Repository navigation
Develop - #4
Merged
Merged
Develop#4
Conversation
Add plan, research, data model, contracts (settings form, telemetry window, lib modules, audio) and quickstart validation guide. Sensor speed is taken as m/s, as in DFM Speed Announcer v2.1. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Features branch from develop as feature/NNN-name, releases go through release/x.y.z to main and are tagged <script>-v<version>. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Aaron supplied a car HUD speedometer as the style to follow. The photo isn't committed (third-party product image); its elements are described instead: dark round face, ~270 deg sweep, rim ticks and numbers, glowing value arc, red overspeed zone, big center number, subtle max marker, and secondary rows. Adds a per-size layout (full dial when double, compact arc when single; exact window sizes to be measured in planning) and updates FR-014/014a/016/017 and the gauge acceptance scenarios. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NZc7boqR33nD9xyYsGNLBg
Aaron: the gauge assumes the DS-24 II display; older transmitters are not supported for the gauge. JETI lists the II display as 4", 480x480 px, not the original DS-24's 320x240. - Spec 001: DS-24 II / DC-24 II is the gauge target; new FR-013a shows a notice on other transmitters while callouts/warnings keep working; window sizes must be measured with the probe and recorded in plan.md. - tools/probe/PROBE.lua: dev tool that prints the small/large window and form canvas sizes plus font heights (copied to the emulator's Apps). - docs/jeti-api-notes.md: target hardware section with a table to fill. - CLAUDE.md, README: note the target and the probe. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NZc7boqR33nD9xyYsGNLBg
Measured in the emulator (fw 6.04) with PROBE.lua: small 157x60, large 157x127. Lua on the II keeps the original's window sizes, not 480x480. - docs/jeti-api-notes.md, spec 001, CLAUDE.md: record sizes; layouts sized for them (compact arc when single, ~110-120 px dial with corner rows when double). - Spec 001: add a full-screen window (dial + side panel, like the reference). FR-013: two windows, the per-app maximum: single/double (auto) and full screen. - PROBE.lua 1.1: MODE switch; MODE 2 measures full-screen sizes 3 and 4. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NZc7boqR33nD9xyYsGNLBg
Measured with PROBE.lua 1.1 in the emulator: full-screen sizes 3 and 4 are both 320x260 on the II, so the spec drops the status-bar option. Spec 001, API notes and CLAUDE.md now list all three window sizes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NZc7boqR33nD9xyYsGNLBg
The reference photo lives at docs/vendor/gauge-reference.jpg on the dev PC (gitignored, third-party product image) so Claude Code can open it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NZc7boqR33nD9xyYsGNLBg
41 tasks in 8 phases, organized by user story. US1 (callouts) is the MVP; MANUAL tasks mark the emulator and transmitter checks. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ous/jeti-lua-apps into feature/001-speed-gauge
tools/emulator/sensors.json for LeonAirRC's Emulator Telemetry app (installed separately, emulator only): MSpeed 450 Velocity 0-450 km/h on P5 and GPS Speed 0-300 km/h on P6. Control fully down = sensor lost. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NZc7boqR33nD9xyYsGNLBg
Per research R3 (API doc + JETI forum), sensors send speed in m/s, so the emulated MSpeed 450 and GPS now report m/s (0-125 and 0-83.3, about 450 and 300 km/h). R3 gains two notes: call system.getSensor* via `system` each time (Emulator Telemetry swaps them) and the emulator sensor ranges. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NZc7boqR33nD9xyYsGNLBg
Plan follows the spec's visual design reference and measured DS-24 II window sizes: dark face, numbered scale, overspeed zone, compact / round / full-screen layouts, size-0 pilot-sized window with a fallback setting, colors without red/orange, and a notice on other transmitters (research R6-R8, R12). Quickstart uses tools/emulator/sensors.json. tasks.md regenerated: 46 tasks; US3 rebuilt for the new layouts. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
New app src/Apps/AG-SpdGa.lua with shared modules ag_dens (ISA density factor) and ag_gauge (dial geometry and drawing), per spec 001: - US1/US2: DFM v2.1 callout timing, stall/overspeed/airspeed-alive warnings with vibration, startup announcement - US3: compact, round and full-screen gauge layouts after the design reference, layout cache, spike-filtered session max, DS-24 II notice - US4: optional true-airspeed correction from elevation and temperature - US5: grouped settings, units conversion, threshold-order warning The max filter also counts a reading held for 1 s; without it a steady speed never registered (research R5). Reused DFM WAVs are credited in src/Apps/AG-SpdGa/CREDITS.txt. PROBE.lua gains a size-0 mode. Verified: check.py, LuaLS (no problems), tests/test_ag_dens.lua and a 48-check mock-API run under Lua 5.3 (fengari). Emulator and transmitter checks (T005, T017, T021, T029, T035, T039, T046) are still open. