fix: remove remaining Vitruvian IP outliers after catalogue swap

.almanac/pages/machine-diagnostics.md

@@ -56,7 +56,7 @@ The screen is driven by live BLE state. `BleRepository` exposes `diagnostics` as
 
 The binary contract is broader than just four fault words. `parseDiagnosticPacket()` accepts an empty payload as a valid zero snapshot, rejects non-empty payloads shorter than `18` bytes, then decodes uptime seconds, four unsigned `16-bit` fault words, six required temperatures, two optional extra temperatures, an optional `52`-byte crash block, and an optional `32-bit` warnings field [@protocol-parser] [@protocol-models]. [[vitruvian-ble-protocol]] is the neighboring page for the broader scan, rep, monitor, and command surface around this diagnostic characteristic.
 
-Fault decoding is intentionally category-specific. `DiagnosticFaultDecoder` always projects the packet into four display slots named `Vee`, `Other`, `Motor A`, and `Motor B`, then maps bitmasks to labels such as `TI restarted`, `Over voltage`, `Encoder`, or `Motor overtemp` instead of exposing only raw integers [@fault-decoder]. The view-model tests pin that labeling contract and verify that the export text includes both the human label and the raw hex code [@diagnostics-tests].
+Fault decoding is intentionally category-specific. `DiagnosticFaultDecoder` always projects the packet into four display slots named `Controller`, `Other`, `Motor A`, and `Motor B`, then maps bitmasks to labels such as `Controller restarted`, `Overvoltage`, `Encoder fault`, or `Motor over-temperature` instead of exposing only raw integers [@fault-decoder]. The view-model tests pin that labeling contract and verify that the export text includes both the human label and the raw hex code [@diagnostics-tests].
 
 The export contract is deliberately narrow. `buildDiagnosticsExportText()` prepends app version, export timestamp, `REDACTED_DIAGNOSTICS` classification, and a privacy line that excludes workout history, profiles, auth or session tokens, Supabase config, and keystore data before listing the current packet contents [@diagnostics-vm]. The copy button stays disabled until a packet exists, so the feature exports the latest machine snapshot rather than an empty template [@diagnostics-screen].
 

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -28,7 +28,7 @@ body:
     id: app-version
     attributes:
       label: App Version
-      description: What version of Vitruvian Phoenix are you using?
+      description: What version of Project Phoenix are you using?
       placeholder: "e.g., 1.0.0"
     validations:
       required: true

