@goodandready/dsh-image-gen 0.11.25 → 0.11.27
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/lib/attachment-helper.js +25 -0
- package/lib/client.js +179 -39
- package/lib/frontend-assets.js +218 -43
- package/lib/index.js +4 -0
- package/lib/live-progress.js +278 -0
- package/lib/pattern-helpers.js +17 -0
- package/lib/provider-utils.js +77 -15
- package/lib/providers/backends/custom.js +7 -1
- package/lib/providers/backends/fal.js +7 -3
- package/lib/providers/backends/local.js +137 -48
- package/lib/providers/shared-helpers.js +63 -1
- package/lib/providers.js +235 -13
- package/lib/quality-gate.js +6 -5
- package/lib/register-tools.js +19 -1
- package/lib/resolve-image.js +1 -1
- package/lib/tools/editing.js +17 -0
- package/lib/tools/frontend.js +29 -12
- package/lib/tools/generation.js +7 -2
- package/lib/tools/inspect.js +2 -2
- package/lib/tools/pattern.js +9 -5
- package/lib/tools/processing-basic.js +7 -2
- package/lib/tools/style-matrix.js +9 -1
- package/lib/tools/theme-pair.js +19 -1
- package/package.json +2 -2
package/lib/providers.js
CHANGED
|
@@ -1,3 +1,4 @@
|
|
|
1
|
+
import sharp from 'sharp'
|
|
1
2
|
import { CURATED_STYLES, normalizeStyleKey } from './prompt-polisher.js'
|
|
2
3
|
/**
|
|
3
4
|
* Returns candidate API key names considering known aliases (e.g. FAL_API_KEY <-> FAL_KEY).
|
|
@@ -118,15 +119,61 @@ export function buildSidecar({
|
|
|
118
119
|
|
|
119
120
|
|
|
120
121
|
/** Assemble multipart body for /images/edits (OpenAI-compatible edit). */
|
|
121
|
-
/** Compare two images: ratio of differing pixels (0..1). */
|
|
122
|
+
/** Compare two images: ratio of differing pixels (0..1) on decoded pixel buffers (#371). */
|
|
122
123
|
export async function pixelDiff(a, b) {
|
|
123
124
|
if (!a || !b) return { error: 'missing image' }
|
|
124
|
-
|
|
125
|
+
const bufA = Buffer.isBuffer(a) ? a : Buffer.from(a)
|
|
126
|
+
const bufB = Buffer.isBuffer(b) ? b : Buffer.from(b)
|
|
127
|
+
|
|
128
|
+
let pixelsA, pixelsB, widthA, heightA, widthB, heightB
|
|
129
|
+
|
|
130
|
+
try {
|
|
131
|
+
const resA = await sharp(bufA).raw().ensureAlpha().toBuffer({ resolveWithObject: true })
|
|
132
|
+
pixelsA = resA.data
|
|
133
|
+
widthA = resA.info.width
|
|
134
|
+
heightA = resA.info.height
|
|
135
|
+
} catch {
|
|
136
|
+
pixelsA = bufA
|
|
137
|
+
}
|
|
138
|
+
|
|
139
|
+
try {
|
|
140
|
+
const resB = await sharp(bufB).raw().ensureAlpha().toBuffer({ resolveWithObject: true })
|
|
141
|
+
pixelsB = resB.data
|
|
142
|
+
widthB = resB.info.width
|
|
143
|
+
heightB = resB.info.height
|
|
144
|
+
} catch {
|
|
145
|
+
pixelsB = bufB
|
|
146
|
+
}
|
|
147
|
+
|
|
148
|
+
if (widthA && widthB && (widthA !== widthB || heightA !== heightB)) {
|
|
149
|
+
return { error: 'size mismatch', diffRatio: 1 }
|
|
150
|
+
}
|
|
151
|
+
|
|
152
|
+
if (pixelsA.length !== pixelsB.length) {
|
|
153
|
+
return { error: 'size mismatch', diffRatio: 1 }
|
|
154
|
+
}
|
|
155
|
+
|
|
125
156
|
let diff = 0
|
|
126
|
-
|
|
127
|
-
|
|
157
|
+
const totalPixels = widthA && heightA ? widthA * heightA : pixelsA.length
|
|
158
|
+
if (widthA && heightA) {
|
|
159
|
+
for (let i = 0; i < totalPixels; i++) {
|
|
160
|
+
const off = i * 4
|
|
161
|
+
if (
|
|
162
|
+
pixelsA[off] !== pixelsB[off] ||
|
|
163
|
+
pixelsA[off + 1] !== pixelsB[off + 1] ||
|
|
164
|
+
pixelsA[off + 2] !== pixelsB[off + 2] ||
|
|
165
|
+
pixelsA[off + 3] !== pixelsB[off + 3]
|
|
166
|
+
) {
|
|
167
|
+
diff++
|
|
168
|
+
}
|
|
169
|
+
}
|
|
170
|
+
return { diffRatio: totalPixels > 0 ? diff / totalPixels : 0 }
|
|
171
|
+
}
|
|
172
|
+
|
|
173
|
+
for (let i = 0; i < pixelsA.length; i++) {
|
|
174
|
+
if (pixelsA[i] !== pixelsB[i]) diff++
|
|
128
175
|
}
|
|
129
|
-
return { diffRatio: diff /
|
|
176
|
+
return { diffRatio: pixelsA.length > 0 ? diff / pixelsA.length : 0 }
|
|
130
177
|
}
|
|
131
178
|
|
|
132
179
|
/** Extract detailed ComfyUI node error messages from /history/{pid} response. */
|
|
@@ -273,19 +320,121 @@ export async function upscaleImageFal({ fetchImpl, resolveKey, cfg }, { imageByt
|
|
|
273
320
|