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Fixes found in the JETI Studio emulator (DS-24 II, firmware 6.04): - Callouts never played: the ms time counter is a wrapped, negative 32-bit value. All time checks now compare differences. - Single and double windows were clipped: the II desktop draws each window's title bar inside the canvas, hiding ~25 px at the bottom. Layouts draw into h - TITLE_H; the compact layout is reworked for ~34 px. - The dial face drew past the window border (the renderer does not clip): ag_gauge.face can now cut the circle flat at maxY. - Full screen uses size 4, as DFM-InsP does; size 3 is covered by the desktop's model tile. Refinements from the user's review: - Full-screen window titled "Speed Gauge"; larger scale and unit text on the full-screen dial. - Max marker thicker and longer; new Yellow preset is its default. - Glow bands fading toward the center behind the value arc and the overspeed zone (ag_gauge.arc gains alpha). - Past the overspeed mark the value arc switches to the overspeed color. Spec (FR-013, FR-014, FR-017), research, data model, contracts and quickstart updated to match. docs/jeti-api-notes.md records the visible window sizes, scale, fonts, device string and Emulated Telemetry behavior. PROBE.lua gains a size-0 mode, a ruler, an isPlayback log and an empty-title test. Verified: check.py, LuaLS, test_ag_dens.lua and the mock-API harness (51 checks including wrapped clock, visible-area bounds and overspeed color); layouts reviewed in the emulator. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Full-screen gauge: - MAX moves from the side panel to the open bottom of the dial, centered under the speed, without a unit. - Side panel rows are spread over the visible height (about 10 px apart): STALL, OVERSPEED, AIR DENSITY (+N%, OFF or GPS, always shown), and while correction is on ELEVATION (as set, ft or m) and RAW SENSOR (uncorrected speed). Fix: when the sensor is lost, RAW SENSOR shows "---" instead of keeping its last value. Spec, plan, contract and quickstart brought up to date with the review decisions: glow and overspeed coloring, yellow max marker, visible window sizes and title bar, compact layout, settings labels, and the now-verified emulator assumptions. Verified: check.py, LuaLS, and the mock-API harness (72 checks, including text-overlap and visible-area checks for every layout). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Measured in the emulator with system.getCPU(): the CPU figure in the app list is the highest per-call budget use (the script is killed at 100%), not an average. Start-up 24%, loop 0-1%, small/double draw up to 18%, full-screen draw up to 43% (down from 66%). - Glow bands use a coarse 10-degree dial and fewer, wider bands; visible edges keep 5-degree steps (10 degrees looked blocky on the II). - Drawing split into fixed parts, moving parts and scale labels. - Off-screen image caching was tried and removed: on the II emulator the image showed nothing and the figure did not drop. - SC-007 revised to "worst single call below 50%, full screen included"; the user accepted the full-screen glow as designed. Plan, research, quickstart and tasks updated (T030 closed). docs/jeti-api-notes.md records what the CPU figure means, how to log it, that alpha glow (not fills) is the cost, and that off-screen images don't work. PROBE.lua prints Lua memory every 5 s. Verified: check.py, LuaLS, mock-API harness (73 checks). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Aaron compared generated samples (Lessac, Amy, Ryan) and chose Amy as the clearest. - Clarification session 2026-09-28; User Story 6 (P3) with acceptance scenarios; edge cases for partial installs and number+unit splicing. - FR-030-FR-037: number files 0-500, unit and warning phrases; Voice setting (Speed Gauge / Transmitter); startup check with per-callout fallback to the transmitter voice and DFM's recordings (never silent); generator in tools/voice/; generated files not committed (Amy model is CC BY-SA, training data license undocumented); Piper/Amy credits. - FR-010 and assumptions updated; SC-009 (timing, from measured Amy durations) and SC-010 (no silent callouts without voice files). - .gitignore: src/Apps/AG-SpdGa/voice/. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NZc7boqR33nD9xyYsGNLBg
Aaron asked for tighter, realistic limits (proposed -10..8,000 ft and -10..120 F). Checking real airfields showed -10..8,000 ft would exclude Leadville (9,934 ft), Telluride (9,078 ft), Furnace Creek (-208 ft), Thermal (-114 ft) and Schiphol (-11 ft); he chose -300..10,000 ft. For temperature he chose the wider -20..130 F for margin. - Clarifications session 2026-09-30. - FR-020: limits table (imperial and metric: -90..3,050 m, -29..54 C); standard temperature stays within limits across the elevation range. - Edge case: values saved outside the limits are clamped on load. - data-model.md: elev/temp ranges updated to match. Code (form limits, unit-switch clamps, clamp on load) still uses the old limits and needs a follow-up change. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NZc7boqR33nD9xyYsGNLBg