.phoenix-review/BUG-FIX-PLAN.md

@@ -142,7 +142,7 @@
 - F173 | impact=250 | Domain | medium / failure-point | file=shared/src/commonMain/kotlin/com/devil/phoenixproject/domain/voice/SafeWordListener.kt | lines=26, 38-44 | The common listener contract exposes `startListening()`, `isListening`, and `detectedWord`, but no error/status channel for unavailable recognition, denied microphone/speech permission, unsupported on-device recognition, or start failures. The platform implementations can fail by logging and leaving `isListening` false, which leaves callers unable to distinguish “not yet started”, “already stopped”, and “emergency voice stop is unavailable”. For a safety-oriented emergency stop path, silent non-operation can make the workout UI appear configured while no voice stop will ever fire. | fix: Add a shared status/error flow or sealed state such as `Idle`, `Listening`, `PermissionDenied`, `RecognitionUnavailable`, and `StartFailed`, and have calibration/workout UI surface failures instead of relying only on logs and `isListening`.
 - F174 | impact=250 | Domain | medium / failure-point | file=shared/src/commonMain/kotlin/com/devil/phoenixproject/domain/voice/SafeWordListener.kt | lines=41-44 | The API documents that emitted detections occur when results “match the safe word”, but it does not define or enforce what a valid safe word is. The current platform implementations perform whole-token matching after splitting transcripts on whitespace, while settings/calibration paths only require a non-blank string. A user can configure a multi-word phrase or punctuation-bearing value that appears valid in common code but can never match the actual token-based listener. | fix: Define the safe-word normalization/validation contract in common code. Either restrict configured safe words to one normalized token and reject/trim punctuation at save/calibration time, or update platform matching to normalize transcripts and support the same phrase rules accepted by the UI.
 - F304 | impact=250 | Utilities | medium / bug | file=shared/src/commonMain/kotlin/com/devil/phoenixproject/util/DataBackupManager.kt | lines=1355-1380 | The streaming personal-record importer drops `PersonalRecordBackup.cableCount` by passing `cable_count = null` to `queries.upsertPR`. The non-streaming importer preserves this field at lines 631-643, so large backups imported through the streaming path lose cable-count metadata while small backups do not. | fix: Pass `pr.cableCount?.toLong()` in the streaming path to match `importFromJson()`.
-- F306 | impact=250 | Utilities | medium / bug | file=shared/src/commonMain/kotlin/com/devil/phoenixproject/util/BlePacketFactory.kt | lines=75-86 | `createStopCommand()` is documented as the primary STOP command but emits legacy opcode `0x05`, while `BleConstants.Commands.STOP_COMMAND` and `createOfficialStopPacket()` use official stop opcode `0x50`. This creates an attractive but unsafe API footgun: callers choosing `createStopCommand()` by name/comment may send a different stop semantics than the official stop/clear-fault packet. | fix: Rename/deprecate the legacy helper (for example `createLegacyStopCommand()`), make the primary helper delegate to `createOfficialStopPacket()`, and keep tests explicit about the legacy opcode only where it is intentionally required.
+- F306 | impact=250 | Utilities | medium / bug | file=shared/src/commonMain/kotlin/com/devil/phoenixproject/util/BlePacketFactory.kt | lines=75-86 | `createStopCommand()` is documented as the primary STOP command but emits legacy opcode `0x05`, while `BleConstants.Commands.STOP_COMMAND` and `createSoftStopPacket()` use 0x50 soft-stop opcode `0x50`. This creates an attractive but unsafe API footgun: callers choosing `createStopCommand()` by name/comment may send a different stop semantics than the 0x50 soft-stop/clear-fault packet. | fix: Rename/deprecate the legacy helper (for example `createLegacyStopCommand()`), make the primary helper delegate to `createSoftStopPacket()`, and keep tests explicit about the legacy opcode only where it is intentionally required.
 - F311 | impact=250 | Utilities | medium / bug | file=shared/src/commonMain/kotlin/com/devil/phoenixproject/util/HardwareDetection.kt | lines=8-13, 20-29 | The file-level documentation says name-prefix hardware detection was flawed and should be avoided, but `detectModel()` still infers `VFormTrainer`/`TrainerPlus` from `Vee_` and `VIT` prefixes. That stale detection result is used when reporting `ConnectionState.Connected`, so UI or downstream logic can still receive a model classification the comments say is unreliable. | fix: Either remove model inference and return `Unknown` until firmware-backed detection exists, or update the documentation and tests to explain why these prefixes are now reliable. Prefer firmware/version-characteristic detection before exposing a concrete hardware model.
 - F313 | impact=250 | Utilities | medium / bug | file=shared/src/commonMain/kotlin/com/devil/phoenixproject/util/KmpUtils.kt | lines=232-236 | `formatFloat()` builds the decimal part from `rounded - intPart`, but `toLong()` truncates negative values toward zero. Negative values with decimal places therefore format with a negative fractional component, e.g. the current algorithm formats `-1.23` with two decimals as `-1.-23` and `-0.25` as `0.-25`. Any UI using this helper for signed deltas, percentages, velocities, or corrections can display malformed numbers. | fix: Format the absolute fractional component separately from the sign, or use a locale-stable multiplatform formatter that rounds the whole scaled integer first and then reconstructs sign, integer, and padded fractional digits from absolute values. Add tests for negative values such as `-1.23`, `-5.5`, and `-0.25`.
 - F314 | impact=250 | Utilities | medium / bug | file=shared/src/commonMain/kotlin/com/devil/phoenixproject/util/KmpUtils.kt | lines=245, 283 | `formatDouble()` downcasts every `Double` to `Float` before formatting. This loses precision for normal `Double` values and turns large finite doubles outside the `Float` range into infinities, which then hit the non-finite guard and display as zero. The `Double.format()` extension exposes this behavior broadly, so any future Double-valued analytics/statistics display can silently show rounded or completely wrong values. | fix: Implement `formatDouble()` using Double arithmetic end-to-end and a Double finite check rather than delegating through Float. Add regression tests with a high-precision value and a finite value greater than `Float.MAX_VALUE` to prevent silent zeroing.

