parakeet v2 support (#116)

* parakeet v2 support

* Update model.rs

* fix parakeet size

* new onboarding UI for parakeet v2 introduction

* tweak height
This commit is contained in:
CJ Pais 2025-10-28 16:32:26 -07:00 committed by GitHub
parent ae406aebe2
commit 65c1e2dbcb
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
8 changed files with 130 additions and 59 deletions

View file

@ -31,6 +31,8 @@ pub struct ModelInfo {
pub partial_size: u64, pub partial_size: u64,
pub is_directory: bool, pub is_directory: bool,
pub engine_type: EngineType, pub engine_type: EngineType,
pub accuracy_score: f32, // 0.0 to 1.0, higher is more accurate
pub speed_score: f32, // 0.0 to 1.0, higher is faster
} }
#[derive(Debug, Clone, Serialize, Deserialize)] #[derive(Debug, Clone, Serialize, Deserialize)]
@ -62,6 +64,7 @@ impl ModelManager {
let mut available_models = HashMap::new(); let mut available_models = HashMap::new();
// TODO this should be read from a JSON file or something..
available_models.insert( available_models.insert(
"small".to_string(), "small".to_string(),
ModelInfo { ModelInfo {
@ -76,6 +79,8 @@ impl ModelManager {
partial_size: 0, partial_size: 0,
is_directory: false, is_directory: false,
engine_type: EngineType::Whisper, engine_type: EngineType::Whisper,
accuracy_score: 0.60,
speed_score: 0.85,
}, },
); );
@ -94,6 +99,8 @@ impl ModelManager {
partial_size: 0, partial_size: 0,
is_directory: false, is_directory: false,
engine_type: EngineType::Whisper, engine_type: EngineType::Whisper,
accuracy_score: 0.75,
speed_score: 0.60,
}, },
); );
@ -111,6 +118,8 @@ impl ModelManager {
partial_size: 0, partial_size: 0,
is_directory: false, is_directory: false,
engine_type: EngineType::Whisper, engine_type: EngineType::Whisper,
accuracy_score: 0.80,
speed_score: 0.40,
}, },
); );
@ -128,10 +137,31 @@ impl ModelManager {
partial_size: 0, partial_size: 0,
is_directory: false, is_directory: false,
engine_type: EngineType::Whisper, engine_type: EngineType::Whisper,
accuracy_score: 0.85,
speed_score: 0.30,
},
);
// Add NVIDIA Parakeet models (directory-based)
available_models.insert(
"parakeet-tdt-0.6b-v2".to_string(),
ModelInfo {
id: "parakeet-tdt-0.6b-v2".to_string(),
name: "Parakeet V2".to_string(),
description: "English only. The best model for English speakers.".to_string(),
filename: "parakeet-tdt-0.6b-v2-int8".to_string(), // Directory name
url: Some("https://blob.handy.computer/parakeet-v2-int8.tar.gz".to_string()),
size_mb: 473, // Approximate size for int8 quantized model
is_downloaded: false,
is_downloading: false,
partial_size: 0,
is_directory: true,
engine_type: EngineType::Parakeet,
accuracy_score: 0.85,
speed_score: 0.85,
}, },
); );
// Add NVIDIA Parakeet model (directory-based)
available_models.insert( available_models.insert(
"parakeet-tdt-0.6b-v3".to_string(), "parakeet-tdt-0.6b-v3".to_string(),
ModelInfo { ModelInfo {
@ -140,12 +170,14 @@ impl ModelManager {
description: "Fast and accurate".to_string(), description: "Fast and accurate".to_string(),
filename: "parakeet-tdt-0.6b-v3-int8".to_string(), // Directory name filename: "parakeet-tdt-0.6b-v3-int8".to_string(), // Directory name
url: Some("https://blob.handy.computer/parakeet-v3-int8.tar.gz".to_string()), url: Some("https://blob.handy.computer/parakeet-v3-int8.tar.gz".to_string()),
size_mb: 850, // Approximate size for int8 quantized model size_mb: 478, // Approximate size for int8 quantized model
is_downloaded: false, is_downloaded: false,
is_downloading: false, is_downloading: false,
partial_size: 0, partial_size: 0,
is_directory: true, is_directory: true,
engine_type: EngineType::Parakeet, engine_type: EngineType::Parakeet,
accuracy_score: 0.80,
speed_score: 0.85,
}, },
); );

