Hermes Speech v0.3.3

What’s changed

• Fixed plugin startup in Hermes managed runtimes where PyYAML is not installed.
• The update checker now uses the plugin version already parsed by Hermes instead of reading plugin.yaml itself.
• No new Python dependency is required.

After updating, restart the Hermes gateway.

Hermes Speech v0.3.2

What’s changed

• Update-available notices now remind users to restart the Hermes gateway after an update completes.
• The reminder is included in /speech update output, automatic /speech notices, structured management-tool results, and the bundled agent skills.
• Update checks remain read-only and never install or modify plugin source automatically.

Hermes Speech v0.3.1

What’s changed

• Preview voices directly from voice-search results with /speech tts voice preview [optional sample text]. Previews use the selected TTS model and speed without changing the saved voice.
• Voice-search results now explain how to preview and select a result.
• Balance responses no longer expose individual promotional grant names, eligibility, or expiry details. Paid and aggregate promotional balance totals remain available.

Updating

Ask your Hermes agent: “Update the hermes-speech plugin.”

The /speech update command remains a read-only update check and never modifies plugin files.

Hermes Speech v0.3.0

Hermes Speech v0.3.0 makes plugin updates notification-only and prepares the plugin for Hermes catalog validation.

Highlights:

• /speech update now checks the latest stable release and tells the user to ask their agent to update; it never modifies plugin source.
• Removed all self-update, Git, subprocess, and checkout-replacement code from the plugin.
• Legacy update_plugin tool calls now degrade to a read-only check and host-managed update instruction.
• Bundled workflows now use namespaced ctx.register_skill() registration without modifying skills.external_dirs.
• Declared both registered tools and compatibility metadata in plugin.yaml.
• Catalog admission validation passes capability, collision, and security checks with no warnings.

To apply future updates, ask Hermes to update the plugin or run hermes plugins update hermes-speech, then restart Hermes.

Hermes Speech v0.2.0

What’s changed

• Updated voice discovery for the new AllModels voices API with server-side text search across names, descriptions, languages, categories, and labels.
• Avoids downloading or caching full voice-catalogue rows; only bounded compact metadata from returned searches is retained for fallback.
• Improved voice compatibility handling across model selection, setup, commands, and native TTS.
• Updated the management skill to offer playable voice samples after presenting candidates without changing the saved TTS configuration.
• Expanded integration coverage for catalogue search, caching, configuration, and provider behavior.

AllModels: https://allmodels.io

Hermes Speech v0.1.1

Patch release fixing update-check test isolation.

• Keeps simulated release metadata inside each test's temporary Hermes home.
• Waits for the final background refresh in the coalescing/rate-limit test.
• Prevents test runs from writing fake update metadata into a user's real ~/.hermes cache.

No runtime speech behavior or defaults changed.

Hermes Speech v0.1.0

Initial release of the native AllModels speech plugin for Hermes Agent.

Includes guided and conversational setup, native TTS/STT providers, model and voice management, signup, balance and top-up support, synchronous and streaming TTS, safe diagnostics, and opt-out stable-release update checks with explicit self-update support.