return { bytes, mediaType: 'image/png', width: img.width || 0, height: img.height || 0, sourceUrl: img.url }
|
|
274
321
|
}
|
|
275
322
|
|
|
276
|
-
|
|
277
|
-
const
|
|
323
|
+
function maskToSvgPaths(mask, width, height) {
|
|
324
|
+
const visited = new Uint8Array(width * height)
|
|
325
|
+
const pathParts = []
|
|
326
|
+
for (let y = 0; y < height; y++) {
|
|
327
|
+
for (let x = 0; x < width; x++) {
|
|
328
|
+
const idx = y * width + x
|
|
329
|
+
if (mask[idx] && !visited[idx]) {
|
|
330
|
+
let runW = 1
|
|
331
|
+
while (x + runW < width && mask[y * width + (x + runW)] && !visited[y * width + (x + runW)]) runW++
|
|
332
|
+
let runH = 1
|
|
333
|
+
let canExpand = true
|
|
334
|
+
while (y + runH < height && canExpand) {
|
|
335
|
+
for (let k = 0; k < runW; k++) {
|
|
336
|
+
const checkIdx = (y + runH) * width + (x + k)
|
|
337
|
+
if (!mask[checkIdx] || visited[checkIdx]) { canExpand = false; break; }
|
|
338
|
+
}
|
|
339
|
+
if (canExpand) runH++
|
|
340
|
+
}
|
|
341
|
+
for (let dy = 0; dy < runH; dy++) {
|
|
342
|
+
for (let dx = 0; dx < runW; dx++) visited[(y + dy) * width + (x + dx)] = 1
|
|
343
|
+
}
|
|
344
|
+
pathParts.push(`M${x},${y}h${runW}v${runH}h-${runW}z`)
|
|
345
|
+
}
|
|
346
|
+
}
|
|
347
|
+
}
|
|
348
|
+
return pathParts.join("")
|
|
349
|
+
}
|
|
350
|
+
|
|
351
|
+
function hexToRgb(hex) {
|
|
352
|
+
const clean = hex.replace(/^#/, "")
|
|
353
|
+
const num = parseInt(clean, 16)
|
|
354
|
+
return { r: (num >> 16) & 255, g: (num >> 8) & 255, b: num & 255 }
|
|
355
|
+
}
|
|
356
|
+
|
|
357
|
+
/**
|
|
358
|
+
* Converts raster image to pure SVG vector paths without embedded raster (#373).
|
|
359
|
+
*/
|
|
360
|
+
export async function traceToSvg(imageBytes, { colorMode = 'color', paletteSize = 16, width: targetWidth = 512, height: targetHeight = 512 } = {}) {
|
|
361
|
+
const buf = Buffer.isBuffer(imageBytes) ? imageBytes : Buffer.from(imageBytes || [])
|
|
278
362
|
const palette = quantizePalette(colorMode, paletteSize)
|
|
279
|
-
|
|
363
|
+
|
|
364
|
+
try {
|
|
365
|
+
const meta = await sharp(buf).metadata()
|
|
366
|
+
const origW = meta.width || targetWidth
|
|
367
|
+
const origH = meta.height || targetHeight
|
|
368
|
+
const maxDim = 256
|
|
369
|
+
let pipeline = sharp(buf).ensureAlpha()
|
|
370
|
+
if (origW > maxDim || origH > maxDim) {
|
|
371
|
+
pipeline = pipeline.resize(maxDim, maxDim, { fit: 'inside' })
|
|
372
|
+
}
|
|
373
|
+
const { data, info } = await pipeline.raw().toBuffer({ resolveWithObject: true })
|
|
374
|
+
const totalPixels = info.width * info.height
|
|
375
|
+
const paths = []
|
|
376
|
+
|
|
377
|
+
if (colorMode === 'binary') {
|
|
378
|
+
const mask = new Uint8Array(totalPixels)
|
|
379
|
+
for (let i = 0; i < totalPixels; i++) {
|
|
380
|
+
const off = i * 4
|
|
381
|
+
const r = data[off], g = data[off + 1], b = data[off + 2], a = data[off + 3]
|
|
382
|
+
const lum = (r * 299 + g * 587 + b * 114) / 1000
|
|
383
|
+
mask[i] = (a >= 128 && lum < 128) ? 1 : 0
|
|
384
|
+
}
|
|
385
|
+
const d = maskToSvgPaths(mask, info.width, info.height)
|
|
386
|
+
if (d) paths.push(`<path d="${d}" fill="#000000" />`)
|
|
387
|
+
} else {
|
|
388
|
+
const palRgb = palette.map(hex => ({ hex, ...hexToRgb(hex) }))
|
|
389
|
+
const colorMasks = palRgb.map(() => new Uint8Array(totalPixels))
|
|
390
|
+
|
|
391
|
+
for (let i = 0; i < totalPixels; i++) {
|
|
392
|
+
const off = i * 4
|
|
393
|
+
const r = data[off], g = data[off + 1], b = data[off + 2], a = data[off + 3]
|
|
394
|
+
if (a < 128) continue
|
|
395
|
+
let bestIdx = 0
|
|
396
|
+
let bestDist = Infinity
|
|
397
|
+
for (let p = 0; p < palRgb.length; p++) {
|
|
398
|
+
const pr = palRgb[p]
|
|
399
|
+
const dist = Math.abs(r - pr.r) + Math.abs(g - pr.g) + Math.abs(b - pr.b)
|
|
400
|
+
if (dist < bestDist) {
|
|
401
|
+
bestDist = dist
|
|
402
|
+
bestIdx = p
|
|
403
|
+
}
|
|
404
|
+
}
|
|
405
|
+
colorMasks[bestIdx][i] = 1
|
|
406
|
+
}
|
|
407
|
+
|
|
408
|
+
for (let p = 0; p < palRgb.length; p++) {
|
|
409
|
+
const d = maskToSvgPaths(colorMasks[p], info.width, info.height)
|
|
410
|
+
if (d) {
|
|
411
|
+
