Replace sharp with wasm-vips for image processing

Fixes #696

- Replaced sharp dependency with wasm-vips (WebAssembly build of libvips)
- Eliminates native build requirements that caused installation failures
- Added vips.ts singleton wrapper for async initialization
- Updated image-resize.ts and image-convert.ts to use wasm-vips API
- Added unit tests for image processing functionality
This commit is contained in:
Mario Zechner 2026-01-13 18:33:27 +01:00
parent 09d409cc92
commit e45fc5f91b
7 changed files with 273 additions and 547 deletions

View file

@ -5,6 +5,7 @@
### Changed
- Light theme colors adjusted for WCAG AA compliance (4.5:1 contrast ratio against white backgrounds)
- Replaced `sharp` with `wasm-vips` for image processing (resize, PNG conversion). Eliminates native build requirements that caused installation failures on some systems. ([#696](https://github.com/badlogic/pi-mono/issues/696))
### Added

View file

@ -51,7 +51,7 @@
"marked": "^15.0.12",
"minimatch": "^10.1.1",
"proper-lockfile": "^4.1.2",
"sharp": "^0.34.2"
"wasm-vips": "^0.0.16"
},
"devDependencies": {
"@types/diff": "^7.0.2",

View file

@ -1,3 +1,5 @@
import { getVips } from "./vips.js";
/**
* Convert image to PNG format for terminal display.
* Kitty graphics protocol requires PNG format (f=100).
@ -11,16 +13,23 @@ export async function convertToPng(
return { data: base64Data, mimeType };
}
const vips = await getVips();
if (!vips) {
// wasm-vips not available
return null;
}
try {
const sharp = (await import("sharp")).default;
const buffer = Buffer.from(base64Data, "base64");
const pngBuffer = await sharp(buffer).png().toBuffer();
const img = vips.Image.newFromBuffer(buffer);
const pngBuffer = img.writeToBuffer(".png");
img.delete();
return {
data: pngBuffer.toString("base64"),
data: Buffer.from(pngBuffer).toString("base64"),
mimeType: "image/png",
};
} catch {
// Sharp not available or conversion failed
// Conversion failed
return null;
}
}

