From 02cb745046062b142eaf1e5a4892f9fe93ae17d1 Mon Sep 17 00:00:00 2001 From: 100yenadmin Date: Mon, 5 Oct 2026 18:15:52 +0900 Subject: [PATCH 1/8] Add Hermes Agent integration page --- developer-guide/integrations/hermes-agent.mdx | 78 +++++++++++++++++++ docs.json | 3 +- 2 files changed, 80 insertions(+), 1 deletion(-) create mode 100644 developer-guide/integrations/hermes-agent.mdx diff --git a/developer-guide/integrations/hermes-agent.mdx b/developer-guide/integrations/hermes-agent.mdx new file mode 100644 index 00000000..21881058 --- /dev/null +++ b/developer-guide/integrations/hermes-agent.mdx @@ -0,0 +1,78 @@ +--- +title: "Hermes Agent" +description: "Give Hermes Agent a Fish Audio voice with text-to-speech, speech-to-text, voice cloning, and voice design" +icon: "terminal" +--- + +[Hermes Agent](https://github.com/NousResearch/hermes-agent) is an open-source AI agent by Nous Research. It runs in the terminal, in messaging apps such as Telegram and Discord, and in a desktop app. The official Fish Audio plugin for Hermes makes Fish Audio the agent's speech and transcription provider. It also gives the agent tools to search, clone, and design voices. + +## Prerequisites + +- Hermes Agent 0.21.5 or newer, on macOS or Linux +- A Fish Audio API key from [fish.audio/app/api-keys](https://fish.audio/app/api-keys) + +## Installation + +Install the plugin and enable it: + +```bash +hermes plugins install https://github.com/100yenadmin/fish-audio-hermes --enable +``` + +Hermes 0.21.5 installs the plugin's Python dependencies automatically. Newer Hermes builds ask first; add `--yes-deps` to answer yes. + +## Configuration + +Save your API key on the machine that runs Hermes: + +```bash +hermes fish login +``` + +`login` checks the key against your Fish Audio account and saves it to the active Hermes profile. Where no speech or transcription provider is set, it selects Fish Audio; it asks before it replaces another provider. + +You can also add the key in the Hermes desktop app under **Plugins** > **Fish Audio**, or with `hermes tools` > **Text-to-Speech** > **Fish Audio**. + +Check the setup: + +```bash +hermes fish doctor +``` + +`doctor` makes one short billed synthesis. Add `--no-synth` to skip it. Restart a running Hermes gateway or the desktop app to pick up a new key. + +## Choose a voice + +Search the voice library from any chat and switch voices: + +```text +/fish voices narrator +/fish use +``` + +Or set a voice in the profile's `config.yaml`: + +```yaml +tts: + provider: fish-audio + fish-audio: + voice: 933563129e564b19a115bedd57b7406a # Sarah + model: s2.1-pro +``` + +## Features + +- **Text-to-speech**: voice replies, read-aloud, and voice notes use Fish Audio. CLI voice mode and desktop voice replies stream sentence by sentence. +- **Speech-to-text**: voice notes and desktop dictation are transcribed with `transcribe-1-pro`. +- **Agent tools**: + - `fish_speak` for expressive and multi-speaker speech, with word timestamps and subtitles; + - `fish_voices` to search, clone, design, and manage voices; + - `fish_transcribe` for transcripts with speaker turns and SRT subtitles. +- **Commands**: `/fish status`, `/fish voices`, `/fish use`, `/fish model`, `/fish preview`, and `/fish balance` work in every chat. `hermes fish login`, `status`, `doctor`, and `use` work in the terminal. + +Synthesis, transcription, cloning, and voice design are billed to your Fish Audio API credits, except requests on the free `s2.1-pro-free` model. Cloning and deleting voices go through Hermes's approval prompt. + +## Resources + +- [GitHub](https://github.com/100yenadmin/fish-audio-hermes) +- [Hermes Agent](https://github.com/NousResearch/hermes-agent) diff --git a/docs.json b/docs.json index 17f06c13..3bd1cf57 100644 --- a/docs.json +++ b/docs.json @@ -211,7 +211,8 @@ "developer-guide/integrations/livekit", "developer-guide/integrations/n8n", "developer-guide/integrations/telnyx", - "developer-guide/integrations/slng" + "developer-guide/integrations/slng", + "developer-guide/integrations/hermes-agent" ] }, { From a3e4dbe6951f1d6625f984cdfe5b8f76289eba3c Mon Sep 17 00:00:00 2001 From: Eva Date: Mon, 5 Oct 2026 19:49:59 +0900 Subject: [PATCH 2/8] Hermes Agent: the desktop Voices page --- developer-guide/integrations/hermes-agent.mdx | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/developer-guide/integrations/hermes-agent.mdx