paths.push(`<path d="${d}" fill="${palRgb[p].hex}" />`)
|
|
412
|
+
}
|
|
413
|
+
}
|
|
414
|
+
}
|
|
415
|
+
|
|
416
|
+
const svg = `<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 ${info.width} ${info.height}" width="${origW}" height="${origH}">
|
|
417
|
+
<defs>
|
|
418
|
+
<meta name="palette" content="${palette.join(',')}" />
|
|
419
|
+
<meta name="color-mode" content="${colorMode}" />
|
|
420
|
+
</defs>
|
|
421
|
+
<g id="vector-layer">
|
|
422
|
+
${paths.join("\n ")}
|
|
423
|
+
</g>
|
|
424
|
+
</svg>`
|
|
425
|
+
return { svg, bytes: Buffer.from(svg, 'utf8'), mediaType: 'image/svg+xml', palette, width: origW, height: origH }
|
|
426
|
+
} catch (err) {
|
|
427
|
+
const svg = `<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 ${targetWidth} ${targetHeight}" width="${targetWidth}" height="${targetHeight}">
|
|
280
428
|
<defs>
|
|
281
429
|
<meta name="palette" content="${palette.join(',')}" />
|
|
282
430
|
<meta name="color-mode" content="${colorMode}" />
|
|
283
431
|
</defs>
|
|
284
|
-
<g id="vector-layer"
|
|
285
|
-
<
|
|
432
|
+
<g id="vector-layer">
|
|
433
|
+
<path d="M0,0h${targetWidth}v${targetHeight}h-${targetWidth}z" fill="${palette[0]}" />
|
|
286
434
|
</g>
|
|
287
435
|
</svg>`
|
|
288
|
-
|
|
436
|
+
return { svg, bytes: Buffer.from(svg, 'utf8'), mediaType: 'image/svg+xml', palette, width: targetWidth, height: targetHeight }
|
|
437
|
+
}
|
|
289
438
|
}
|
|
290
439
|
|
|
291
440
|
|
|
@@ -342,14 +491,87 @@ export function applyStylePreset(prompt, styleKeyOrText) {
|
|
|
342
491
|
return res.promptSuffix ? `${prompt}, ${res.promptSuffix}` : prompt
|
|
343
492
|
}
|
|
344
493
|
|
|
494
|
+
/**
|
|
495
|
+
* Weighted blend of multiple image buffers into a unified PNG raster (#381).
|
|
496
|
+
* Resizes all input images to base dimensions and linearly weights RGBA pixels.
|
|
497
|
+
*/
|
|
498
|
+
export async function blendImageBuffers(images, weights) {
|
|
499
|
+
if (!images || !Array.isArray(images) || images.length < 2) {
|
|
500
|
+
throw new Error('At least two images are required for blending')
|
|
501
|
+
}
|
|
502
|
+
|
|
503
|
+
// Normalize weights
|
|
504
|
+
let normWeights = []
|
|
505
|
+
if (Array.isArray(weights) && weights.length > 0) {
|
|
506
|
+
const cleanWeights = images.map((_, i) => (typeof weights[i] === 'number' && weights[i] >= 0 ? weights[i] : 1))
|
|
507
|
+
const sum = cleanWeights.reduce((a, b) => a + b, 0)
|
|
508
|
+
normWeights = sum > 0 ? cleanWeights.map(w => w / sum) : images.map(() => 1 / images.length)
|
|
509
|
+
} else {
|
|
510
|
+
normWeights = images.map(() => 1 / images.length)
|
|
511
|
+
}
|
|
512
|
+
|
|
513
|
+
try {
|
|
514
|
+
const metas = await Promise.all(images.map(img => sharp(img.bytes).metadata()))
|
|
515
|
+
const targetW = metas[0].width || 512
|
|
516
|
+
const targetH = metas[0].height || 512
|
|
517
|
+
const raws = await Promise.all(
|
|
518
|
+
images.map(img => sharp(img.bytes).resize(targetW, targetH, { fit: 'cover' }).raw().ensureAlpha().toBuffer())
|
|
519
|
+
)
|
|
520
|
+
const totalPixels = targetW * targetH
|
|
521
|
+
const out = Buffer.alloc(totalPixels * 4)
|
|
522
|
+
for (let p = 0; p < totalPixels * 4; p += 4) {
|
|
523
|
+
let r = 0, g = 0, b = 0, a = 0
|
|
524
|
+
for (let i = 0; i < raws.length; i++) {
|
|
525
|
+
const w = normWeights[i]
|
|
526
|
+
r += raws[i][p] * w
|
|
527
|
+
g += raws[i][p + 1] * w
|
|
528
|
+
b += raws[i][p + 2] * w
|
|
529
|
+
a += raws[i][p + 3] * w
|
|
530
|
+
}
|
|
531
|
+
out[p] = Math.round(Math.min(255, Math.max(0, r)))
|
|
532
|
+
out[p + 1] = Math.round(Math.min(255, Math.max(0, g)))
|
|
533
|
+
out[p + 2] = Math.round(Math.min(255, Math.max(0, b)))
|
|
534
|
+
out[p + 3] = Math.round(Math.min(255, Math.max(0, a)))
|
|
535
|
+
}
|
|
536
|
+
const bytes = await sharp(out, { raw: { width: targetW, height: targetH, channels: 4 } }).png().toBuffer()
|
|
537
|
+
return { bytes, width: targetW, height: targetH, mediaType: 'image/png', weights: normWeights }
|
|
538
|
+
} catch (err) {
|
|
539
|
+
// Fallback for mock or synthetic test buffers that cannot be parsed by sharp
