v0.4.0 — Fix streaming, auth, theming & dialogs

Fixes

• Chat streaming (CRITICAL): Messages now accumulate properly instead of showing only the last token
• Assistant text color: Readable on both light and dark themes (was invisible on light)
• Auth handling: Tailscale/local mode no longer sends hardcoded token
• Connect params: Mode changed from ui to operator, version from BuildConfig
• Session rename dialog: No longer targets wrong session after context menu dismissal
• Version display: Settings shows actual build version instead of hardcoded v0.3.0
• Error logging: WebSocket failures now include HTTP status codes
• Build: targetSdk 34→35, versionCode 11

v0.3.1 — Notification Fix & Message Preview

Changes

• Notifications preference enforced: Toggle in settings now actually controls whether notifications are posted
• Message preview in notifications: Chat notifications now show the actual response text (up to 300 chars) instead of generic "New response received"
• Notification ID fix: Fixed bug where wrong notification ID was used for notify() call

Full Changelog

v0.3.0...v0.3.1

v0.3.0-alpha — Phase 2: Full Feature Set

Phase 2 Release — 6 Major Features

Session Management (2A)

• Side drawer with session list (swipe from left or tap hamburger menu)
• Create, switch, delete, compact, and rename sessions
• Sessions persisted locally via Room DB
• Dynamic session key — no more hardcoded "android://companion"

Settings Editor (2B)

• Dynamic config form generated from server JSON schema
• Text, number, toggle, and select field types
• Collapsible sections, search filtering
• Save with optimistic concurrency, snackbar feedback

Chat Attachments (2C)

• Camera capture and gallery photo picker
• Image compression: resize to 1024px max edge, JPEG compress to <512KB
• Attachment thumbnails in input bar and message bubbles
• Base64-encoded attachments sent via gateway API

Log Viewer (2D)

• Live log streaming from gateway (2s polling)
• Color-coded by level: ERROR=red, WARN=yellow, DEBUG=dim
• Filter chips (ALL/ERROR/WARN/INFO/DEBUG)
• Auto-scroll with manual override, scroll-to-bottom FAB
• Accessible from Settings → Diagnostics → View Logs

Push Notifications (2E)

• Foreground service (BillBotService) with connectedDevice type
• Three notification channels: chat (HIGH), alerts (DEFAULT), service (LOW)
• Background chat notifications with 5s throttle
• Health/shutdown alert notifications
• Toggle in Settings → Notifications

Login/Auth (2F)

• Expandable authentication section on connection screen
• Three modes: Tailscale (default), Token, Password
• Segmented button selector with mode-specific input fields
• Auth params passed through WebSocket handshake
• Scope-based feature gating from server response

Technical

• Room DB: version 1→2 (sessions table)
• 14 new files, 17 modified files, ~2500 lines changed
• New dependency: lifecycle-process for ProcessLifecycleOwner
• APK: 19MB debug build

v0.2.8 — Chat persists through phone lock

What's New

Chat messages now survive phone lock, app kills, and gateway restarts.

How it works

• Messages are saved to a local SQLite database (Room) as you chat
• When you unlock your phone, the app shows your cached messages immediately — before the WebSocket even reconnects
• Once connected, it refreshes from the server and updates the cache
• "New Chat" clears both server session and local cache

Also includes v0.2.7 fixes

• Stop button works (was sending malformed abort)
• New Chat button works (wrong parameter name)
• No more stuck "thinking" on reconnect

v0.2.7 — Fix stuck thinking & broken stop/reset

Fixes

• Stop button now works: chat.abort was missing the required sessionKey parameter — server rejected every abort attempt
• New Chat button now works: sessions.reset was sending sessionKey instead of key — server schema mismatch caused rejection
• No more stuck "thinking" on reconnect: App now resets generating state when loading history after reconnect
• Abort is resilient: Stop button clears UI state immediately, even if the server call fails

Root Cause

The gateway got stuck on a reasoning-only response from the model. The app couldn't recover because both abort and reset were sending malformed RPC params that the server rejected.

v0.2.6 - Radeon VII Hardware Metrics

What's New

• Radeon VII hardware monitoring via LibreHardwareMonitor
  • GPU temp (color-coded: green <75, yellow <85, red 85+)
  • Hot spot temperature
  • GPU utilization with progress bar
  • Core / memory clocks in MHz
  • Power draw (W / 300W TDP)
  • Fan speed (RPM + %)
• Radeon VII now appears in Temperatures, GPU Utilization, and Power Draw metric cards
• Graceful degradation: shows "Hardware monitor offline" when LHM is down instead of "not available"
• Removed all "llama.cpp limitation" placeholder notes

Backend (oogleyskr/billbot)

• OpenClaw polls LibreHardwareMonitor data.json endpoint every 15s
• LHM JSON tree parser with fallback sensor name matching
• hwMonitorEnabled/Host/Port config fields in memoryCortex section
• LHM failure is supplementary — does not degrade overall Memory Cortex status

v0.2.5 - Chat Persistence + New Chat Button

Changes

• Fix chat persistence on screen lock — chat history now reloads from the gateway when the Activity recreates (e.g. after locking the phone), so messages no longer disappear
• New Chat button — top-right icon in the chat screen with a confirmation dialog that resets the server-side session

Install

Download app-debug.apk below and sideload on your Android device.

v0.2.4 - Unified Services + Dashboard Fixes

Changes

Devices view

• Unified Services card: Providers, SSH Tunnels, Memory Cortex, and all Multimodal services combined into one card with status dots and latency
• DGX Spark card now includes system metrics: CPU, RAM, Network In/Out are shown inside the DGX Spark card (no more separate "DGX System" section at the bottom)

Metrics view

• Radeon VII noted: Temperature, GPU Utilization, and Power Draw sections note that Radeon VII hardware metrics aren't available (llama.cpp only exposes VRAM and tok/s — to get temp/util/power we'd need AMD GPU monitoring on Windows)
• Radeon VII VRAM already shows correctly in the VRAM Usage section
• DGX labels fixed to say "DGX Spark" consistently

Install

Download app-debug.apk and install. Requires Tailscale connection.

v0.2.3 - Dashboard Metrics View + Model Names

Changes

• Dashboard Devices/Metrics toggle: Switch between per-device cards (Devices) and grouped metric view (Metrics)
• Metrics view: Temperatures, VRAM, GPU Utilization, Power Draw, CPU/RAM, Network, Service Health all grouped across devices
• Model names on GPU cards: DGX Spark shows "LLM: gpt-oss-120b", RTX 3090 shows all multimodal service models
• Removed Inference card from dashboard (now covered by Tokens page)
• Oogley identity: BillBot now recognizes you as Oogley from the Android app (SOUL.md updated)
• Version bump to 0.2.3

Install

Download app-debug.apk and install on your Android device. Requires Tailscale connection.

v0.2.2 - Chat Loading + RTX 3090 Dashboard

Changes

• Chat loading indicator: Loading spinner now shows immediately when you send a message, not just when BillBot starts streaming
• RTX 3090 on token dashboard: Now shows in the By Device view with its role (Multimodal Services: STT, Vision, TTS, ImageGen)
• Device cards show descriptive roles for non-LLM hardware
• Version bump to 0.2.2

Install

Download app-debug.apk and install on your Android device. Requires Tailscale connection.