View file

@ -16,9 +16,9 @@
"label": "main", "label": "main",
"title": "Handy", "title": "Handy",
"width": 680, "width": 680,
"height": 520, "height": 570,
"minWidth": 680, "minWidth": 680,
"minHeight": 520, "minHeight": 570,
"resizable": true, "resizable": true,
"maximizable": false, "maximizable": false,
"visible": false "visible": false

View file

@ -4,7 +4,6 @@ import { Toaster } from "sonner";
import "./App.css"; import "./App.css";
import AccessibilityPermissions from "./components/AccessibilityPermissions"; import AccessibilityPermissions from "./components/AccessibilityPermissions";
import Footer from "./components/footer"; import Footer from "./components/footer";
import HandyTextLogo from "./components/icons/HandyTextLogo";
import Onboarding from "./components/onboarding"; import Onboarding from "./components/onboarding";
import { Sidebar, SidebarSection, SECTIONS_CONFIG } from "./components/Sidebar"; import { Sidebar, SidebarSection, SECTIONS_CONFIG } from "./components/Sidebar";
import { useSettings } from "./hooks/useSettings"; import { useSettings } from "./hooks/useSettings";
@ -67,14 +66,7 @@ function App() {
}; };
if (showOnboarding) { if (showOnboarding) {
return ( return <Onboarding onModelSelected={handleModelSelected} />;
<div className="h-screen flex flex-col">
<div className="flex-1 flex flex-col items-center justify-center p-4 gap-2">
<HandyTextLogo width={200} />
<Onboarding onModelSelected={handleModelSelected} />
</div>
</div>
);
} }
return ( return (

