M57andM57
adde03a4de
* fix(chat): stop duplicating interim commentary before tool calls (issue #842) Interim commentary the agent narrates before each tool call rendered TWICE. Two mechanisms, both now covered: 1. The old tool.start orphan seal + message.complete's full text: when the complete payload repeats the commentary, the final bubble duplicated it. onMessageComplete now strips the sealed-orphan prefix (tracked in StreamingState) from the final text. 2. On-device capture + DB forensics (2026-08-08) exposed the real repeat offender: the app seals the RAW streamed text — which carries leading blank lines the model emits before its narration — while the backend persists a CLEANED copy (leading whitespace stripped). A transcript reload ~10s later merged the REST page, exact-content matching failed, and the cleaned REST copy rendered on top of the live bubble. sameLogicalMessage now compares trimmed content, so the sealed live bubble is covered by the REST row. Also: - session.interrupt seals the in-flight streaming message before clearing streaming state, so interrupting a commentary turn keeps the partial text - regression tests cover both turn shapes (commentary in the final row → stripped; absent → kept) and the whitespace-drift reload case * fix(chat): match tool rows by gateway call id, drop reasoning-echo dupes (#842) REST transcript rows carry tool_call_id; live WS bubbles keep tool_id from tool.start. Match tool rows on that 1:1 identity so MCP/web rows (stored as raw <untrusted_tool_result> text with no JSON key) dedupe during the post-turn transcript merge instead of ghosting after the final message. reasoning.available echo (narration text == message body, sent ~40ms before each tool.start) no longer overwrites the real reasoning trace, so narration bubbles don't render their own text twice (body + Reasoning card). Room v5 migration adds tool_call_id; regression tests use captured on-device payload shapes. * fix(chat): flush throttled tokens before orphan seal + prefix-safe dedupe (#842) The tool.start seal raced the 33ms throttled token flush: deltas landing just before the transition stayed in the buffer, so the sealed narration orphan came out truncated (on-device capture: orphan ended '...chickpea', backend persisted '...chickpea goodness:') and failed the logical-match against the clean REST row — the commentary ghosted on the next transcript sync. - ChatStreamingController.flushPendingTokens(): force-flushes the throttled TOKEN buffer on MessageStart/MessageComplete/MessageDone/ToolStart so the reducer seals the COMPLETE narration. - sameLogicalMessage: accept prefix-covering (both sides >=40 chars) so a residual seal drift can never re-add a substantial narration. - Regression tests: controller flush + truncated-orphan prefix match. --------- Co-authored-by: M57 <hy4ri@users.noreply.github.com>
Hermes Mobile
Native Android companion app for your Hermes AI agent.
Overview
Hermes Mobile is the native Android client for Hermes Agent. It connects securely to your local Hermes gateway (REST API and WebSocket TUI Gateway) over LAN, giving you pocket control over your AI assistant.
Screenshots
Chat, automation, productivity, and agent configuration — from your phone.
Features
- Real-Time Chat: Message your agent with Room-backed local database history and inline reply notifications.
- System Config: Manage active profiles, installed skills, plugins, toolsets, and LLM model/provider selections.
- Operations: Stream and filter live logs, manage cron jobs, edit environment keys, test webhooks, and monitor processes.
- Gateway Status: Monitor WebSocket connection, MCP servers, messaging channels, and OAuth providers.
- Productivity: View and manage tasks via integrated Kanban boards, track agent milestones, and browse session history.
- Analytics & Billing: Usage analytics dashboard and billing/subscription management.
- Theming: 6 built-in color presets (Default, Monochrome, Gruvbox, Catppuccin, AMOLED, Neon Noir) plus Material You dynamic colors on supported devices.
- Modern UX: Native Material 3 design with pull-to-refresh, scroll-aware TopBar, and customizable bottom navigation.