.phoenix-review/CONSOLIDATED-REPORT.md

@@ -323,7 +323,7 @@
 - F303 | module=Utilities | category=failure-point | file=shared/src/commonMain/kotlin/com/devil/phoenixproject/util/DataBackupManager.kt | lines=1138-1215 | description: Metric batching counts only metrics that are actually inserted (`batchCount++` is inside `metric.sessionId in importedSessionIds`). If a large backup contains mostly metrics for skipped/pre-existing sessions, the second batching loop can consume the entire remaining `metricSamples` array in one transaction because `batchCount` never reaches `IMPORT_BATCH_SIZE`. This defeats the OOM/lock-avoidance goal of streaming import for exactly the large-file path. | fix: Limit each transaction by rows processed/seen, not rows inserted, or maintain separate `rowsInBatch` and `insertedInBatch` counters.
 - F304 | module=Utilities | category=bug | file=shared/src/commonMain/kotlin/com/devil/phoenixproject/util/DataBackupManager.kt | lines=1355-1380 | description: The streaming personal-record importer drops `PersonalRecordBackup.cableCount` by passing `cable_count = null` to `queries.upsertPR`. The non-streaming importer preserves this field at lines 631-643, so large backups imported through the streaming path lose cable-count metadata while small backups do not. | fix: Pass `pr.cableCount?.toLong()` in the streaming path to match `importFromJson()`.
 - F305 | module=Utilities | category=failure-point | file=shared/src/commonMain/kotlin/com/devil/phoenixproject/util/DataBackupManager.kt | lines=976-980, 1025-1737 | description: The streaming importer is order-dependent even though JSON object member order is not semantically significant. Child arrays such as `metricSamples`, `routineExercises`, `cycleDays`, `plannedSets`, and `completedSets` are imported only if their parent IDs have already been seen in earlier fields. A valid backup with fields reordered can silently skip children or log warnings while returning success. | fix: Make streaming import robust to object order by doing multiple passes over a seekable source, buffering small parent/child ID sets and deferred child rows, or enforcing/exporting a documented field order and rejecting out-of-order backups instead of silently skipping data.
-- F306 | module=Utilities | category=bug | file=shared/src/commonMain/kotlin/com/devil/phoenixproject/util/BlePacketFactory.kt | lines=75-86 | description: `createStopCommand()` is documented as the primary STOP command but emits legacy opcode `0x05`, while `BleConstants.Commands.STOP_COMMAND` and `createOfficialStopPacket()` use official stop opcode `0x50`. This creates an attractive but unsafe API footgun: callers choosing `createStopCommand()` by name/comment may send a different stop semantics than the official stop/clear-fault packet. | fix: Rename/deprecate the legacy helper (for example `createLegacyStopCommand()`), make the primary helper delegate to `createOfficialStopPacket()`, and keep tests explicit about the legacy opcode only where it is intentionally required.
+- F306 | module=Utilities | category=bug | file=shared/src/commonMain/kotlin/com/devil/phoenixproject/util/BlePacketFactory.kt | lines=75-86 | description: `createStopCommand()` is documented as the primary STOP command but emits legacy opcode `0x05`, while `BleConstants.Commands.STOP_COMMAND` and `createSoftStopPacket()` use 0x50 soft-stop opcode `0x50`. This creates an attractive but unsafe API footgun: callers choosing `createStopCommand()` by name/comment may send a different stop semantics than the 0x50 soft-stop/clear-fault packet. | fix: Rename/deprecate the legacy helper (for example `createLegacyStopCommand()`), make the primary helper delegate to `createSoftStopPacket()`, and keep tests explicit about the legacy opcode only where it is intentionally required.
 - F307 | module=Utilities | category=failure-point | file=shared/src/commonMain/kotlin/com/devil/phoenixproject/util/ConnectivityChecker.kt | lines=7-12 | description: The common contract only exposes a synchronous `isOnline()` method and has no lifecycle/dispose hook or observable connectivity-change signal. Platform implementations that allocate long-lived resources or background monitors cannot be cancelled through the common API, and common sync code cannot subscribe to a reconnect event after setting a waiting-for-connectivity state. | fix: Add lifecycle and state-change semantics to the expect contract, such as `close()`/`dispose()` plus a `StateFlow<Boolean>` or callback registration, and make DI scopes own the checker lifecycle explicitly.
 - F308 | module=Utilities | category=bug | file=shared/src/commonMain/kotlin/com/devil/phoenixproject/util/Constants.kt | lines=139-142 | description: `ProtocolConstants` declares packet sizes that contradict the actual packet builders and tests. `ECHO_PACKET_SIZE` is `29`, but `BlePacketFactory.createEchoControl()` builds a 32-byte frame and tests assert 32. `ACTIVATION_PACKET_SIZE` is `97`, but `BleConstants.ActivationPacket.SIZE`, `BlePacketFactory.createProgramParams()`, model documentation, and tests all use 96 bytes. Any code using these constants for allocation, validation, or MTU checks would reject valid packets or allocate the wrong size. | fix: Define packet sizes from a single source of truth (`BleConstants.ActivationPacket.SIZE` and the factory Echo size), update Echo to 32 and activation to 96, and add tests covering `ProtocolConstants` parity with factory output.
 - F309 | module=Utilities | category=failure-point | file=shared/src/commonMain/kotlin/com/devil/phoenixproject/util/DeviceInfo.kt | lines=71-73 | description: The common API promises device info as a JSON `String`, but the contract does not require escaping or a structured serializer-backed result. Platform values such as model names, device names, build strings, or bundle values can contain quotes, backslashes, or control characters; manual string-building actual implementations can then return invalid JSON or malformed metadata. | fix: Replace `toJson(): String` with a serializable data model or require all actual implementations to use `kotlinx.serialization`/a shared escaping helper. Add tests with quotes, backslashes, and newlines in representative fields.