b/developer-guide/integrations/hermes-agent.mdx index 21881058..fa0b899d 100644 --- a/developer-guide/integrations/hermes-agent.mdx +++ b/developer-guide/integrations/hermes-agent.mdx @@ -50,6 +50,8 @@ Search the voice library from any chat and switch voices: /fish use ``` +In the Hermes desktop app, open **Voices** to search the library, preview a voice, and click **Use**. To add the page, switch on **Desktop** for Fish Audio under **Capabilities** > **Plugins**, then restart the Hermes gateway once. + Or set a voice in the profile's `config.yaml`: ```yaml @@ -68,9 +70,10 @@ tts: - `fish_speak` for expressive and multi-speaker speech, with word timestamps and subtitles; - `fish_voices` to search, clone, design, and manage voices; - `fish_transcribe` for transcripts with speaker turns and SRT subtitles. +- **Voices page in the desktop app**: browse and preview the voice library, choose the agent's voice, clone or design voices, and check your API credit. - **Commands**: `/fish status`, `/fish voices`, `/fish use`, `/fish model`, `/fish preview`, and `/fish balance` work in every chat. `hermes fish login`, `status`, `doctor`, and `use` work in the terminal. -Synthesis, transcription, cloning, and voice design are billed to your Fish Audio API credits, except requests on the free `s2.1-pro-free` model. Cloning and deleting voices go through Hermes's approval prompt. +Synthesis, transcription, cloning, and voice design are billed to your Fish Audio API credits, except requests on the free `s2.1-pro-free` model. Cloning and deleting voices from chat go through Hermes's approval prompt; on the desktop Voices page, you confirm them on the page. ## Resources From cdbe3f9519f7b7c66940e8f9ca187b5531e61769 Mon Sep 17 00:00:00 2001 From: Eva Date: Mon, 5 Oct 2026 20:59:48 +0900 Subject: [PATCH 3/8] Hermes Agent: noninteractive login and free-model wording --- developer-guide/integrations/hermes-agent.mdx | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/developer-guide/integrations/hermes-agent.mdx b/developer-guide/integrations/hermes-agent.mdx index fa0b899d..5f2427d6 100644 --- a/developer-guide/integrations/hermes-agent.mdx +++ b/developer-guide/integrations/hermes-agent.mdx @@ -29,7 +29,7 @@ Save your API key on the machine that runs Hermes: hermes fish login ``` -`login` checks the key against your Fish Audio account and saves it to the active Hermes profile. Where no speech or transcription provider is set, it selects Fish Audio; it asks before it replaces another provider. +`login` checks the key against your Fish Audio account and saves it to the active Hermes profile. Where no speech or transcription provider is set, it selects Fish Audio. It asks before it replaces another provider; run without a terminal, it keeps the other provider unless you pass `--yes`. You can also add the key in the Hermes desktop app under **Plugins** > **Fish Audio**, or with `hermes tools` > **Text-to-Speech** > **Fish Audio**. @@ -73,7 +73,7 @@ tts: - **Voices page in the desktop app**: browse and preview the voice library, choose the agent's voice, clone or design voices, and check your API credit. - **Commands**: `/fish status`, `/fish voices`, `/fish use`, `/fish model`, `/fish preview`, and `/fish balance` work in every chat. `hermes fish login`, `status`, `doctor`, and `use` work in the terminal. -Synthesis, transcription, cloning, and voice design are billed to your Fish Audio API credits, except requests on the free `s2.1-pro-free` model. Cloning and deleting voices from chat go through Hermes's approval prompt; on the desktop Voices page, you confirm them on the page. +Synthesis, transcription, cloning, and voice design are billed to your Fish Audio API credits, except requests that run on the free `s2.1-pro-free` model. Cloning and deleting voices from chat go through Hermes's approval prompt; on the desktop Voices page, you confirm them on the page. ## Resources From 3a7c485a21971d6111fc89a5c872d6cea962b3bb Mon Sep 17 00:00:00 2001 From: Eva Date: Sat, 10 Oct 2026 15:15:23 +0900 Subject: [PATCH 4/8] Hermes Agent: install from the plugin catalog; streaming needs Hermes 0.21.6 --- developer-guide/integrations/hermes-agent.mdx | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/developer-guide/integrations/hermes-agent.mdx b/developer-guide/integrations/hermes-agent.mdx index 