View file

@ -1,4 +1,5 @@
import type { ImageContent } from "@mariozechner/pi-ai";
import { getVips } from "./vips.js";
export interface ImageResizeOptions {
maxWidth?: number; // Default: 2000
@ -29,9 +30,9 @@ const DEFAULT_OPTIONS: Required<ImageResizeOptions> = {
/** Helper to pick the smaller of two buffers */
function pickSmaller(
a: { buffer: Buffer; mimeType: string },
b: { buffer: Buffer; mimeType: string },
): { buffer: Buffer; mimeType: string } {
a: { buffer: Uint8Array; mimeType: string },
b: { buffer: Uint8Array; mimeType: string },
): { buffer: Uint8Array; mimeType: string } {
return a.buffer.length <= b.buffer.length ? a : b;
}
@ -39,7 +40,7 @@ function pickSmaller(
* Resize an image to fit within the specified max dimensions and file size.
* Returns the original image if it already fits within the limits.
*
* Uses sharp for image processing. If sharp is not available (e.g., in some
* Uses wasm-vips for image processing. If wasm-vips is not available (e.g., in some
* environments), returns the original image unchanged.
*
* Strategy for staying under maxBytes:
@ -52,12 +53,29 @@ export async function resizeImage(img: ImageContent, options?: ImageResizeOption
const opts = { ...DEFAULT_OPTIONS, ...options };
const buffer = Buffer.from(img.data, "base64");
let sharp: typeof import("sharp") | undefined;
const vipsOrNull = await getVips();
if (!vipsOrNull) {
// wasm-vips not available - return original image
// We can't get dimensions without vips, so return 0s
return {
data: img.data,
mimeType: img.mimeType,
originalWidth: 0,
originalHeight: 0,
width: 0,
height: 0,
wasResized: false,
};
}
// Capture non-null reference for use in nested functions
const vips = vipsOrNull;
// Load image to get metadata
let sourceImg: InstanceType<typeof vips.Image>;
try {
sharp = (await import("sharp")).default;
sourceImg = vips.Image.newFromBuffer(buffer);
} catch {
// Sharp not available - return original image
// We can't get dimensions without sharp, so return 0s
// Failed to load image
return {
data: img.data,
mimeType: img.mimeType,
@ -69,16 +87,14 @@ export async function resizeImage(img: ImageContent, options?: ImageResizeOption
};
}
const sharpImg = sharp(buffer);
const metadata = await sharpImg.metadata();
const originalWidth = metadata.width ?? 0;
const originalHeight = metadata.height ?? 0;
const format = metadata.format ?? img.mimeType?.split("/")[1] ?? "png";
const originalWidth = sourceImg.width;
const originalHeight = sourceImg.height;
// Check if already within all limits (dimensions AND size)
const originalSize = buffer.length;
if (originalWidth <= opts.maxWidth && originalHeight <= opts.maxHeight && originalSize <= opts.maxBytes) {
sourceImg.delete();
const format = img.mimeType?.split("/")[1] ?? "png";
return {
data: img.data,
mimeType: img.mimeType ?? `image/${format}`,
@ -104,37 +120,45 @@ export async function resizeImage(img: ImageContent, options?: ImageResizeOption
}
// Helper to resize and encode in both formats, returning the smaller one
async function tryBothFormats(
function tryBothFormats(
width: number,
height: number,
jpegQuality: number,
