Xiaoma AI Hermes Chinese Enhancement Pack
A one-command Chinese localization patch for Hermes Desktop, TUI and CLI, auto-matched to your Hermes version so Chinese users can read every screen.
- Source repo
- fresh-claw/hermes-zh-cn
- Stars
- ★ 16
- Last updated
- 1mo ago
- License
- MIT
- Primary language
- Shell
- FA score
- 36/100 · Major gaps
At a glance
- Works with
- Platform-specific
- You'll need
- Typical use
- Chinese newcomers to Hermes who cannot read the English startup screen and quick commands.
- Main limitation
- Only useful within the Hermes ecosystem; it does nothing for other agent frameworks.
- Source review
- 36/100 · Major gaps
What does this agent do, and when should you use it?
The Xiaoma AI Hermes Chinese Enhancement Pack (fresh-claw/hermes-zh-cn) is a localization patch and installer for Hermes Desktop, the Hermes Agent TUI, and the Hermes CLI. The intended flow is that the user sends one sentence to Hermes, which reads the useai.live/hermes page and follows its installation guide. The installer detects or installs the official desktop app first, identifies the local Hermes version within 4 seconds, selects a matching pack for 0.15.x, 0.14.x, or 0.13.x, and falls back to a legacy compatibility pack otherwise. After verifying SHA256 checksums, the patch installs to ~/.xiaoma-hermes/current and modifies Hermes program files such as hermes_cli/banner.py and ui-tui sources (with backup) to localize the startup banner, quick commands, gateway status, and approval prompts. The project is MIT-licensed and runs on Windows, macOS, Linux, and WSL2.
The repository consists of installer entry points under web/ and per-version Chinese packs under packages/. The flow: Hermes reads the body text, agent., and latest. served at useai.live/hermes; the Windows entry (install.ps1) detects the official desktop app and invokes official install parameters if missing; the macOS (install.command), Linux, and WSL2 entry (install.sh) detects a local Hermes and runs the official install script if absent. The installer then matches packages/0.15.x, 0.14.x, 0.13.x, or legacy zh-CN packs to the detected version, verifies SHA256, installs to ~/.xiaoma-hermes/current, sets display.language=zh (or updates ~/.hermes/config.yaml directly), and backs up then modifies hermes_cli/banner.py, skin_engine.py, tips.py, commands.py, bundles.py, mcp_picker.py, secrets_cli.py, acp_adapter/*.py, gateway/platforms/*.py, and ui-tui/src/**/*.ts. It also writes ~/.hermes/skills/xiaoma-hermes-zh/SKILL.md and provides a wrapper at ~/.xiaoma-hermes/bin/hermes that checks latest. before launch to apply updates.
- Chinese newcomers to Hermes who cannot read the English startup screen and quick commands.
- Users who installed localization but still see English in skill categories, gateway status, or approval prompts.
- Beginners on Windows, macOS, Linux, or WSL2 who want one command to install the desktop app and the Chinese patch together.
- Users running Hermes versions outside 0.13.x–0.15.x who need the legacy compatibility pack.
- Heavy users who want the Chinese pack to auto-update before each Hermes launch.
How do you install or deploy this agent?
The recommended path is sending one sentence to Hermes, which reads the official site and installs automatically:
text
请访问 useai.live/hermes 安装汉化补丁
You can also run the one-shot installer from a terminal:
bash
bash web/install.shWindows uses the PowerShell entry install.ps1; macOS can double-click install.command. All installation methods, system requirements, and troubleshooting are documented at the official site: https://useai.live/hermesHow do you use this agent?
After installation, the patch sets display.language=zh; restart Hermes and the startup banner, quick commands, tool sets, and gateway status appear in Chinese. Optional auto-update runs through the wrapper command, which checks latest. before launch and re-runs the installer when a new version exists:
bash
~/.xiaoma-hermes/bin/hermesFor local testing, point the installer at a local base URL:
bash
XIAOMA_HERMES_BASE_URL=http://127.0.0.1:4173 bash web/install.shWhat are this agent's strengths and limitations?
- One command covers Windows, macOS, Linux, and WSL2, and auto-installs the official desktop app if missing.
- Matches Chinese packs to Hermes versions (0.15.x/0.14.x/0.13.x/legacy) with automatic fallback to the compatibility pack.
