@goodandready/dsh-image-gen 0.10.17 → 0.10.19

package/lib/tools/processing.js

@@ -321,4 +321,176 @@ export function registerProcessingTools(ctx, deps) {
       }),
     )
   }, 'dsh-image-gen: tool blend_images')
+
+  ctx.effect(() => {
+    ctx.tools.register(
+      defineTool({
+        name: 'assemble_image_grid',
+        description:
+          'Assemble 2 to 4 images into a clean side-by-side, 2x2 grid, or vertical collage comparison. '
+          + 'Generates a composite visual with labels, optimal spacing, and lossless fidelity. '
+          + 'Returns safe dual-output with an inline preview and downloadable image asset.',
+        parameters: {
+          images: {
+            type: 'array',
+            items: { type: 'string' },
+            required: true,
+            description: 'List of 2 to 4 image paths or attachment ids to assemble into a grid.',
+          },
+          layout: {
+            type: 'string',
+            enum: ['side_by_side', 'grid_2x2', 'vertical'],
+            description: 'Grid layout arrangement: "side_by_side" (default), "grid_2x2", or "vertical".',
+          },
+          title: {
+            type: 'string',
+            description: 'Optional headline title displayed above the assembled collage.',
+          },
+          output_name: {
+            type: 'string',
+            description: 'Optional custom file name stem for the output grid image.',
+          },
+        },
+        output: {
+          schema: {
+            type: 'object',
+            additionalProperties: true,
+            properties: {
+              summary: { type: 'string' },
+              path: { type: 'string' },
+              url: { type: 'string' },
+              layout: { type: 'string' },
+              imageCount: { type: 'number' },
+              format: { type: 'string' },
+              attachment: { type: 'object', additionalProperties: true },
+            },
+          },
+        },
+        async execute(args, exec) {
+          try {
+            const rawRefs = Array.isArray(args.images) ? args.images : [args.images].filter(Boolean)
+            if (rawRefs.length < 2) throw new Error('assemble_image_grid requires at least 2 images (received ' + rawRefs.length + ')')
+            const limitRefs = rawRefs.slice(0, 4)
+
+            const sources = []
+            for (const r of limitRefs) {
+              const s = await resolveSource(ctx, exec, r)
+              if (s && s.bytes) sources.push(s)
+            }
+            if (sources.length < 2) throw new Error('Could not resolve at least 2 valid image sources from provided references')
+
+            const layout = args.layout || 'side_by_side'
+            const title = args.title ? String(args.title).trim() : ''
+
+            const tileW = 512
+            const tileH = 512
+            const gap = 16
+            const pad = 16
+            const headerH = title ? 48 : 0
+
+            let cols = 2
+            let rows = 1
+            if (layout === 'side_by_side') {
+              cols = sources.length
+              rows = 1
+            } else if (layout === 'vertical') {
+              cols = 1
+              rows = sources.length
+            } else if (layout === 'grid_2x2') {
+              cols = 2
+              rows = Math.ceil(sources.length / 2)
+            }
+
+            const totalW = pad * 2 + cols * tileW + (cols - 1) * gap
+            const totalH = pad * 2 + headerH + rows * tileH + (rows - 1) * gap
+
+            const labels = ['A', 'B', 'C', 'D']
+            let tilesSvg = ''
+
+            sources.forEach((src, idx) => {
+              let c = idx
+              let r = 0
+              if (layout === 'side_by_side') {
+                c = idx
+                r = 0
+              } else if (layout === 'vertical') {
+                c = 0
+                r = idx
+              } else if (layout === 'grid_2x2') {
+                c = idx % 2
+                r = Math.floor(idx / 2)
+              }
+
+              const x = pad + c * (tileW + gap)
+              const y = pad + headerH + r * (tileH + gap)
+              const b64 = Buffer.isBuffer(src.bytes) ? src.bytes.toString('base64') : Buffer.from(src.bytes).toString('base64')
+              const mime = src.mediaType || 'image/png'
+              const dataUri = `data:${mime};base64,${b64}`
+              const label = labels[idx] || String(idx + 1)
+
+              tilesSvg += `
+  <g transform="translate(${x}, ${y})">
+    <clipPath id="clip-${idx}"><rect width="${tileW}" height="${tileH}" rx="8" ry="8"/></clipPath>
+    <rect width="${tileW}" height="${tileH}" rx="8" fill="#18191c" stroke="#2c2e33" stroke-width="1"/>
+    <image href="${dataUri}" width="${tileW}" height="${tileH}" preserveAspectRatio="xMidYMid slice" clip-path="url(#clip-${idx})"/>
+    <rect x="12" y="12" width="28" height="24" rx="4" fill="rgba(0,0,0,0.65)"/>
+    <text x="26" y="29" fill="#ffffff" font-family="-apple-system,BlinkMacSystemFont,'Segoe UI',Roboto,sans-serif" font-size="13" font-weight="700" text-anchor="middle">${label}</text>
+  </g>`
+            })
+
+            const escapedTitle = title
+              .replace(/&/g, '&amp;')
+              .replace(/</g, '&lt;')
+              .replace(/>/g, '&gt;')
+
+            const svgContent = `<?xml version="1.0" encoding="UTF-8"?>
+<svg xmlns="http://www.w3.org/2000/svg" width="${totalW}" height="${totalH}" viewBox="0 0 ${totalW} ${totalH}">
+  <defs>
+    <style>
+      .bg { fill: #0f1012; }
+      .title { fill: #f3f4f6; font-family: -apple-system,BlinkMacSystemFont,'Segoe UI',Roboto,sans-serif; font-size: 16px; font-weight: 600; }
+    </style>
+  </defs>
+  <rect width="100%" height="100%" rx="12" class="bg"/>
+  ${title ? `<text x="${pad}" y="${pad + 24}" class="title">${escapedTitle}</text>` : ''}
+  ${tilesSvg}
+</svg>`
+
+            const svgBuffer = Buffer.from(svgContent, 'utf8')
+            const cfg = live()
+            const stem = `${slugify(args.output_name || 'grid-collage')}-${Date.now().toString(36)}`
+            const name = `${stem}.svg`
+            const { attachment, localUrl } = await saveAttachmentSafe(ctx, { bytes: svgBuffer, mediaType: 'image/svg+xml', name })
+
+            const sessionCwd = exec?.agent?.session?.header?.cwd
+            const outDir = path.resolve(sessionCwd || process.cwd(), cfg.outputDir || 'generated/images')
+            await mkdir(outDir, { recursive: true })
+            const filePath = path.join(outDir, name)
+            await writeFile(filePath, svgBuffer)
+
+            const summary = `### Assembled Image Grid (${layout})\n`
+              + `- **Images**: ${sources.length} sources combined (${layout})\n`
+              + `- **Canvas Size**: ${totalW} × ${totalH} px\n`
+              + `- **File**: \`${filePath}\`\n\n`
+              + `![${title || 'Assembled Grid'}](${localUrl || filePath})`
+
+            return toLosslessJson({
+              summary,
+              path: filePath,
+              url: localUrl,
+              layout,
+              imageCount: sources.length,
+              width: totalW,
+              height: totalH,
+              format: 'svg',
+              attachment,
+            })
+          } catch (err) {
+            throw new Error(sanitizeErrorAndLogs(err.message || err))
+          }
+        },
+      }),
+    )
+  }, 'dsh-image-gen: tool assemble_image_grid')
+
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@goodandready/dsh-image-gen",
-  "version": "0.10.17",
+  "version": "0.10.19",
   "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).",
   "keywords": [
     "deepseek-harness",