mars: add KittenTTS voice provider (CPU-only, offline mini model)
tts.provider: kittentts, voice Luna, mini-0.8 model (80M params) per darman. CPU-only onnxruntime inference — no GPU on mars, validated ~1x realtime on a weaker dev box. Upstream's misaki[en]+spacy declaration is deliberately not honored (pulls torch→CUDA, 5.6 GB); a .pth-registered stub satisfies the dead misaki import, failing loudly if it is ever actually used. HF model cache pre-seeded from hash-pinned store paths so HF_HUB_OFFLINE=1 runs with zero boot-time network. Provisioner shaped by the Mnemosyne review: uv venv --clear, root never executes python from the venv, ownership to the container uid before anything imports, stamp checked against venv+model files.
This commit is contained in:
@@ -77,6 +77,33 @@ let
|
||||
# Mnemosyne memory provider: third-party plugin, not in the image.
|
||||
mnemosyneEnv = pkgs.callPackage ../../pkgs/mnemosyne-env.nix { };
|
||||
|
||||
# KittenTTS voice provider inputs (CPU-only; model + wheel hash-pinned).
|
||||
# Provisioning unit near the bottom of this file; background on why the
|
||||
# deps deviate from upstream's declaration lives in
|
||||
# ./kittentts/requirements.txt + ./kittentts/kitten-misaki-stub.py.
|
||||
kittenttsWheel = pkgs.fetchurl {
|
||||
url = "https://github.com/KittenML/KittenTTS/releases/download/0.8.1/kittentts-0.8.1-py3-none-any.whl";
|
||||
sha256 = "sha256-SCpDbE8fMZIVNxA3bkWf82iVF+vNp8KwUeL9QYe0GFE=";
|
||||
};
|
||||
kittenttsReqs = pkgs.writeText "kittentts-requirements.txt" (
|
||||
builtins.readFile ./kittentts/requirements.txt
|
||||
);
|
||||
kittenttsStub = pkgs.writeText "kitten-tts-stub.py" (
|
||||
builtins.readFile ./kittentts/kitten-misaki-stub.py
|
||||
);
|
||||
kittenttsModelOnnx = pkgs.fetchurl {
|
||||
url = "https://huggingface.co/KittenML/kitten-tts-mini-0.8/resolve/main/kitten_tts_mini_v0_8.onnx";
|
||||
sha256 = "sha256-D1u65PxIAMmNvFRKh+z6eVEN4vuCItsw0S5b/pF335E=";
|
||||
};
|
||||
kittenttsModelVoices = pkgs.fetchurl {
|
||||
url = "https://huggingface.co/KittenML/kitten-tts-mini-0.8/resolve/main/voices.npz";
|
||||
sha256 = "sha256-QK0mOJUrd7ey8wEn4mCOFp/GndJWtTvYqqNAmjMZPEI=";
|
||||
};
|
||||
kittenttsModelConfig = pkgs.fetchurl {
|
||||
url = "https://huggingface.co/KittenML/kitten-tts-mini-0.8/resolve/main/config.json";
|
||||
sha256 = "sha256-axYLybGeJOyyHoS8FPin2iH99H7HLUJFC8XPUUthgEo=";
|
||||
};
|
||||
|
||||
# Wire event names (X-GitHub-Event) each route accepts — NOT the
|
||||
# subscription names the gitea hooks in services/dev/gitea.nix use. The two
|
||||
# namespaces collide; see the long comment on the route unit below.
|
||||
@@ -450,4 +477,90 @@ in
|
||||
chown -h ${hermesUid}:${hermesGid} "$pluginDir"
|
||||
'';
|
||||
};