): Promise<{ buffer: Buffer; mimeType: string }> {
const resized = await sharp!(buffer)
.resize(width, height, { fit: "inside", withoutEnlargement: true })
.toBuffer();
): { buffer: Uint8Array; mimeType: string } {
// Load image fresh and resize using scale factor
// (Using newFromBuffer + resize instead of thumbnailBuffer to avoid lazy re-read issues)
const img = vips.Image.newFromBuffer(buffer);
const scale = Math.min(width / img.width, height / img.height);
const resized = scale < 1 ? img.resize(scale) : img;
const [pngBuffer, jpegBuffer] = await Promise.all([
sharp!(resized).png({ compressionLevel: 9 }).toBuffer(),
sharp!(resized).jpeg({ quality: jpegQuality }).toBuffer(),
]);
const pngBuffer = resized.writeToBuffer(".png");
const jpegBuffer = resized.writeToBuffer(".jpg", { Q: jpegQuality });
if (resized !== img) {
resized.delete();
}
img.delete();
return pickSmaller({ buffer: pngBuffer, mimeType: "image/png" }, { buffer: jpegBuffer, mimeType: "image/jpeg" });
}
// Clean up the source image
sourceImg.delete();
// Try to produce an image under maxBytes
const qualitySteps = [85, 70, 55, 40];
const scaleSteps = [1.0, 0.75, 0.5, 0.35, 0.25];
let best: { buffer: Buffer; mimeType: string };
let best: { buffer: Uint8Array; mimeType: string };
let finalWidth = targetWidth;
let finalHeight = targetHeight;
// First attempt: resize to target dimensions, try both formats
best = await tryBothFormats(targetWidth, targetHeight, opts.jpegQuality);
best = tryBothFormats(targetWidth, targetHeight, opts.jpegQuality);
if (best.buffer.length <= opts.maxBytes) {
return {
data: best.buffer.toString("base64"),
data: Buffer.from(best.buffer).toString("base64"),
mimeType: best.mimeType,
originalWidth,
originalHeight,
@ -146,11 +170,11 @@ export async function resizeImage(img: ImageContent, options?: ImageResizeOption
// Still too large - try JPEG with decreasing quality (and compare to PNG each time)
for (const quality of qualitySteps) {
best = await tryBothFormats(targetWidth, targetHeight, quality);
best = tryBothFormats(targetWidth, targetHeight, quality);
if (best.buffer.length <= opts.maxBytes) {
return {
data: best.buffer.toString("base64"),
data: Buffer.from(best.buffer).toString("base64"),
mimeType: best.mimeType,
originalWidth,
originalHeight,
@ -172,11 +196,11 @@ export async function resizeImage(img: ImageContent, options?: ImageResizeOption
}
for (const quality of qualitySteps) {
best = await tryBothFormats(finalWidth, finalHeight, quality);
best = tryBothFormats(finalWidth, finalHeight, quality);
if (best.buffer.length <= opts.maxBytes) {
return {
data: best.buffer.toString("base64"),
data: Buffer.from(best.buffer).toString("base64"),
mimeType: best.mimeType,
originalWidth,
originalHeight,
@ -191,7 +215,7 @@ export async function resizeImage(img: ImageContent, options?: ImageResizeOption
// Last resort: return smallest version we produced even if over limit
// (the API will reject it, but at least we tried everything)
return {
data: best.buffer.toString("base64"),
data: Buffer.from(best.buffer).toString("base64"),
mimeType: best.mimeType,
originalWidth,
originalHeight,