Aaron: Speed Gauge should be able to use the MSpeed's temperature, with the sensor selectable or the temperature set manually. - Clarification recorded; User Story 4 gains scenarios 7-9 (sensor reading, fallback on loss, no jitter). - FR-038-FR-043: Temperature source Standard/Manual/Sensor (default Standard); sensor list limited to temperature values, stored by identity; readings used only if valid and within FR-020 limits, else standard temperature; read at most every 5 s with 1 C hysteresis; settings show the temperature and its source; stall/landing checks and GPS unaffected. - Settings table, edge cases (warm fuselage, late/missing sensor) and assumptions (MSpeed EX reports temperature; pressure still out of scope) updated. - data-model.md: tSrc/tId/tPar/tLbl replace tStd (27 keys), with a cfgV 1 -> 2 migration. - tools/emulator/sensors.json: MSpeed 450 Temperature, -29..54 C on P7. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NZc7boqR33nD9xyYsGNLBg
Aaron: when the sensor reads out of range (e.g. heat-soaked in the sun before flight) use the default, and resume the sensor once it's back in range; show the sensor temperature on the full-screen gauge. - FR-040: out-of-range/invalid readings use standard temperature; resume automatically after 10 s valid and in range (no flapping); note that in-range heat soak can't be detected. - User Story 4 scenario 8a: 140 F at takeoff -> standard; cools to 95 F in flight -> sensor resumes. - FR-044: full-screen side panel Temperature row (value, unit, source, "SENSOR OUT" while rejected); full-screen layout text now six rows. - Clarifications recorded. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NZc7boqR33nD9xyYsGNLBg
Add a "PR source is develop" check that fails any pull request into main whose source is not this repo's develop branch (required by the main ruleset), and document the branch flow in CLAUDE.md: feature/* branches from develop and merges back by PR; develop goes to main by PR only. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UXZ3gE435jMKRzKW9FRfiy
Spec 001 plan, tasks and implementation for the 2026-09-28/30 spec changes, plus fixes from the first run on a real DS-24 II (2026-10-03). - Temperature source Standard / Manual / Sensor (FR-038..FR-044): sensor read every 5 s, reject-and-resume, 1 C hysteresis, live status in settings, TEMP row on the full-screen panel. - FR-020 limits (-300..10000 ft, -20..130 F), clamp on load, cfgV 1 -> 2 migration of the 0.1.0 tStd checkbox. - App voice (Piper "Amy", FR-030..FR-037): tools/voice/make_voice.py generates 512 WAVs locally (gitignored, CC BY-SA model); per-phrase fallback to the transmitter voice and DFM's recordings. Speech at 1.3x after listening on the transmitter. - Layouts redesigned for the transmitter's real Lua windows (150x23, 150x68, 316x159; the emulator reports larger ones): bar strip, small dial, full screen with one-line panel rows. The full-screen window is named "Speed Gauge (full screen)". - Smooth dial: tools/dial/make_dial.py draws anti-aliased face/track PNGs, since the transmitter enlarges Lua lines without smoothing. - Grouped constants into tables to stay under Lua's 200-local limit. - Probe: sensor unit strings, MODE 5 (untitled full-screen window). - docs/jeti-api-notes.md: transmitter sizes, fonts, scaling, no emulator audio, local-variable limit. Open: SC-009's 1.3 s limit (longest number now 1.50 s), manual checks. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Follow-up from transmitter testing on 2026-10-03. - Glow: nested filled ring segments with solved per-layer alphas, on a 5-degree path; no seams from overlapping polyline joints. - Solid arcs: pre-drawn anti-aliased ring images (overspeed zone, speed arc in each color, speed arc past the limit) from tools/dial/make_dial.py, shown through lcd.setClipping in three 90-degree sectors. setClipping also moves the drawing origin, so images are drawn at minus the clip corner. - ag_gauge: new band() (filled ring segment), drawn in pieces of at most 18 dial steps. Speed Gauge is the module's only user. - JETI Studio's simulator crashes in Qt6Core once filled arcs are drawn; the emulator's windows (widths 157, 320) draw arcs and glow as polylines instead. The transmitter keeps images and filled bands. - "Gauge full scale" renamed "Gauge max limit", with explanatory hints. - Voice generator: default speed 1.3, hint wording. - RAW row shows "---" before the first reading. - TEMPORARY: DEBUG_CPU readout on the full-screen gauge (remove before release, task T080). - docs/jeti-api-notes.md: text vs line resolution, setClipping origin, DFM-InsP uses the same 318 x 159 area. Open: live pitot test, emulator crash confirmation, SC-009 limit. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The emulator's full-screen gauge stopped with the speed at full scale (no new Windows crash logged, so most likely the per-call CPU limit). In the emulator's windows the polyline glow now draws every other band at double width: 5 lines per arc instead of 9, same area and fade. The transmitter's filled bands are unchanged. Confirmed working in the emulator. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…voice From the first live tests on the transmitter (2026-10-03). - Full screen went blank at the overspeed setting: the live glow bands pushed one redraw to the CPU limit. The glow is now baked into the ring images (tools/dial/make_dial.py); no live glow when they exist. drawRing clips by quadrant, so neighbors join exactly along the dial's center lines. - "Callouts start above" (vArm, default 30 mph): normal and continuous callouts and "airspeed alive" wait until first exceeded, then stay armed for the session. Replaces the half-of-landing-speed rule, which continuous mode bypassed (FR-009). - "Landing speed callouts" on/off (landOn, default on). Off: landing speed no longer makes callouts fast and short; it still arms the stall warning (FR-006, FR-008). Settings keys: 30, the limit. - Short hints under both callout switches. - Voice generator default speed 1.5, so every number up to 199 fits SC-009's 1.3 s (longest 1.28 s). Voice WAVs stay uncommitted (FR-035). - Spec, data model, contracts, quickstart, research and tasks (T076, T086-T089) updated. Open: live pitot retest, T088 (voice detection), T080 (remove DEBUG_CPU). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Voice files (user decision, 2026-10-03): - Commit the 512 generated voice WAVs, CREDITS.txt and index.txt in src/Apps/AG-SpdGa/voice/ (1.5x, longest number 1.28 s). - These files are CC BY-SA 4.0, like the Amy voice model, not MIT like the rest of the repository; the license of the model's training recordings is undocumented. Labeled in CREDITS.md, README.md, the folder's CREDITS.txt and tools/voice/README.md. - Spec FR-035 changed from "MUST NOT be committed" to committed; the voice/ line is removed from .gitignore. From the live test (no crashes; worst CPU 14% at speed): - Stall warning at most twice per slowdown; the count resets once above landing speed (FR-010, US2 #1). - Continuous callouts speak only while above "Callouts start above" (FR-009a); hint updated. - "Gauge max limit" hint lists the MSpeed (20-350 km/h) and MSpeed 450 EX (80-450 km/h) ranges. - Quickstart 7a/7b; tasks T062, T090. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
From live testing on the transmitter (2026-10-04). - CPU readout off (DEBUG_CPU = false; T080): worst full-screen redraw was 14% at speed. - No callout below 5 mph or its equivalent in any units (FR-008b). - "Longest time between callouts" 2-60 s (was 10-60), never below the shortest: it stops at the shortest, and raising the shortest raises it; corrected on load too (FR-005). - Max-speed callout "max N unit" once the session max hasn't risen for 1 s and beats the last announced max by the callout sensitivity (FR-019a). New voice phrase max.wav (513 files); the generator's new --missing option made it without changing the committed WAVs. - Spec, data model, contracts, quickstart (13c-13e) and tasks (T091). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- APP_VERSION 0.3.0 after live pitot testing on the DS-24 II. - CLAUDE.md: target window sizes are the transmitter's (150x23, 150x68, 316x159), not the emulator's; emulator plays no Lua audio. - docs/jeti-api-notes.md: facts verified on the transmitter (device string, size 0, 22.05 kHz WAV, the degree sign, renderer reuse, temperature unit, CPU); local-variable count 191 of 200. - DEBUG_CPU comment: diagnostic, off for release. - plan.md release note; tasks T043 done, T070 partly. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Brings in develop's "main only from develop" branch rule (5747cc5) so the pull request into develop merges cleanly. CLAUDE.md conflict: kept develop's Branches section (enforced on GitHub) and folded in this branch's details that still fit: matching branch and spec-folder names, the Spec Kit note, release tags after the develop -> main PR, ask before pushing/PRs/tagging. Dropped the release/x.y.z and hotfix flow, which the new rule doesn't allow. README's branching table updated to match. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- README apps table: Speed Gauge is a release candidate (0.3.0). - tasks.md: manual checks the user confirmed on the transmitter (2026-10-04) ticked: T017, T021, T029, T035, T039, T046, T072, T077. T088 closed (no "Voice files missing" seen); T070 closed with the queue pause and warning cut-off judged "fine in use". - Still open: T059 (temperature source checks) and T066 (voice checks), which the user hasn't tested separately. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Feature/001 speed gauge
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
README: Speed Gauge released (0.3.0)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.