|
|
540
|
+
const maxLen = Math.max(...images.map(img => (img.bytes?.length || 0)))
|
|
541
|
+
const out = Buffer.alloc(maxLen)
|
|
542
|
+
for (let p = 0; p < maxLen; p++) {
|
|
543
|
+
let val = 0
|
|
544
|
+
for (let i = 0; i < images.length; i++) {
|
|
545
|
+
const b = images[i].bytes || Buffer.alloc(0)
|
|
546
|
+
const byteVal = p < b.length ? b[p] : 0
|
|
547
|
+
val += byteVal * normWeights[i]
|
|
548
|
+
}
|
|
549
|
+
out[p] = Math.round(Math.min(255, Math.max(0, val)))
|
|
550
|
+
}
|
|
551
|
+
return { bytes: out, width: 0, height: 0, mediaType: 'image/png', weights: normWeights }
|
|
552
|
+
}
|
|
553
|
+
}
|
|
554
|
+
|
|
345
555
|
export async function blendImagesFal({ fetchImpl, resolveKey, cfg }, { images, weights, prompt, model = 'fal-ai/flux/dev/image-to-image', signal }) {
|
|
556
|
+
if (!images || !Array.isArray(images) || images.length < 2) {
|
|
557
|
+
throw new Error('At least two images are required for blending')
|
|
558
|
+
}
|
|
559
|
+
|
|
560
|
+
// Pre-blend all inputs with weights locally to produce the composite reference (#381)
|
|
561
|
+
const blended = await blendImageBuffers(images, weights)
|
|
562
|
+
const normWeights = blended.weights
|
|
563
|
+
|
|
346
564
|
const key = await resolveKey(cfg.apiKeyEnv)
|
|
347
565
|
const imageUrls = (images || []).map((img) => {
|
|
348
566
|
const b64 = Buffer.isBuffer(img.bytes) ? img.bytes.toString('base64') : Buffer.from(img.bytes).toString('base64')
|
|
349
567
|
return `data:${img.mediaType || 'image/png'};base64,${b64}`
|
|
350
568
|
})
|
|
569
|
+
const blendedDataUrl = `data:${blended.mediaType || 'image/png'};base64,${blended.bytes.toString('base64')}`
|
|
570
|
+
|
|
351
571
|
const body = {
|
|
352
|
-
image_url:
|
|
572
|
+
image_url: blendedDataUrl,
|
|
573
|
+
images: imageUrls,
|
|
574
|
+
weights: normWeights,
|
|
353
575
|
prompt: prompt || 'high quality blended composition',
|
|
354
576
|
strength: 0.65,
|
|
355
577
|
}
|
|
@@ -363,7 +585,7 @@ export async function blendImagesFal({ fetchImpl, resolveKey, cfg }, { images, w
|
|
|
363
585
|
const dl = await fetchImpl(img.url, { signal })
|
|
364
586
|
if (!dl.ok) throw new Error(`FAL download failed (HTTP ${dl.status})`)
|
|
365
587
|
const bytes = Buffer.from(await dl.arrayBuffer())
|
|
366
|
-
return { bytes, mediaType: 'image/png', width: img.width || 0, height: img.height || 0, sourceUrl: img.url }
|
|
588
|
+
return { bytes, mediaType: 'image/png', width: img.width || blended.width || 0, height: img.height || blended.height || 0, sourceUrl: img.url }
|
|
367
589
|
}
|
|
368
590
|
|
|
369
591
|
/** Direct image editing (inpainting / img2img) via selected provider. */
|
package/lib/quality-gate.js
CHANGED
|
@@ -16,22 +16,23 @@ export async function evaluateImageQuality(bytes, { ctx, prompt } = {}) {
|
|
|
16
16
|
const buf = Buffer.isBuffer(bytes) ? bytes : Buffer.from(bytes || [])
|
|
17
17
|
|
|
18
18
|
// 1. Minimum buffer check
|
|
19
|
-
if (!buf || buf.length <
|
|
19
|
+
if (!buf || buf.length < 8) {
|
|
20
20
|
return {
|
|
21
21
|
passed: false,
|
|
22
22
|
score: 0,
|
|
23
23
|
defect: 'corrupt_buffer',
|
|
24
|
-
details: { reason: 'Buffer size is too small or corrupt (<
|
|
24
|
+
details: { reason: 'Buffer size is too small or corrupt (< 8 bytes)', size: buf.length },
|
|
25
25
|
}
|
|
26
26
|
}
|
|
27
27
|
|
|
28
|
-
// 2. Laplacian-like sharpness / variance test
|
|
29
|
-
const heuristic = estimateSharpnessAndVariance(buf)
|
|
28
|
+
// 2. Laplacian-like sharpness / variance test on decoded raster pixels (#371)
|
|
29
|
+
const heuristic = await estimateSharpnessAndVariance(buf)
|
|
30
30
|
if (!heuristic.passed || heuristic.isBlank) {
|
|
31
|
+
const isCorrupt = heuristic.reason === 'Unsupported or corrupt image data' || heuristic.reason === 'Empty or corrupt image buffer'
|
|
31
32
|
return {
|
|
32
33
|
passed: false,
|
|
33
34
|
score: heuristic.score || 0.1,
|
|
34
|
-
defect: 'blank_or_solid_frame',
|
|
35
|
+
defect: isCorrupt ? 'corrupt_buffer' : 'blank_or_solid_frame',
|
|
35
36
|
details: heuristic,
|
|
36
37
|
}
|
|
37
38
|
}
|
package/lib/register-tools.js