5f2427d6..e135ad27 100644 --- a/developer-guide/integrations/hermes-agent.mdx +++ b/developer-guide/integrations/hermes-agent.mdx @@ -13,12 +13,14 @@ icon: "terminal" ## Installation -Install the plugin and enable it: +Fish Audio is in the Hermes plugin catalog. Install it and enable it: ```bash -hermes plugins install https://github.com/100yenadmin/fish-audio-hermes --enable +hermes plugins install fish-audio --enable ``` +With the Hermes desktop app installed, you can open [hermes://plugin/install?catalog=fish-audio](hermes://plugin/install?catalog=fish-audio) instead. The app finds Fish Audio in the catalog and asks you to confirm the install. + Hermes 0.21.5 installs the plugin's Python dependencies automatically. Newer Hermes builds ask first; add `--yes-deps` to answer yes. ## Configuration @@ -64,7 +66,7 @@ tts: ## Features -- **Text-to-speech**: voice replies, read-aloud, and voice notes use Fish Audio. CLI voice mode and desktop voice replies stream sentence by sentence. +- **Text-to-speech**: voice replies, read-aloud, and voice notes use Fish Audio. On Hermes 0.21.6 or newer, CLI voice mode and desktop voice replies stream sentence by sentence. Hermes 0.21.5 waits for each sentence in CLI voice mode, and for the whole reply in the desktop app, to finish synthesizing before it plays. - **Speech-to-text**: voice notes and desktop dictation are transcribed with `transcribe-1-pro`. - **Agent tools**: - `fish_speak` for expressive and multi-speaker speech, with word timestamps and subtitles; From eaba1f3db0ed404180b3579004a32fa3c4ac136a Mon Sep 17 00:00:00 2001 From: Eva Date: Sun, 11 Oct 2026 18:28:23 +0900 Subject: [PATCH 5/8] Hermes Agent: voice names and the Fish Audio settings page (plugin 1.4.0) --- developer-guide/integrations/hermes-agent.mdx | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-) diff --git a/developer-guide/integrations/hermes-agent.mdx b/developer-guide/integrations/hermes-agent.mdx index e135ad27..0df7cb66 100644 --- a/developer-guide/integrations/hermes-agent.mdx +++ b/developer-guide/integrations/hermes-agent.mdx @@ -52,7 +52,14 @@ Search the voice library from any chat and switch voices: /fish use ``` -In the Hermes desktop app, open **Voices** to search the library, preview a voice, and click **Use**. To add the page, switch on **Desktop** for Fish Audio under **Capabilities** > **Plugins**, then restart the Hermes gateway once. +Save a short name for a voice, then use the name wherever a voice ID goes: + +```text +/fish alias add narrator +/fish use narrator +``` + +In the Hermes desktop app, open **Voices** to search the library, preview a voice, and click **Use**. To add the page, switch on **Desktop** for Fish Audio under **Capabilities** > **Plugins**, then restart the Hermes gateway once. On Hermes 0.21.6 or newer, **Settings** > **Plugins** > **Fish Audio** shows the agent's voice, speech model, and saved voice names, and lets you change them. Or set a voice in the profile's `config.yaml`: @@ -73,7 +80,7 @@ tts: - `fish_voices` to search, clone, design, and manage voices; - `fish_transcribe` for transcripts with speaker turns and SRT subtitles. - **Voices page in the desktop app**: browse and preview the voice library, choose the agent's voice, clone or design voices, and check your API credit. -- **Commands**: `/fish status`, `/fish voices`, `/fish use`, `/fish model`, `/fish preview`, and `/fish balance` work in every chat. `hermes fish login`, `status`, `doctor`, and `use` work in the terminal. +- **Commands**: `/fish status`, `/fish voices`, `/fish use`, `/fish model`, `/fish preview`, `/fish alias`, and `/fish balance` work in every chat. `hermes fish login`, `status`, `doctor`, and `use` work in the terminal. Synthesis, transcription, cloning, and voice design are billed to your Fish Audio API credits, except requests that run on the free `s2.1-pro-free` model. Cloning and deleting voices from chat go through Hermes's approval prompt; on the desktop Voices page, you confirm them on the page. From a26a953cd25caf7017ff02b6ea7e51d4ff0a37aa Mon Sep 17 00:00:00 2001 From: Eva Date: Mon, 12 Oct 2026 00:07:08 +0900 Subject: [PATCH 6/8] Hermes Agent: live dictation (plugin 1.5.0) --- developer-guide/integrations/hermes-agent.mdx | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/developer-guide/integrations/hermes-agent.mdx b/developer-guide/integrations/hermes-agent.mdx