View file

@ -2,12 +2,12 @@ import React from "react";
import { Download } from "lucide-react"; import { Download } from "lucide-react";
import { ModelInfo } from "../../lib/types"; import { ModelInfo } from "../../lib/types";
import { formatModelSize } from "../../lib/utils/format"; import { formatModelSize } from "../../lib/utils/format";
import Badge from "../ui/Badge";
interface ModelCardProps { interface ModelCardProps {
model: ModelInfo; model: ModelInfo;
variant?: "default" | "featured"; variant?: "default" | "featured";
disabled?: boolean; disabled?: boolean;
badgeText?: string;
className?: string; className?: string;
onSelect: (modelId: string) => void; onSelect: (modelId: string) => void;
} }
@ -16,31 +16,18 @@ const ModelCard: React.FC<ModelCardProps> = ({
model, model,
variant = "default", variant = "default",
disabled = false, disabled = false,
badgeText,
className = "", className = "",
onSelect, onSelect,
}) => { }) => {
const isFeatured = variant === "featured"; const isFeatured = variant === "featured";
const baseButtonClasses = const baseButtonClasses =
"relative rounded-xl p-3 p-4 text-left transition-all duration-200 disabled:opacity-50 disabled:cursor-not-allowed focus:outline-none focus:ring-2 focus:ring-logo-primary/25 active:scale-[0.98] cursor-pointer group"; "flex justify-between items-center rounded-xl p-3 px-4 text-left transition-all duration-200 disabled:opacity-50 disabled:cursor-not-allowed focus:outline-none focus:ring-2 focus:ring-logo-primary/25 active:scale-[0.98] cursor-pointer group";
const variantClasses = isFeatured const variantClasses = isFeatured
? "border-2 border-logo-primary/40 bg-logo-primary/5 hover:border-logo-primary/60 hover:bg-logo-primary/10 hover:shadow-lg hover:scale-[1.02] disabled:hover:border-logo-primary/40 disabled:hover:bg-logo-primary/5 disabled:hover:shadow-none disabled:hover:scale-100" ? "border-2 border-logo-primary/25 bg-logo-primary/5 hover:border-logo-primary/40 hover:bg-logo-primary/8 hover:shadow-lg hover:scale-[1.02] disabled:hover:border-logo-primary/25 disabled:hover:bg-logo-primary/5 disabled:hover:shadow-none disabled:hover:scale-100"
: "border-2 border-mid-gray/20 hover:border-logo-primary/50 hover:bg-logo-primary/5 hover:shadow-lg hover:scale-[1.02] disabled:hover:border-mid-gray/20 disabled:hover:bg-transparent disabled:hover:shadow-none disabled:hover:scale-100"; : "border-2 border-mid-gray/20 hover:border-logo-primary/50 hover:bg-logo-primary/5 hover:shadow-lg hover:scale-[1.02] disabled:hover:border-mid-gray/20 disabled:hover:bg-transparent disabled:hover:shadow-none disabled:hover:scale-100";
const titleClasses =
"text-lg font-semibold text-text group-hover:text-logo-primary transition-colors";
const descriptionClasses = "text-text/60 text-sm leading-relaxed";
const sizeRowClasses =
"mt-1 flex items-center gap-2 text-xs text-text/60 tabular-nums";
const containerSpacing = "space-y-3";
const sizeIconClasses = "h-3.5 w-3.5 text-text/45";
return ( return (
<button <button
onClick={() => onSelect(model.id)} onClick={() => onSelect(model.id)}
@ -50,26 +37,36 @@ const ModelCard: React.FC<ModelCardProps> = ({
.join(" ")} .join(" ")}
type="button" type="button"
> >
{badgeText && ( <div className="flex flex-col items-ce">
<div className="absolute -top-2 -right-2 bg-logo-primary text-white text-sm px-4 py-2 rounded-full font-medium shadow-md"> <div className="flex items-center gap-4">
{badgeText} <h3 className="text-lg font-semibold text-text group-hover:text-logo-primary transition-colors">
{model.name}
</h3>
<DownloadSize sizeMb={model.size_mb} />
{isFeatured && <Badge variant="primary">Recommended</Badge>}
</div> </div>
)} <p className="text-text/60 text-sm leading-relaxed">
{model.description}
</p>
</div>
<div className={containerSpacing}> <div className="space-y-1">
<div className="space-y-0"> <div className="flex items-center gap-2">
<h3 className={titleClasses}>{model.name}</h3> <p className="text-xs text-text/70 w-16 text-right">accuracy</p>
<p className={descriptionClasses}>{model.description}</p> <div className="w-20 h-2 bg-mid-gray/20 rounded-full overflow-hidden">
<div className={sizeRowClasses}> <div
<Download className="h-full bg-logo-primary rounded-full transition-all duration-300"
aria-hidden="true" style={{ width: `${model.accuracy_score * 100}%` }}
className={sizeIconClasses} />
strokeWidth={1.75} </div>
</div>
<div className="flex items-center gap-2">
<p className="text-xs text-text/70 w-16 text-right">speed</p>
<div className="w-20 h-2 bg-mid-gray/20 rounded-full overflow-hidden">
<div
className="h-full bg-logo-primary rounded-full transition-all duration-300"
style={{ width: `${model.speed_score * 100}%` }}
/> />
<span className="sr-only">Download size</span>
<span className="font-medium text-text/70">
{formatModelSize(model.size_mb)}
</span>
</div> </div>
</div> </div>
</div> </div>
@ -77,4 +74,20 @@ const ModelCard: React.FC<ModelCardProps> = ({
); );
}; };
const DownloadSize = ({ sizeMb }: { sizeMb: number }) => {
return (
<div className="flex items-center gap-1.5 text-xs text-text/60 tabular-nums">
<Download
aria-hidden="true"
className="h-3.5 w-3.5 text-text/45"
strokeWidth={1.75}
/>
<span className="sr-only">Download size</span>
<span className="font-medium text-text/70">
{formatModelSize(sizeMb)}
</span>
</div>
);
};
export default ModelCard; export default ModelCard;