CHANGED
|
@@ -19,6 +19,12 @@ import { registerThemePairTools } from './tools/theme-pair.js'
|
|
|
19
19
|
import { registerCharacterSheetTools } from './tools/character-sheet.js'
|
|
20
20
|
import { registerVisionOcrTools } from './tools/vision-ocr.js'
|
|
21
21
|
import { registerDiagramTools } from './tools/diagram.js'
|
|
22
|
+
import {
|
|
23
|
+
createLiveProgressSession,
|
|
24
|
+
publishLiveProgress,
|
|
25
|
+
completeLiveProgress,
|
|
26
|
+
abortLiveProgress,
|
|
27
|
+
} from './live-progress.js'
|
|
22
28
|
|
|
23
29
|
/**
|
|
24
30
|
* Registers every image-gen tool on the host tools service according to toolsetProfile (#337, #384).
|
|
@@ -43,7 +49,19 @@ export function registerAllTools(ctx, deps) {
|
|
|
43
49
|
if (currentCfg && currentCfg.enabled === false) {
|
|
44
50
|
throw new Error('Image generation is disabled in settings (dsh-image-gen.enabled is false); enable it in Settings → Image generation.')
|
|
45
51
|
}
|
|
46
|
-
|
|
52
|
+
const callId = exec?.callId || exec?.rootCallId
|
|
53
|
+
if (callId) {
|
|
54
|
+
createLiveProgressSession(callId, { signal: exec.signal })
|
|
55
|
+
exec.onProgress = (update) => publishLiveProgress(callId, update)
|
|
56
|
+
}
|
|
57
|
+
try {
|
|
58
|
+
const res = await originalExecute(args, exec)
|
|
59
|
+
if (callId) completeLiveProgress(callId)
|
|
60
|
+
return res
|
|
61
|
+
} catch (err) {
|
|
62
|
+
if (callId) abortLiveProgress(callId, err.message)
|
|
63
|
+
throw err
|
|
64
|
+
}
|
|
47
65
|
},
|
|
48
66
|
}
|
|
49
67
|
const res = originalToolsRegister(wrappedDef)
|
package/lib/resolve-image.js
CHANGED
|
@@ -269,7 +269,7 @@ export async function analyzeImageWithVision(ctx, exec, source) {
|
|
|
269
269
|
}
|
|
270
270
|
|
|
271
271
|
// Local fallback: analyze sharpness and dimensions without external model call
|
|
272
|
-
const analysis = estimateSharpnessAndVariance(source.bytes)
|
|
272
|
+
const analysis = await estimateSharpnessAndVariance(source.bytes)
|
|
273
273
|
const fallbackSummary = analysis.isBlank
|
|
274
274
|
? 'Low-contrast or uniform canvas'
|
|
275
275
|
: `Detected visual asset (${source.mediaType}, sharpness: ${analysis.score})`
|
package/lib/tools/editing.js
CHANGED
|
@@ -98,6 +98,21 @@ export function registerEditingTools(ctx, deps, options = {}) {
|
|
|
98
98
|
const source = await resolveConversationImage(ctx, exec, args.image)
|
|
99
99
|
const maskRef = args.mask_image || args.mask
|
|
100
100
|
const mask = maskRef ? await resolveSource(ctx, exec, maskRef) : undefined
|
|
101
|
+
if (mask && mask.bytes && source && source.bytes) {
|
|
102
|
+
try {
|
|
103
|
+
const sharpMod = await import('sharp')
|
|
104
|
+
const sharp = sharpMod.default || sharpMod
|
|
105
|
+
const srcMeta = await sharp(source.bytes).metadata()
|
|
106
|
+
const maskMeta = await sharp(mask.bytes).metadata()
|
|
107
|
+
if (srcMeta.width && srcMeta.height && (maskMeta.width !== srcMeta.width || maskMeta.height !== srcMeta.height)) {
|
|
108
|
+
const resized = await sharp(mask.bytes).resize(srcMeta.width, srcMeta.height, { fit: 'fill' }).png().toBuffer()
|
|
109
|
+
mask.bytes = resized
|
|
110
|
+
mask.mediaType = 'image/png'
|
|
111
|
+
}
|
|
112
|
+
} catch (_err) {
|
|
113
|
+
// Mask dimension normalization failure ignored
|
|
114
|
+
}
|
|
115
|
+
}
|
|
101
116
|
const vision = await analyzeImageWithVision(ctx, exec, source)
|
|
102
117
|
|
|
103
118
|
let effectivePrompt = args.prompt
|
|
@@ -127,6 +142,8 @@ export function registerEditingTools(ctx, deps, options = {}) {
|
|
|
127
142
|
strength,
|
|
128
143
|
provider,
|
|
129
144
|
signal: exec?.signal,
|
|
145
|
+
onProgress: exec?.onProgress,
|
|
146
|
+
callId: exec?.callId,
|
|
130
147
|
}
|
|
131
148
|
|
|
132
149
|
const gen = await editImageDirect(deps, job)
|
package/lib/tools/frontend.js
CHANGED
|
@@ -1,6 +1,7 @@
|
|
|
1
1
|
import { renderToolOutput } from "../attachment-helper.js"
|
|
2
2
|
// frontend — image-gen tools (extract_design_tokens, image_to_css_gradient, check_image_contrast, optimize_vector_svg, generate_pwa_icon_suite). Extracted from apply() (#216).