View file

@ -0,0 +1,40 @@
/**
* Singleton wrapper for wasm-vips initialization.
* wasm-vips requires async initialization, so we cache the instance.
*/
import type Vips from "wasm-vips";
let vipsInstance: Awaited<ReturnType<typeof Vips>> | null = null;
let vipsInitPromise: Promise<Awaited<ReturnType<typeof Vips>> | null> | null = null;
/**
* Get the initialized wasm-vips instance.
* Returns null if wasm-vips is not available or fails to initialize.
*/
export async function getVips(): Promise<Awaited<ReturnType<typeof Vips>> | null> {
if (vipsInstance) {
return vipsInstance;
}
if (vipsInitPromise) {
return vipsInitPromise;
}
vipsInitPromise = (async () => {
try {
const VipsInit = (await import("wasm-vips")).default;
vipsInstance = await VipsInit();
return vipsInstance;
} catch {
// wasm-vips not available
return null;
}
})();
const result = await vipsInitPromise;
if (!result) {
vipsInitPromise = null; // Allow retry on failure
}
return result;
}

View file

@ -0,0 +1,144 @@
/**
* Tests for image processing utilities using wasm-vips.
*/
import { describe, expect, it } from "vitest";
import { convertToPng } from "../src/utils/image-convert.js";
import { formatDimensionNote, resizeImage } from "../src/utils/image-resize.js";
import { getVips } from "../src/utils/vips.js";
// Small 2x2 red PNG image (base64)
const TINY_PNG = "iVBORw0KGgoAAAANSUhEUgAAAAIAAAACCAIAAAD91JpzAAAADklEQVQI12P4z8DAwMAAAA0BA/m5sb9AAAAAAElFTkSuQmCC";
// Small 2x2 blue JPEG image (base64)
const TINY_JPEG =
"/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRofHh0aHBwgJC4nICIsIxwcKDcpLDAxNDQ0Hyc5PTgyPC4zNDL/2wBDAQkJCQwLDBgNDRgyIRwhMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjL/wAARCAACAAIDASIAAhEBAxEB/8QAFQABAQAAAAAAAAAAAAAAAAAAAAn/xAAUEAEAAAAAAAAAAAAAAAAAAAAA/8QAFQEBAQAAAAAAAAAAAAAAAAAAAAX/xAAUEQEAAAAAAAAAAAAAAAAAAAAA/9oADAMBAAIRAxEAPwCwAB//2Q==";
// 100x100 gray PNG (generated with wasm-vips)
const MEDIUM_PNG_100x100 =
"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";
// 200x200 colored PNG (generated with wasm-vips)
const LARGE_PNG_200x200 =
"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";
describe("wasm-vips initialization", () => {
it("should initialize wasm-vips successfully", async () => {
const vips = await getVips();
expect(vips).not.toBeNull();
});
it("should return cached instance on subsequent calls", async () => {
const vips1 = await getVips();
const vips2 = await getVips();
expect(vips1).toBe(vips2);
});
});
describe("convertToPng", () => {
it("should return original data for PNG input", async () => {
const result = await convertToPng(TINY_PNG, "image/png");
expect(result).not.toBeNull();
expect(result!.data).toBe(TINY_PNG);
expect(result!.mimeType).toBe("image/png");
});
it("should convert JPEG to PNG", async () => {
const result = await convertToPng(TINY_JPEG, "image/jpeg");
expect(result).not.toBeNull();
expect(result!.mimeType).toBe("image/png");
// Result should be valid base64
expect(() => Buffer.from(result!.data, "base64")).not.toThrow();
// PNG magic bytes
const buffer = Buffer.from(result!.data, "base64");
expect(buffer[0]).toBe(0x89);
expect(buffer[1]).toBe(0x50); // 'P'
expect(buffer[2]).toBe(0x4e); // 'N'
expect(buffer[3]).toBe(0x47); // 'G'
});
});
describe("resizeImage", () => {
it("should return original image if within limits", async () => {
const result = await resizeImage(
{ type: "image", data: TINY_PNG, mimeType: "image/png" },
{ maxWidth: 100, maxHeight: 100, maxBytes: 1024 * 1024 },
);
expect(result.wasResized).toBe(false);
expect(result.data).toBe(TINY_PNG);
expect(result.originalWidth).toBe(2);
expect(result.originalHeight).toBe(2);
expect(result.width).toBe(2);
expect(result.height).toBe(2);
});
it("should resize image exceeding dimension limits", async () => {
const result = await resizeImage(
{ type: "image", data: MEDIUM_PNG_100x100, mimeType: "image/png" },
{ maxWidth: 50, maxHeight: 50, maxBytes: 1024 * 1024 },
);
expect(result.wasResized).toBe(true);
expect(result.originalWidth).toBe(100);
expect(result.originalHeight).toBe(100);
expect(result.width).toBeLessThanOrEqual(50);
expect(result.height).toBeLessThanOrEqual(50);
});
it("should resize image exceeding byte limit", async () => {
const originalBuffer = Buffer.from(LARGE_PNG_200x200, "base64");
const originalSize = originalBuffer.length;
// Set maxBytes to less than the original image size
const result = await resizeImage(
{ type: "image", data: LARGE_PNG_200x200, mimeType: "image/png" },
{ maxWidth: 2000, maxHeight: 2000, maxBytes: Math.floor(originalSize / 2) },
);
// Should have tried to reduce size
const resultBuffer = Buffer.from(result.data, "base64");
expect(resultBuffer.length).toBeLessThan(originalSize);
});
it("should handle JPEG input", async () => {
const result = await resizeImage(
{ type: "image", data: TINY_JPEG, mimeType: "image/jpeg" },
{ maxWidth: 100, maxHeight: 100, maxBytes: 1024 * 1024 },
);
expect(result.wasResized).toBe(false);
expect(result.originalWidth).toBe(2);
expect(result.originalHeight).toBe(2);
});
});
describe("formatDimensionNote", () => {
it("should return undefined for non-resized images", () => {
const note = formatDimensionNote({
data: "",
mimeType: "image/png",
originalWidth: 100,
originalHeight: 100,
width: 100,
height: 100,
wasResized: false,
});
expect(note).toBeUndefined();
});
it("should return formatted note for resized images", () => {
const note = formatDimensionNote({
data: "",
mimeType: "image/png",
originalWidth: 2000,
originalHeight: 1000,
width: 1000,
height: 500,
wasResized: true,
});
expect(note).toContain("original 2000x1000");
expect(note).toContain("displayed at 1000x500");
expect(note).toContain("2.00"); // scale factor
});
});