default: ;NL-OV-ZWL-STDSHGN-WKC Bot)

Changed

🔄 config.py: Added BOT_DEVICE_NAME constant for bot mode device name
🔄 bot.py: Removed hardcoded BOT_NAME prefix ("Zwolle Bot") from bot reply messages — bot replies no longer include a name prefix
🔄 filter_panel.py: BOT checkbox toggle now triggers device name save/rename via command queue
🔄 commands.py: Added set_bot_name and restore_name command handlers for device name switching
🔄 shared_data.py: Added original_device_name field for storing the pre-bot device name

Removed

❌ BOT_NAME constant from bot.py — bot reply prefix removed; replies no longer prepend a bot display name

[1.2.0] - 2026-02-08 — Contact Maintenance Feature

Added

✅ Pin/Unpin contacts (Iteration A) — Toggle to pin individual contacts, protecting them from bulk deletion
- Persistent pin state stored in ~/.meshcore-gui/cache/<ADDRESS>_pins.json
- Pinned contacts visually marked with yellow background
- Pinned contacts sorted to top of contact list
- Pin state survives app restart
- New service: services/pin_store.py — JSON-backed persistent pin storage
✅ Bulk delete unpinned contacts (Iteration B) — Remove all unpinned contacts from device in one action
- "🧹 Clean up" button in contacts panel with confirmation dialog
- Shows count of contacts to be removed vs. pinned contacts kept
- Progress status updates during removal
- Automatic device resync after completion
- New service: services/contact_cleaner.py — ContactCleanerService with purge statistics
✅ Auto-add contacts toggle (Iteration C) — Control whether device automatically adds new contacts from mesh adverts
- "📥 Auto-add" checkbox in contacts panel (next to Clean up button)
- Syncs with device via set_manual_add_contacts() SDK call
- Inverted logic handled internally (UI "Auto-add ON" = set_manual_add_contacts(false))
- Optimistic update with automatic rollback on BLE failure
- State synchronized from device on each GUI update cycle

Changed

🔄 contacts_panel.py: Added pin checkbox per contact, purge button, auto-add toggle, DM dialog (all existing functionality preserved)
🔄 commands.py: Added purge_unpinned and set_auto_add command handlers
🔄 shared_data.py: Added auto_add_enabled field with thread-safe getter/setter
🔄 protocols.py: Added set_auto_add_enabled and is_auto_add_enabled to Writer and Reader protocols
🔄 dashboard.py: Passes PinStore and set_auto_add_enabled callback to ContactsPanel
🔄 UI language: All Dutch strings in contacts_panel.py and commands.py translated to English

Fixed

🛠 Route table names and IDs not displayed — Route tables in both current messages (RoutePage) and archive messages (ArchivePage) now correctly show node names and public key IDs for sender, repeaters and receiver

Changed

🔄 CHANGELOG.md: Corrected version numbering to semantic versioning, fixed inaccurate references (archive button location, filter state persistence)
🔄 README.md: Added Message Archive feature, updated project structure, configuration table and architecture diagram
🔄 MeshCore_GUI_Design.docx: Added ArchivePage, MessageArchive, Models components; updated project structure, protocols, configuration and version history

[1.1.0] - 2026-02-07 — Archive Viewer Feature

Added

✅ Archive Viewer Page (/archive) — Full-featured message archive browser
- Pagination (50 messages per page, configurable)
- Channel filter dropdown (All + configured channels)
- Time range filter (24h, 7d, 30d, 90d, All time)
- Text search (case-insensitive)
- Filter state stored in instance variables (reset on page reload)
- Message cards with same styling as main messages panel
- Clickable messages for route visualization (where available)
- 💬 Reply functionality — Expandable reply panel per message
- 🗺️ Inline route table — Expandable route display per archive message with sender, repeaters and receiver (names, IDs, node types)
- (Note: Reply panels and inline route tables removed in v1.8.0, replaced by click-to-route navigation via message hash)

✅ MessageArchive.query_messages() method
- Filter by: time range, channel, text search, sender
- Pagination support (limit, offset)
- Returns tuple: (messages, total_count)
- Sorting: Newest first
✅ UI Integration
- "📚 Archive" button in Messages panel header (opens in new tab)
- Back to Dashboard button in archive page

✅ Reply Panel
- Expandable reply per message (💬 Reply button)
- Pre-filled with @sender mention
- Channel selector
- Send button with success notification
- Auto-close expansion after send

Changed

🔄 SharedData.get_snapshot(): Now includes 'archive' field
🔄 MessagesPanel: Added archive button in header row
🔄 Both entry points (__main__.py and meshcore_gui.py): Register /archive route

Performance

Query: ~10ms for 10k messages with filters
Memory: ~10KB per page (50 messages)
No impact on main UI (separate page)

Known Limitations

~~Route visualization only works for messages in recent buffer (last 100)~~ — Fixed in v1.8.0: archive messages now support click-to-route via get_message_by_hash() fallback
Text search is linear scan (no indexing yet)
Sender filter exists in API but not in UI yet

[1.0.3] - 2026-02-07 — Critical Bugfix: Archive Overwrite Prevention

Fixed

🛠 CRITICAL: Fixed bug where archive was overwritten instead of appended on restart
🛠 Archive now preserves existing data when read errors occur
🛠 Buffer is retained for retry if existing archive cannot be read

Changed

🔄 _flush_messages(): Early return on read error instead of overwriting
🔄 _flush_rxlog(): Early return on read error instead of overwriting
🔄 Better error messages for version mismatch and JSON decode errors

Details

Problem: If the existing archive file had a JSON parse error or version mismatch, the flush operation would proceed with existing_messages = [], effectively overwriting all historical data with only the new buffered messages.