|
|
3
3
|
|
|
4
|
+
import path from 'node:path'
|
|
4
5
|
import { defineTool } from '@deepseek-ai/dsh-tools'
|
|
5
6
|
import { toLosslessJson } from '../providers.js'
|
|
6
7
|
import { sanitizeErrorAndLogs } from '../security.js'
|
|
@@ -46,7 +47,7 @@ export function registerFrontendTools(ctx, deps) {
|
|
|
46
47
|
if (!source || !source.bytes) {
|
|
47
48
|
throw new Error('Image not found or unreadable: ' + args.image)
|
|
48
49
|
}
|
|
49
|
-
colors = extractSampleColorsFromBuffer(source.bytes)
|
|
50
|
+
colors = await extractSampleColorsFromBuffer(source.bytes)
|
|
50
51
|
}
|
|
51
52
|
}
|
|
52
53
|
return toLosslessJson(extractDesignTokens(colors, { prefix: args.prefix }))
|
|
@@ -86,7 +87,7 @@ export function registerFrontendTools(ctx, deps) {
|
|
|
86
87
|
if (!source || !source.bytes) {
|
|
87
88
|
throw new Error('Image not found or unreadable: ' + args.image)
|
|
88
89
|
}
|
|
89
|
-
colors = extractSampleColorsFromBuffer(source.bytes)
|
|
90
|
+
colors = await extractSampleColorsFromBuffer(source.bytes)
|
|
90
91
|
}
|
|
91
92
|
}
|
|
92
93
|
return toLosslessJson(generateCssGradient(colors, args.type || 'mesh'))
|
|
@@ -129,7 +130,7 @@ export function registerFrontendTools(ctx, deps) {
|
|
|
129
130
|
if (!source || !source.bytes) {
|
|
130
131
|
throw new Error('Image not found or unreadable: ' + args.image)
|
|
131
132
|
}
|
|
132
|
-
colors = extractSampleColorsFromBuffer(source.bytes)
|
|
133
|
+
colors = await extractSampleColorsFromBuffer(source.bytes)
|
|
133
134
|
}
|
|
134
135
|
}
|
|
135
136
|
return toLosslessJson(checkWcagContrast(colors, args.text_color || '#ffffff'))
|
|
@@ -209,15 +210,31 @@ export function registerFrontendTools(ctx, deps) {
|
|
|
209
210
|
},
|
|
210
211
|
},
|
|
211
212
|
async execute(args, exec) {
|
|
212
|
-
|
|
213
|
-
|
|
214
|
-
|
|
215
|
-
|
|
216
|
-
|
|
217
|
-
|
|
218
|
-
|
|
219
|
-
|
|
220
|
-
|
|
213
|
+
try {
|
|
214
|
+
let sourceBytes = null
|
|
215
|
+
if (args.image) {
|
|
216
|
+
const source = await resolveSource(ctx, exec, args.image)
|
|
217
|
+
if (source && source.bytes) {
|
|
218
|
+
sourceBytes = source.bytes
|
|
219
|
+
}
|
|
220
|
+
}
|
|
221
|
+
const sessionCwd = exec?.agent?.session?.header?.cwd || process.cwd()
|
|
222
|
+
const outDir = path.resolve(sessionCwd, args.out_dir || '.')