.phoenix-review/utilities-part-2.md

@@ -35,8 +35,8 @@ Findings: 11 total
 - Category: bug
 - Severity: medium
 - Line numbers: 75-86
-- Description: `createStopCommand()` is documented as the primary STOP command but emits legacy opcode `0x05`, while `BleConstants.Commands.STOP_COMMAND` and `createOfficialStopPacket()` use official stop opcode `0x50`. This creates an attractive but unsafe API footgun: callers choosing `createStopCommand()` by name/comment may send a different stop semantics than the official stop/clear-fault packet.
-- Suggested fix direction: Rename/deprecate the legacy helper (for example `createLegacyStopCommand()`), make the primary helper delegate to `createOfficialStopPacket()`, and keep tests explicit about the legacy opcode only where it is intentionally required.
+- Description: `createStopCommand()` is documented as the primary STOP command but emits legacy opcode `0x05`, while `BleConstants.Commands.STOP_COMMAND` and `createSoftStopPacket()` use 0x50 soft-stop opcode `0x50`. This creates an attractive but unsafe API footgun: callers choosing `createStopCommand()` by name/comment may send a different stop semantics than the 0x50 soft-stop/clear-fault packet.
+- Suggested fix direction: Rename/deprecate the legacy helper (for example `createLegacyStopCommand()`), make the primary helper delegate to `createSoftStopPacket()`, and keep tests explicit about the legacy opcode only where it is intentionally required.
 
 #### Finding 3
 - Category: bug

ANDROID_INSTALL.md

@@ -1,6 +1,6 @@
 # Android Installation Guide
 
-This guide explains how to install Vitruvian Phoenix on your Android device.
+This guide explains how to install Project Phoenix on your Android device.
 
 ## Prerequisites
 
@@ -47,7 +47,7 @@ Android requires permission to install apps from outside the Play Store.
 1. Open the downloaded APK file
 2. Tap **Install**
 3. Wait for installation to complete
-4. Tap **Open** or find **Vitruvian Phoenix** in your app drawer
+4. Tap **Open** or find **Project Phoenix** in your app drawer
 
 ---
 
@@ -64,7 +64,7 @@ When you first launch the app, you'll be asked to grant permissions:
 - This is an Android requirement - the app does not track your location
 - Tap **Allow** when prompted
 
-**Note:** If you deny permissions, the app cannot connect to your trainer. You can always grant permissions later in Settings > Apps > Vitruvian Phoenix > Permissions.
+**Note:** If you deny permissions, the app cannot connect to your trainer. You can always grant permissions later in Settings > Apps > Project Phoenix > Permissions.
 
 ---
 
@@ -95,12 +95,12 @@ When a new version is released:
 ### App Crashes on Launch
 
 - Make sure your device is running Android 8.0 or later
-- Try clearing app data: Settings > Apps > Vitruvian Phoenix > Storage > Clear data
+- Try clearing app data: Settings > Apps > Project Phoenix > Storage > Clear data
 - Report the issue on GitHub with your device model and Android version
 
 ### Bluetooth Permission Denied
 
-1. Go to **Settings > Apps > Vitruvian Phoenix > Permissions**
+1. Go to **Settings > Apps > Project Phoenix > Permissions**
 2. Enable **Nearby devices** (Android 12+) or **Location** (Android 11 and below)
 3. Restart the app
 

androidApp/build.gradle.kts

@@ -321,7 +321,7 @@ android {
         targetSdk = 37
         // Fail fast if CI injects an invalid version code instead of silently shipping a default.
         versionCode = injectedVersionCode ?: 5
-        versionName = "0.9.6"
+        versionName = "1.0.0"
 
         // Supabase config injected from local.properties
         buildConfigField("String", "SUPABASE_URL", "\"$supabaseUrl\"")

androidApp/src/main/res/xml/backup_rules.xml

@@ -1,6 +1,6 @@
 <?xml version="1.0" encoding="utf-8"?>
 <!-- API 23-30 backup exclusion rules (android:fullBackupContent) -->
-<!-- Excludes Vitruvian database and sensitive preferences from Android auto-backup -->
+<!-- Excludes the app database and sensitive preferences from Android auto-backup -->
 <full-backup-content>
     <exclude domain="database" path="vitruvian.db" />
     <exclude domain="database" path="vitruvian.db-journal" />

androidApp/src/main/res/xml/data_extraction_rules.xml

@@ -1,6 +1,6 @@
 <?xml version="1.0" encoding="utf-8"?>
 <!-- API 31+ data extraction rules (android:dataExtractionRules) -->
-<!-- Excludes Vitruvian database and sensitive preferences from both cloud backup and device transfer -->
+<!-- Excludes the app database and sensitive preferences from both cloud backup and device transfer -->
 <data-extraction-rules>
     <cloud-backup>
         <exclude domain="database" path="vitruvian.db" />