Quick Start
Prerequisites
- JDK 21+ (required for Kotlin compilation and the Gradle toolchain)
- Android Studio (Ladybug+) or a Nix development environment
Build & Deploy
- Clone the repository:
git clone https://github.com/Hy4ri/hermes-mobile.git cd hermes-mobile - Build the debug APK:
./gradlew assembleDebug - Install on your emulator/device:
adb install app/build/outputs/apk/debug/app-debug.apk
Note: For release builds, ensure keystore environment variables (KEYSTORE_PATH, KEYSTORE_PASSWORD, KEY_ALIAS, KEY_PASSWORD) are configured, or let the GitHub Actions release workflow handle it on tag push (v*).
Authentication
Once the app is installed, you need to point it at your Hermes gateway. The app auto-detects which auth mode the dashboard is using — just fill in the fields it shows.
1. Start the dashboard
On your host machine, start the dashboard:
hermes dashboard # loopback (127.0.0.1:9119) — no auth needed
hermes dashboard --host 0.0.0.0 # LAN — requires auth
For LAN access, configure credentials in ~/.hermes/config.yaml:
dashboard:
basic_auth:
username: admin # pick your own
password: hermes # pick your own
2. Connect the app
Tap Sign in on the landing screen and enter the dashboard host and port. The app probes the dashboard and reveals the fields you need:
| Auth mode | When | What you fill |
|---|---|---|
| Token only | Dashboard on same machine (loopback) | Token — grab from ~/.hermes/dashboard-token.txt or ~/.hermes/.env (HERMES_DASHBOARD_SESSION_TOKEN). The app can also auto-extract it from the dashboard page |
| Basic auth | Dashboard on LAN with password gate | Username + Password (default admin / hermes). The app logs in, gets a session cookie, and mints a WebSocket ticket automatically |
The app communicates over plain HTTP — it's designed for trusted local networks only. Do not expose your Hermes gateway to untrusted networks.
Connection profiles
Have multiple gateways? Switch between them in Settings → Connection profiles. Each profile stores its own host, port, and token — just tap to swap.
Pairing (admin)
The Pairing screen lets you approve or revoke agents and services that are trying to connect to your gateway, such as Telegram or Discord sessions.
Project Structure
app/src/main/java/com/m57/hermescontrol/
├── data/ # Local (Room, AuthManager), Remote (Retrofit, OkHttp), WS (WebSocket), Models
├── notification/ # Foreground service + inline reply for chat notifications
├── theme/ # Preset-based design system (6 themes), status colors, spacing, typography
├── ui/ # 28 Compose feature screens + common components (HermesScaffold, StateViews)
├── util/ # CronExpressionFormatter, LocaleContextWrapper
└── Navigation*.kt # Navigation3 wiring, keys, screen registry, controller
Tech Stack
- Language: Kotlin 2.4.10 with KSP 2.3.10 compiler plugin
- UI & Layout: Jetpack Compose (BOM 2026.03.01) & Material 3 / Material You
- Navigation: Navigation3 (Compose-first Routing)
- Networking: Retrofit 3.0.0, OkHttp 5.4.0, Kotlinx Serialization 1.11.0
- Database: Room 2.7.1 with SQLCipher 4.17.0 encryption
- Security:
EncryptedSharedPreferences(AES256-GCM), DataStore - Theming: 6 built-in presets (Default, Monochrome, Gruvbox, Catppuccin, AMOLED, Neon Noir) + Material You dynamic colors
- Image Loading: Coil 2.7.0
- Testing: JUnit 5, MockK, Turbine, Espresso, Compose UI testing
- Formatting:
ktlint1.2.1 style rules (checked automatically in CI)
Contributing
Contributions are welcome! Please read CONTRIBUTING.md for our branch workflow, code style guidelines, and PR checklist.
For developer-specific details, code conventions, and project architecture notes, refer to AGENTS.md.
License
Copyright © 2026 M57 (Hy4ri).
This project is licensed under the Apache License, Version 2.0. See the LICENSE file for details.