|
|
223
|
+
|
|
224
|
+
const res = await generatePwaIconSuite({
|
|
225
|
+
name: args.name || 'App',
|
|
226
|
+
shortName: args.short_name || args.name || 'App',
|
|
227
|
+
themeColor: args.theme_color || '#0f172a',
|
|
228
|
+
backgroundColor: args.background_color || '#ffffff',
|
|
229
|
+
iconsDir: args.icons_dir || 'icons',
|
|
230
|
+
sourcePath: args.image || null,
|
|
231
|
+
imageBytes: sourceBytes,
|
|
232
|
+
outDir,
|
|
233
|
+
})
|
|
234
|
+
return toLosslessJson(res)
|
|
235
|
+
} catch (err) {
|
|
236
|
+
throw new Error(sanitizeErrorAndLogs(err.message || err))
|
|
237
|
+
}
|
|
221
238
|
},
|
|
222
239
|
}),
|
|
223
240
|
)
|
package/lib/tools/generation.js
CHANGED
|
@@ -1,3 +1,4 @@
|
|
|
1
|
+
import { buildPresentationMeta } from '../attachment-helper.js'
|
|
1
2
|
import { registerGenerationPackTool } from './generation-pack.js'
|
|
2
3
|
import { defineTool } from '@deepseek-ai/dsh-tools'
|
|
3
4
|
import path from 'node:path'
|
|
@@ -282,12 +283,16 @@ export function registerGenerationTools(ctx, deps, options = {}) {
|
|
|
282
283
|
const summary = lines.join('\n')
|
|
283
284
|
if (live().deliverAs !== 'image') {
|
|
284
285
|
const urls = list.map((img) => img.url).filter(Boolean)
|
|
285
|
-
|
|
286
|
+
const mdLinks = list.map((img) => ``).join('\n')
|
|
287
|
+
return [{ type: 'text', text: urls.length ? `${summary}\n${mdLinks}` : summary }]
|
|
286
288
|
}
|
|
287
289
|
const blocks = [{ type: 'text', text: summary }]
|
|
288
290
|
for (const img of list) if (img.attachment) blocks.push({ type: 'image', attachment: img.attachment })
|
|
289
291
|
return blocks
|
|
290
292
|
},
|
|
293
|
+
presentationMeta(args, value) {
|
|
294
|
+
return buildPresentationMeta(value)
|
|
295
|
+
},
|
|
291
296
|
},
|
|
292
297
|
isConcurrencySafe: () => false,
|
|
293
298
|
timeoutMs: (config?.timeoutMs || 180000) + 30000,
|
|
@@ -334,7 +339,7 @@ export function registerGenerationTools(ctx, deps, options = {}) {
|
|
|
334
339
|
assertBudgetAvailable(estimatedCost, cfg.dailyBudgetUsd)
|
|
335
340
|
const providers = makeProviders(
|
|
336
341
|
{ fetchImpl: fetch, resolveKey: (ref) => resolveApiKey(ctx, ref), cfg, subscriptionImages },
|
|
337
|
-
{ prompt: effectivePrompt, size, format, seed: args.seed, signal: exec.signal, negativePrompt: effectiveNegative, guidanceScale: effectiveGuidance, source, mask, strength: args.strength, quality: args.quality, style: args.style, aspectPixels, aspectRatio: args.aspect_ratio, faceReference: resolvedFaceRef },
|
|
342
|
+
{ prompt: effectivePrompt, size, format, seed: args.seed, signal: exec.signal, negativePrompt: effectiveNegative, guidanceScale: effectiveGuidance, source, mask, strength: args.strength, quality: args.quality, style: args.style, aspectPixels, aspectRatio: args.aspect_ratio, faceReference: resolvedFaceRef, onProgress: exec?.onProgress, callId: exec?.callId },
|
|
338
343
|
)
|
|
339
344
|
// Subscription providers return {ok:false, reason} instead of throwing:
|
|
340
345
|
// rejection reaches the model as text. Without this guard,
|
package/lib/tools/inspect.js
CHANGED
|
@@ -35,7 +35,7 @@ export function registerInspectTools(ctx, deps, options = {}) {
|
|
|
35
35
|
try {
|
|
36
36
|
const a = await resolveSource(ctx, exec, args.image_a)
|
|
37
37
|
const b = await resolveSource(ctx, exec, args.image_b)
|
|
38
|
-
return pixelDiff(a && a.bytes, b && b.bytes)
|
|
38
|
+
return await pixelDiff(a && a.bytes, b && b.bytes)
|
|
39
39
|
} catch (err) {
|
|
40
40
|
throw new Error(sanitizeErrorAndLogs(err.message || err))
|
|
41
41
|
}
|
|
@@ -69,7 +69,7 @@ export function registerInspectTools(ctx, deps, options = {}) {
|
|
|
69
69
|
try {
|
|
70
70
|
const source = await resolveSource(ctx, exec, args.image)
|
|
71
71
|
if (!source || !source.bytes) throw new Error('Image not found: ' + args.image)
|
|
72
|
-
const analysis = estimateSharpnessAndVariance(source.bytes)
|
|
72
|
+
const analysis = await estimateSharpnessAndVariance(source.bytes)
|
|
73
73
|
const elementNote = args.expected_elements ? ` Verified presence of: ${args.expected_elements}.` : ''
|
|
74
74
|
return {
|
|
75
75
|
qualityScore: analysis.score,
|
package/lib/tools/pattern.js
CHANGED
|
@@ -99,11 +99,15 @@ export function registerPatternTools(ctx, deps) {
|
|
|
99
99
|
const filePath = path.join(outDir, name)
|
|
100
100
|
await writeFile(filePath, gen.bytes)
|
|
101
101
|
|
|
102
|
-
if (gen.bytes
|
|
103
|
-
|
|
104
|
-
|
|
105
|
-
|
|
106
|
-
exec
|
|
102
|
+
if (gen.bytes) {
|
|
103
|
+
try {
|
|
104
|
+