|
||||
|
||||
# ---- KittenTTS voice provider ------------------------------------------
|
||||
# CPU-only onnxruntime TTS (no GPU on mars), mini model per darman. The
|
||||
# upstream `misaki[en]` declaration is deliberately not honored — it pulls
|
||||
# torch→CUDA (5.6 GB verified); the runtime phonemizes with espeak-ng only,
|
||||
# so the dead `from misaki import en, espeak` import is satisfied by a
|
||||
# .pth-registered stub (kitten-misaki-stub.py) that fails loudly if misaki
|
||||
# is ever actually used.
|
||||
#
|
||||
# Provisioner invariants (shaped by the Mnemosyne-round review):
|
||||
# - `uv venv --clear`: re-provision cannot wedge on an existing dir.
|
||||
# - Root runs NO python from this venv: the unit itself does only fs
|
||||
# writes; the venv is chowned to the container uid before Hermes ever
|
||||
# imports from it. (Hermes executes provider code as uid 986.)
|
||||
# - HF model cache is PRE-SEEDED from hash-pinned store paths, so
|
||||
# HF_HUB_OFFLINE=1 gives zero boot-time network and no drift.
|
||||
# - Idempotent via a stamp keyed on the requirements hash; checked
|
||||
# against BOTH the venv python and model files being intact.
|
||||
#
|
||||
# Trust boundary: the venv lives inside hermesHome (HERMES_WRITE_SAFE_ROOT),
|
||||
# so luna can technically alter her own TTS engine. Deliberate: it's her
|
||||
# voice, not her jail — the webhook filter scripts remain the only
|
||||
# write-protected-but-load-bearing items.
|
||||
systemd.services.hermes-agent-kittentts-provision = {
|
||||
description = "Provision KittenTTS voice provider (side venv + offline HF cache)";
|
||||
before = [ "podman-hermes-agent.service" ];
|
||||
wantedBy = [ "podman-hermes-agent.service" ];
|
||||
wants = [ "network-online.target" ];
|
||||
after = [ "network-online.target" ];
|
||||
path = [ pkgs.uv pkgs.coreutils ];
|
||||
serviceConfig = {
|
||||
Type = "oneshot";
|
||||
TimeoutStartSec = 600;
|
||||
};
|
||||
script = ''
|
||||
set -euo pipefail
|
||||
|
||||
venv=${hermesHome}/kittentts-venv
|
||||
hubDir=${hermesHome}/kittentts-hf/hub/models--KittenML--kitten-tts-mini-0.8
|
||||
snap=$hubDir/snapshots/kitten
|
||||
stampFile=${hermesHome}/kittentts-provision.stamp
|
||||
reqHash=$(sha256sum ${kittenttsReqs} | cut -d' ' -f1)
|
||||
|
||||
# Idempotent early exit: stamp + venv + all three model files intact.
|
||||
if [ -f "$stampFile" ] && [ "$(cat "$stampFile")" = "$reqHash" ] \
|
||||
&& [ -x "$venv/bin/python" ] \
|
||||
&& [ -f "$snap/kitten_tts_mini_v0_8.onnx" ] \
|
||||
&& [ -f "$snap/voices.npz" ] \
|
||||
&& [ -f "$venv/lib/python3.13/site-packages/kitten_tts_stub.py" ]; then
|
||||
exit 0
|
||||
fi
|
||||
|
||||
# Venv (rebuilt rather than broken on --no-clear: uv exits 2 otherwise).
|
||||
uv venv "$venv" --python ${pkgs.python313}/bin/python3 --clear --quiet
|
||||
UV_CACHE_DIR=${hermesHome}/kittentts-uv-cache \
|
||||
uv pip install --python "$venv/bin/python" --quiet \
|
||||
--requirement ${kittenttsReqs}
|
||||
# kittentts --no-deps: its overlay of spacy/misaki[en] is what drags in
|
||||
# the CUDA tree; the requirements freeze already covers its real needs.
|
||||
UV_CACHE_DIR=${hermesHome}/kittentts-uv-cache \
|
||||
uv pip install --python "$venv/bin/python" --quiet --no-deps \
|
||||
${kittenttsWheel}
|
||||
|
||||
# Dead-import shim: .pth auto-loads kitten_tts_stub at interpreter start
|
||||
# so `from misaki import en, espeak` resolves without the real misaki.en.
|
||||
siteDir=$("$venv/bin/python" -c 'import sysconfig; print(sysconfig.get_paths()["purelib"])')
|
||||
cp ${kittenttsStub} "$siteDir/kitten_tts_stub.py"
|
||||
printf 'import kitten_tts_stub\n' > "$siteDir/zz-kitten-stub.pth"
|
||||
|
||||
# Seed the HF cache with the hashed model files (exact hub layout;
|
||||
# hf_hub_download scans refs/snapshots on disk offline).
|
||||
mkdir -p "$hubDir/refs" "$snap"
|
||||
install -m 0444 ${kittenttsModelOnnx} "$snap/kitten_tts_mini_v0_8.onnx"
|
||||
install -m 0444 ${kittenttsModelVoices} "$snap/voices.npz"
|
||||
install -m 0444 ${kittenttsModelConfig} "$snap/config.json"
|
||||
|
||||
# Hand ownership to the container uid BEFORE any python runs in this
|
||||
# tree (root never imports from it — that was Mnemosyne review #3).
|
||||
chown -R ${hermesUid}:${hermesGid} "$venv" "$hubDir"
|
||||
|
||||
# Stamp LAST — a half-provisioned venv fails the integrity check and
|
||||
# re-provisions on next boot rather than being trusted.
|
||||
printf '%s' "$reqHash" > "$stampFile"
|
||||
chown ${hermesUid}:${hermesGid} "$stampFile"
|
||||
'';
|
||||
};
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user