@goodandready/dsh-image-gen 0.8.1 → 0.8.2

package/lib/index.js

@@ -23,6 +23,7 @@ import {
   PROVIDER_KEYS,
   buildSidecar,
   makeProviders,
+  normalizeCount,
   normalizeMediaType,
 } from './providers.js'
 
@@ -247,7 +248,7 @@ export function apply(ctx, config) {
         + 'Saves the image to the session workspace and shows it in the conversation. '
         + 'Depending on how the deployment is configured the result carries either the image itself or a link to it; '
         + 'when you only get a link, answer from the prompt and the link rather than claiming to see the picture. '
-        + 'Use for any text-to-image request.',
+        + 'Use for any text-to-image request. Pass count (1-4) to generate several variations in one call; cost scales with count (default 1).',
       parameters: {
         prompt: {
           type: 'string',
@@ -271,6 +272,12 @@ export function apply(ctx, config) {
           type: 'string',
           description: 'Optional file name stem for the saved image (defaults to a slug of the prompt).',
         },
+        count: {
+          type: 'integer',
+          minimum: 1,
+          maximum: 4,
+          description: 'Number of variations to generate in one call, 1-4 (default 1). Cost scales with count.',
+        },
       },
       output: {
         schema: {
@@ -296,19 +303,34 @@ export function apply(ctx, config) {
                 name: { type: 'string' },
               },
             },
+            images: {
+              type: 'array',
+              items: {
+                type: 'object',
+                additionalProperties: false,
+                properties: {
+                  path: { type: 'string' },
+                  url: { type: 'string' },
+                  width: { type: 'integer' },
+                  height: { type: 'integer' },
+                  seed: { type: 'integer' },
+                  prompt: { type: 'string' },
+                  format: { type: 'string' },
+                },
+              },
+            },
           },
         },
-        render(args, value) {
-          const summary = `Image generated (${value.width}×${value.height}, ${value.format}, seed ${value.seed}): ${value.path}`
-          // "link": the result stays text-only, so a chat model without vision
-          // reads it fine; the card turns the link back into a picture.
-          // "image": the picture travels in the result itself, which only a
-          // vision-capable model — or dsh-vision-bridge — can handle.
+      render(args, value) {
+          const list = value.images || [value]
+          const lines = list.map((img) => `Image generated (${img.width}×${img.height}, ${img.format}, seed ${img.seed}): ${img.path}`)
+          const summary = lines.join('\n')
           if (live().deliverAs !== 'image') {
-            return [{ type: 'text', text: value.url ? `${summary}\n${value.url}` : summary }]
+            const urls = list.map((img) => img.url).filter(Boolean)
+            return [{ type: 'text', text: urls.length ? `${summary}\n${urls.join('\n')}` : summary }]
           }
           const blocks = [{ type: 'text', text: summary }]
-          if (value.attachment) blocks.push({ type: 'image', attachment: value.attachment })
+          for (const img of list) if (img.attachment) blocks.push({ type: 'image', attachment: img.attachment })
           return blocks
         },
       },
@@ -335,77 +357,83 @@ export function apply(ctx, config) {
           { fetchImpl: fetch, resolveKey: (ref) => resolveApiKey(ctx, ref), cfg, subscriptionImages },
           { prompt: args.prompt, size, format, seed: args.seed, signal: exec.signal },
         )
-        const generated = await providers[provider]()
         // Подписочные провайдеры (codex/grok) отдают {ok:false, reason} вместо исключения:
         // отказ должен дойти до модели текстом. Без проверки execute шёл дальше с пустыми
         // байтами, и пользователь получал битую карточку вместо внятного отказа.
-        if (generated && generated.ok === false) throw new Error(generated.reason)
-        const bytes = generated.bytes
+        const guard = (generated) => { if (generated && generated.ok === false) throw new Error(generated.reason) }
+        const one = async (jobSeed) => {
+          const gen = await providers[provider](jobSeed)
+          guard(gen)
+          const bytes = gen.bytes
+          const mediaType = gen.mediaType
+          const extension = mediaType === 'image/jpeg' ? 'jpg' : mediaType === 'image/webp' ? 'webp' : 'png'
+          const stem = `${slugify(args.output_name || args.prompt)}-${Date.now().toString(36)}-${jobSeed}`
+          const name = `${stem}.${extension}`
 
-        const mediaType = generated.mediaType
-        const extension = mediaType === 'image/jpeg' ? 'jpg' : mediaType === 'image/webp' ? 'webp' : 'png'
-        const stem = `${slugify(args.output_name || args.prompt)}-${Date.now().toString(36)}`
-        const name = `${stem}.${extension}`
+          const attachment = await ctx.attachments.saveImage({
+            data: new Uint8Array(bytes),
+            mediaType,
+            name,
+          })
 