const { scoreImageEdgeWrap } = await import('../pattern-helpers.js')
|
|
105
|
+
const seamScore = await scoreImageEdgeWrap(gen.bytes)
|
|
106
|
+
if (exec?.logger?.debug) {
|
|
107
|
+
exec.logger.debug(`[dsh-image-gen] Pattern seam score: ${seamScore}`)
|
|
108
|
+
}
|
|
109
|
+
} catch (_e) {
|
|
110
|
+
// Non-fatal debug metric
|
|
107
111
|
}
|
|
108
112
|
}
|
|
109
113
|
|
|
@@ -193,7 +193,7 @@ export function registerProcessingTools(ctx, deps) {
|
|
|
193
193
|
const source = await resolveSource(ctx, exec, args.image)
|
|
194
194
|
if (!source || !source.bytes) throw new Error('Source image not found: ' + args.image)
|
|
195
195
|
const cfg = live()
|
|
196
|
-
const res = traceToSvg(source.bytes, { colorMode: args.color_mode })
|
|
196
|
+
const res = await traceToSvg(source.bytes, { colorMode: args.color_mode })
|
|
197
197
|
const stem = `${slugify(args.output_name || 'vector')}-${Date.now().toString(36)}`
|
|
198
198
|
const name = `${stem}.svg`
|
|
199
199
|
const sessionCwd = exec.agent?.session?.header?.cwd
|
|
@@ -255,12 +255,17 @@ export function registerProcessingTools(ctx, deps) {
|
|
|
255
255
|
trackAndAssertLoopGuard(exec?.agent?.session?.id || 'default', { limit: cfg.loopGuardLimit })
|
|
256
256
|
assertBudgetAvailable(0.05, cfg.dailyBudgetUsd)
|
|
257
257
|
const refs = Array.isArray(args.images) ? args.images : [args.images].filter(Boolean)
|
|
258
|
+
if (refs.length < 2) {
|
|
259
|
+
throw new Error('At least two images are required for blending')
|
|
260
|
+
}
|
|
258
261
|
const sources = []
|
|
259
262
|
for (const r of refs) {
|
|
260
263
|
const s = await resolveSource(ctx, exec, r)
|
|
261
264
|
if (s && s.bytes) sources.push(s)
|
|
262
265
|
}
|
|
263
|
-
if (sources.length
|
|
266
|
+
if (sources.length < 2) {
|
|
267
|
+
throw new Error('At least two valid images are required to blend')
|
|
268
|
+
}
|
|
264
269
|
const res = await blendImagesFal(
|
|
265
270
|
{ fetchImpl: fetch, resolveKey: (ref) => resolveApiKey(ctx, ref), cfg },
|
|
266
271
|
{ images: sources, weights: args.weights, prompt: args.prompt, signal: exec.signal }
|
|
@@ -107,7 +107,15 @@ export function registerStyleMatrixTools(ctx, deps) {
|
|
|
107
107
|
blindMode: value.blind_mode,
|
|
108
108
|
cells: value.cells || [],
|
|
109
109
|
})
|
|
110
|
-
|
|
110
|
+
const metaTag = `<dsh-image-gen-matrix>${JSON.stringify({ cells: value.cells, base_prompt: value.base_prompt, blind_mode: value.blind_mode })}</dsh-image-gen-matrix>`
|
|
111
|
+
return [{ type: 'text', text: `${summary}\n\n${metaTag}` }]
|
|
112
|
+
},
|
|
113
|
+
presentationMeta(args, value) {
|
|
114
|
+
return {
|
|
115
|
+
base_prompt: value.base_prompt,
|
|
116
|
+
blind_mode: value.blind_mode,
|
|
117
|
+
cells: value.cells || [],
|
|
118
|
+
}
|
|
111
119
|
},
|
|
112
120
|
},
|
|
113
121
|
isConcurrencySafe: () => false,
|
package/lib/tools/theme-pair.js
CHANGED
|
@@ -99,7 +99,25 @@ export function registerThemePairTools(ctx, deps) {
|
|
|
99
99
|
},
|
|
100
100
|
},
|
|
101
101
|
render(_args, value) {
|
|
102
|
-
|
|
102
|
+
const blocks = renderToolOutput(value)
|
|
103
|
+
const metaTag = `<dsh-image-gen-theme-pair>${JSON.stringify({
|
|
104
|
+
light: value.light,
|
|
105
|
+
dark: value.dark,
|
|
106
|
+
html_snippet: value.html_snippet,
|
|
107
|
+
css_snippet: value.css_snippet,
|
|
108
|
+
})}</dsh-image-gen-theme-pair>`
|
|
109
|
+
if (blocks[0] && blocks[0].type === 'text') {
|
|
110
|
+
blocks[0].text += `\n\n${metaTag}`
|
|
111
|
+
}
|
|
112
|
+
return blocks
|
|
113
|
+
},
|
|
114
|
+
presentationMeta(_args, value) {
|
|
115
|
+
return {
|
|
116
|
+
light: value.light,
|
|
117
|
+
dark: value.dark,
|
|
118
|
+
html_snippet: value.html_snippet,
|
|
119
|
+
css_snippet: value.css_snippet,
|
|
120
|
+
}
|
|
103
121
|
},
|
|
104
122
|
},
|
|
105
123
|
isConcurrencySafe: () => false,
|
package/package.json
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@goodandready/dsh-image-gen",
|
|
3
|
-
"version": "0.11.
|
|
4
|
-
"description": "Image generation for DeepSeek Harness: a generate_image tool with pluggable providers
|
|
3
|
+
"version": "0.11.27",
|
|
4
|
+
"description": "Image generation for DeepSeek Harness: a generate_image tool with pluggable providers \u2014 the FAL queue, any OpenAI-compatible images API, or a ChatGPT/Grok subscription with no API key at all. The picture is shown inline in the conversation; the model receives either a link (works with any chat model) or the image itself (needs dsh-vision-bridge or a vision-capable model).",
|
|
5
5
|
"keywords": [
|
|
6
6
|
"deepseek-harness",
|
|
7
7
|
"dsh",
|