index 0df7cb66..75a3d928 100644 --- a/developer-guide/integrations/hermes-agent.mdx +++ b/developer-guide/integrations/hermes-agent.mdx @@ -74,7 +74,7 @@ tts: ## Features - **Text-to-speech**: voice replies, read-aloud, and voice notes use Fish Audio. On Hermes 0.21.6 or newer, CLI voice mode and desktop voice replies stream sentence by sentence. Hermes 0.21.5 waits for each sentence in CLI voice mode, and for the whole reply in the desktop app, to finish synthesizing before it plays. -- **Speech-to-text**: voice notes and desktop dictation are transcribed with `transcribe-1-pro`. +- **Speech-to-text**: voice notes and desktop dictation are transcribed with `transcribe-1-pro`. On Hermes 0.21.6 or newer, set `stt.streaming: true` for live dictation: in CLI voice mode, the TUI, and the desktop composer mic, the text appears phrase by phrase while you speak. Live dictation streams microphone audio to Fish Audio's realtime endpoint and is billed per phrase at the speech-to-text price. - **Agent tools**: - `fish_speak` for expressive and multi-speaker speech, with word timestamps and subtitles; - `fish_voices` to search, clone, design, and manage voices; From e804f7c50edc21ca933aae665d935a562da02c6b Mon Sep 17 00:00:00 2001 From: Eva Date: Tue, 13 Oct 2026 11:32:54 +0900 Subject: [PATCH 7/8] Hermes Agent: install from the Desktop catalog picker instead of the hermes:// link hermes://plugin/install?catalog= links fail in Hermes Desktop until NousResearch/hermes-agent#136483 is fixed. Point to Capabilities > Plugins and the catalog page instead. --- developer-guide/integrations/hermes-agent.mdx | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/developer-guide/integrations/hermes-agent.mdx b/developer-guide/integrations/hermes-agent.mdx index 75a3d928..2460046e 100644 --- a/developer-guide/integrations/hermes-agent.mdx +++ b/developer-guide/integrations/hermes-agent.mdx @@ -19,7 +19,7 @@ Fish Audio is in the Hermes plugin catalog. Install it and enable it: hermes plugins install fish-audio --enable ``` -With the Hermes desktop app installed, you can open [hermes://plugin/install?catalog=fish-audio](hermes://plugin/install?catalog=fish-audio) instead. The app finds Fish Audio in the catalog and asks you to confirm the install. +In the Hermes desktop app, you can instead open **Capabilities** > **Plugins**, search the catalog for Fish Audio, and install it there. The plugin's catalog page is [hermes-agent.nousresearch.com/docs/plugins/fish-audio](https://hermes-agent.nousresearch.com/docs/plugins/fish-audio). Hermes 0.21.5 installs the plugin's Python dependencies automatically. Newer Hermes builds ask first; add `--yes-deps` to answer yes. From 6b04c9c90873f6564e35450dc95c9b73bb82fda7 Mon Sep 17 00:00:00 2001 From: Eva Date: Wed, 14 Oct 2026 02:38:00 +0900 Subject: [PATCH 8/8] Hermes Agent: Drama 3 with plugin 1.6.0 --- developer-guide/integrations/hermes-agent.mdx | 1 + 1 file changed, 1 insertion(+) diff --git a/developer-guide/integrations/hermes-agent.mdx b/developer-guide/integrations/hermes-agent.mdx index 2460046e..80882275 100644 --- a/developer-guide/integrations/hermes-agent.mdx +++ b/developer-guide/integrations/hermes-agent.mdx @@ -74,6 +74,7 @@ tts: ## Features - **Text-to-speech**: voice replies, read-aloud, and voice notes use Fish Audio. On Hermes 0.21.6 or newer, CLI voice mode and desktop voice replies stream sentence by sentence. Hermes 0.21.5 waits for each sentence in CLI voice mode, and for the whole reply in the desktop app, to finish synthesizing before it plays. +- **Drama 3**: with plugin 1.6.0 or newer, `/fish model drama-3` makes [Drama 3](/features/text-to-speech/drama-3) the agent's speech model. Directions in square brackets work in every voice reply. Agents send word tags such as `…` through `fish_speak`, which passes them to Fish Audio unchanged. - **Speech-to-text**: voice notes and desktop dictation are transcribed with `transcribe-1-pro`. On Hermes 0.21.6 or newer, set `stt.streaming: true` for live dictation: in CLI voice mode, the TUI, and the desktop composer mic, the text appears phrase by phrase while you speak. Live dictation streams microphone audio to Fish Audio's realtime endpoint and is billed per phrase at the speech-to-text price. - **Agent tools**: - `fish_speak` for expressive and multi-speaker speech, with word timestamps and subtitles;