View file

@ -2,6 +2,7 @@ import React, { useState, useEffect } from "react";
import { invoke } from "@tauri-apps/api/core"; import { invoke } from "@tauri-apps/api/core";
import { ModelInfo } from "../../lib/types"; import { ModelInfo } from "../../lib/types";
import ModelCard from "./ModelCard"; import ModelCard from "./ModelCard";
import HandyTextLogo from "../icons/HandyTextLogo";
interface OnboardingProps { interface OnboardingProps {
onModelSelected: () => void; onModelSelected: () => void;
@ -48,19 +49,23 @@ const Onboarding: React.FC<OnboardingProps> = ({ onModelSelected }) => {
}; };
return ( return (
<div className="max-w-4xl mx-auto text-center space-y-6"> <div className="h-screen w-screen flex flex-col p-6 gap-4 inset-0">
<p className="text-text/70 max-w-md font-medium mx-auto"> <div className="flex flex-col items-center gap-2 shrink-0">
To get started, choose a transcription model <HandyTextLogo width={200} />
</p> <p className="text-text/70 max-w-md font-medium mx-auto">
To get started, choose a transcription model
</p>
</div>
{error && ( <div className="max-w-[600px] w-full mx-auto text-center flex-1 flex flex-col min-h-0">
<div className="bg-red-500/10 border border-red-500/20 rounded-lg p-4"> {error && (
<p className="text-red-400 text-sm">{error}</p> <div className="bg-red-500/10 border border-red-500/20 rounded-lg p-4 mb-4 shrink-0">
</div> <p className="text-red-400 text-sm">{error}</p>
)} </div>
)}
<div className="max-w-2xl mx-auto mt-4"> {/*<div className="flex flex-col gap-4 bg-background-dark p-4 py-5 w-full rounded-2xl flex-1 overflow-y-auto min-h-0">*/}
<div className="grid grid-cols-1 sm:grid-cols-2 gap-4"> <div className="flex flex-col gap-4 ">
{availableModels {availableModels
.filter((model) => getRecommendedBadge(model.id)) .filter((model) => getRecommendedBadge(model.id))
.map((model) => ( .map((model) => (
@ -68,9 +73,7 @@ const Onboarding: React.FC<OnboardingProps> = ({ onModelSelected }) => {
key={model.id} key={model.id}
model={model} model={model}
variant="featured" variant="featured"
badgeText="Recommended"
disabled={downloading} disabled={downloading}
className="sm:col-span-2"
onSelect={handleDownloadModel} onSelect={handleDownloadModel}
/> />
))} ))}

View file

@ -0,0 +1,27 @@
import React from "react";
interface BadgeProps {
children: React.ReactNode;
variant?: "primary";
className?: string;
}
const Badge: React.FC<BadgeProps> = ({
children,
variant = "primary",
className = "",
}) => {
const variantClasses = {
primary: "bg-logo-primary",
};
return (
<span
className={`inline-flex items-center px-3 py-1 rounded-full text-xs font-medium ${variantClasses[variant]} ${className}`}
>
{children}
</span>
);
};
export default Badge;

View file

@ -13,6 +13,8 @@ interface ModelInfo {
is_downloading: boolean; is_downloading: boolean;
partial_size: number; partial_size: number;
is_directory: boolean; is_directory: boolean;
accuracy_score: number;
speed_score: number;
} }
interface DownloadProgress { interface DownloadProgress {

View file

@ -90,6 +90,8 @@ export const ModelInfoSchema = z.object({
is_downloading: z.boolean(), is_downloading: z.boolean(),
partial_size: z.number(), partial_size: z.number(),
is_directory: z.boolean(), is_directory: z.boolean(),
accuracy_score: z.number(),
speed_score: z.number(),
}); });
export type ModelInfo = z.infer<typeof ModelInfoSchema>; export type ModelInfo = z.infer<typeof ModelInfoSchema>;