- Verifies SHA256 before install and backs up files before modification, reducing patch risk.
- Built-in update mechanism via SKILL.md and the wrapper command keeps the Chinese pack current through latest..
- Only useful within the Hermes ecosystem; it does nothing for other agent frameworks.
- Modifies hermes_cli and ui-tui program files, so Hermes upgrades may require reinstalling the patch.
- Installation and updates depend on the online useai.live/hermes page and latest., so it cannot run fully offline.
- Troubleshooting and version-fit details live mainly on the website rather than in the repository, which has thin docs.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Xiaoma AI Hermes Chinese Enhancement Pack This agent | 36 · Major gaps | — | ★ 16 | 1mo ago | Shell | — |
| Hermes Agent CN Desktop | 41 · Major gaps | Desktop appFree + model costs | ★ 1.7k | 6d ago | TypeScript | — |
| senpi | 74 · Some gaps | CLIFree + model costs | ★ 453 | today | TypeScript | OpenAI API · Claude API |
| zot Coding Agent | 59 · Major gaps | CLIFree + model costs | ★ 345 | 2d ago | Go | OpenAI API · Claude API |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence is limited to README and LICENSE; web/install.sh itself is not provided. The README claims the installer detects versions, installs official Hermes, applies patches, and adds a startup update helper — modifying the local Hermes installation and triggering network downloads with no code to verify scope. A persistent 'startup update helper' is an external effect of unknown risk. Boundary claims ('does not read conversations/API keys') have no code support. Rollback is only a vague 'backed-up patches' mention. Attribution points only to an unverified publisher and the external site useai.live. Deducted: all trust claims lack implementation evidence, and the core flow requires an Agent to execute commands based on an external web page.
The README's version coverage and file listing are broadly self-consistent, but the version matrix conflicts: the description says v0.14.x/0.13.x/legacy while coverage claims 0.15.x, and the implementation is unverifiable. Dependency availability (official installer, network page) cannot be assessed. No evidence at all of failure/error messaging — scored 0. Deducted: version-statement conflicts and unverifiable implementation.
Audience is clear (Chinese Hermes users) and multi-version coverage (0.2–0.15) shows breadth — scored 2. Capability boundaries are only negative safety claims; behavior on patch failure or version mismatch is unstated. Triggering depends on an Agent parsing one sentence and visiting an external page; trigger precision and environment fit (OS, install paths) are unevidenced. Deducted: boundaries, trigger, and environment claims lack testable description.
The file-structure listing is clear and dual-language READMEs are provided — scored 2. Install instructions defer to an external website with no reviewable steps in-repo. Naming wavers between 'Hermes Chinese Enhancement Pack' and '汉化补丁'. No FAQ; no known-limitations section (Boundaries is a safety claim, not functional limits). LICENSE is MIT but dated 'Copyright (c) 2026', a suspicious year. No CHANGELOG; latest. content not provided. Maintenance responsibility (cadence, contact) unevidenced. Deducted: nearly every convention item besides structure lacks direct material.
The output (localized UI) is visible only at description level with no screenshots or patch-file samples — scored 1. Marginal value is real: official language settings reportedly leave some text untranslated — scored 2. Cost-benefit: the user must let an Agent run an install flow pointing at an unverified third-party site in exchange for UI text translation; asymmetric and unproven — scored 1. Deducted: all benefit claims are unverified.
Core claims (version detection, patch application, not reading keys) have no code, tests, or patch files in the provided material to trace — claim_traceability scored 0. No second source for corroboration — scored 0. The README separates factual description from boundary statements, but those boundary statements are unverifiable safety assertions — scored 1. Deducted: nearly all key claims are statically unverifiable.
- The install flow relies on an Agent reading instructions from the third-party site useai.live and executing commands, while the installer script itself is not reviewable here — a high-risk supply-chain path; do not install before auditing web/install.sh and the patch contents.
- Safety boundary claims in the README ('does not read conversations/API keys') are assertions with no supporting code; do not trust them on that basis.
- The 'startup update helper' is a persistent external effect whose network behavior and update source are undisclosed.
- Unverified publisher plus a copyright year of 2026 warrant further verification of project identity and timeline.