Solution: The flush methods now:

Try to read existing archive first
If read fails (JSON error, version mismatch, IO error), abort the flush
Keep buffer intact for next retry
Only clear buffer after successful write

Impact: No data loss on restart or when archive files have issues.

Testing

✅ Added test_append_on_restart_not_overwrite() integration test
✅ Verifies data is appended across multiple sessions
✅ All existing tests still pass

[1.0.2] - 2026-02-07 — RxLog message_hash Enhancement

Added

✅ message_hash field added to RxLogEntry model
✅ RxLog entries now include message_hash for correlation with messages
✅ Archive JSON includes message_hash in rxlog entries

Changed

🔄 events.py: Restructured on_rx_log() to extract message_hash before creating RxLogEntry
🔄 message_archive.py: Updated rxlog archiving to include message_hash field
🔄 Tests updated to verify message_hash persistence

Benefits

Correlation: Link RX log entries to their corresponding messages
Analysis: Track which packets resulted in messages
Debugging: Better troubleshooting of packet processing

[1.0.1] - 2026-02-07 — Entry Point Fix

Fixed

✅ meshcore_gui.py (root entry point) now passes ble_address to SharedData
✅ Archive works correctly regardless of how application is started

Changed

🔄 Both entry points (meshcore_gui.py and meshcore_gui/__main__.py) updated

[1.0.0] - 2026-02-07 — Message & Metadata Persistence

Added

✅ MessageArchive class for persistent storage
✅ Configurable retention periods (MESSAGE_RETENTION_DAYS, RXLOG_RETENTION_DAYS, CONTACT_RETENTION_DAYS)
✅ Automatic daily cleanup of old data
✅ Batch writes for performance
✅ Thread-safe with separate locks
✅ Atomic file writes
✅ Contact retention in DeviceCache
✅ Archive statistics API
✅ Comprehensive tests (20+ unit, 8+ integration)
✅ Full documentation

Storage Locations

~/.meshcore-gui/archive/<ADDRESS>_messages.json
~/.meshcore-gui/archive/<ADDRESS>_rxlog.json

Requirements Completed

R1: All incoming messages persistent ✅
R2: All incoming RxLog entries persistent ✅
R3: Configurable retention ✅
R4: Automatic cleanup ✅
R5: Backward compatibility ✅
R6: Contact retention ✅
R7: Archive stats API ✅
Fix3: Leaflet asset injection is now per page render instead of process-global, and browser bootstrap now retries until the host element, Leaflet runtime, and MeshCore panel runtime are all available. This fixes blank map containers caused by missing or late-loaded JS/CSS assets.
Fix5: Removed per-snapshot map invalidate calls, stopped forcing a default dark theme during map bootstrap, and added client-side interaction/resize guards so zooming stays responsive and the theme no longer jumps back during status-loop updates.

2026-03-09 map hotfix v2

regular map snapshots no longer carry theme state
explicit theme changes are now handled only via the dedicated theme channel
initial map render now sends an ensure_map command plus an immediate theme sync
added no-op ensure_map handling in the Leaflet runtime to avoid accidental fallback behaviour

96 KiB Raw Permalink Blame History Unescape Escape

CHANGELOG

[1.22.1] - 2026-04-27

Added

Changed

Impact

[1.22.0] - 2026-04-21

Added

Changed

Impact

Rationale

Added

Changed

Not changed (deliberately out of scope)

[1.20.2] - 2026-04-19

Fixed

Changed

[1.20.1] - 2026-04-15

Fixed

[1.20.0] - 2026-04-10

Changed

[1.19.0] - 2026-04-06

FIXED

ADDED

CHANGED

IMPACT

[1.18.1] - 2026-04-05

FIXED

CHANGED

[1.18.0] - 2026-04-05

ADDED

CHANGED

IMPACT

RATIONALE

[1.17.1] - 2026-04-04

FIXED

CHANGED

RATIONALE

IMPACT

[1.17.0] - 2026-04-04

ADDED

CHANGED

RATIONALE

IMPACT

[1.16.0] - 2026-04-04

ADDED

CHANGED

RATIONALE

IMPACT

[1.15.0] - 2026-03-16

ADDED

CHANGED

IMPACT

RATIONALE

[1.14.3] - 2026-03-16 — BBS !h / !help NameError fix

Fixed

[1.14.2] - 2026-03-16 — BBS whitelist fix: !bbs channel hook in on_rx_log

Fixed

[1.14.1] - 2026-03-16 — BBS test corrections

Changed

[1.14.0] - 2026-03-14 — BBS (Bulletin Board System)

Added

Changed

Storage

[1.13.5] - 2026-03-14 — Route back-button and map popup flicker fixes

Fixed

Changed

Impact

[1.13.4] - 2026-03-12 — Room Server message classification fix

Fixed

Changed

Impact

[1.13.3] - 2026-03-12 — Active Panel Timer Gating

Changed

Fixed

Impact

[1.13.2] - 2026-03-11 — Map Display Bugfix

Fixed

Changed

Impact

96 KiB

Raw Permalink Blame History