-        const attachment = await ctx.attachments.saveImage({
-          data: new Uint8Array(bytes),
-          mediaType,
-          name,
-        })
+          const sessionCwd = exec.agent?.session?.header?.cwd
+          const outDir = path.resolve(sessionCwd || process.cwd(), cfg.outputDir)
+          await mkdir(outDir, { recursive: true })
+          const filePath = path.join(outDir, name)
+          await writeFile(filePath, bytes)
 
-        // Relative outputDir belongs to the session's workspace, not to the
-        // directory the harness happens to be started from.
-        const sessionCwd = exec.agent?.session?.header?.cwd
-        const outDir = path.resolve(sessionCwd || process.cwd(), cfg.outputDir)
-        await mkdir(outDir, { recursive: true })
-        const filePath = path.join(outDir, name)
-        await writeFile(filePath, bytes)
-        // Sidecar: рядом с картинкой — метаданные генерации. Каталог становится
-        // самодокументируемым (галерея/повтор/диагностика читают их без БД).
-        await writeFile(
-          path.join(outDir, `${stem}.json`),
-          JSON.stringify(buildSidecar({
-            prompt: args.prompt,
-            size,
-            format,
-            seed: generated.seed,
-            provider,
-            deliverAs,
-            width: generated.width || attachment.width,
-            height: generated.height || attachment.height,
-            mediaType,
-            attachmentId: attachment.attachmentId,
-            url: deliverAs === 'image' && generated.sourceUrl ? generated.sourceUrl : localUrl,
-          }), null, 2),
-        )
+          const localUrl = '/dsh-image-gen/image?id=' + encodeURIComponent(attachment.attachmentId)
+            + '&mt=' + encodeURIComponent(attachment.mediaType)
+            + '&b=' + encodeURIComponent(String(attachment.bytes))
+            + '&w=' + encodeURIComponent(String(attachment.width))
+            + '&h=' + encodeURIComponent(String(attachment.height))
 
-        // The plugin's own route outlives the fal.media link, and the card
-        // needs the reference anyway to verify the bytes.
-        const localUrl = '/dsh-image-gen/image?id=' + encodeURIComponent(attachment.attachmentId)
-          + '&mt=' + encodeURIComponent(attachment.mediaType)
-          + '&b=' + encodeURIComponent(String(attachment.bytes))
-          + '&w=' + encodeURIComponent(String(attachment.width))
-          + '&h=' + encodeURIComponent(String(attachment.height))
+          // Sidecar: рядом с картинкой — метаданные генерации. Каталог становится
+          // самодокументируемым (галерея/повтор/диагностика читают их без БД).
+          await writeFile(
+            path.join(outDir, `${stem}.json`),
+            JSON.stringify(buildSidecar({
+              prompt: args.prompt,
+              size,
+              format,
+              seed: gen.seed,
+              provider,
+              deliverAs,
+              width: gen.width || attachment.width,
+              height: gen.height || attachment.height,
+              mediaType,
+              attachmentId: attachment.attachmentId,
+              url: deliverAs === 'image' && gen.sourceUrl ? gen.sourceUrl : localUrl,
+            }), null, 2),
+          )
 
-        return {
-          path: filePath,
-          // Прямой ссылки может и не быть: провайдер, отдающий base64, ничего
-          // не публикует. Тогда и в режиме "image" остаётся наша ссылка.
-          url: deliverAs === 'image' && generated.sourceUrl ? generated.sourceUrl : localUrl,
-          width: generated.width || attachment.width,
-          height: generated.height || attachment.height,
-          seed: generated.seed,
-          prompt: args.prompt,
-          format: mediaType.replace('image/', ''),
-          attachment: {
-            attachmentId: attachment.attachmentId,
-            mediaType: attachment.mediaType,
-            bytes: attachment.bytes,
-            width: attachment.width,
-            height: attachment.height,
-            name: attachment.name,
-          },
+          return {
+            path: filePath,
+            url: deliverAs === 'image' && gen.sourceUrl ? gen.sourceUrl : localUrl,
+            width: gen.width || attachment.width,
+            height: gen.height || attachment.height,
+            seed: gen.seed,
+            prompt: args.prompt,
+            format: mediaType.replace('image/', ''),
+            attachment: {
+              attachmentId: attachment.attachmentId,
+              mediaType: attachment.mediaType,
+              bytes: attachment.bytes,
+              width: attachment.width,
+              height: attachment.height,
+              name: attachment.name,
+            },
+          }
+        }
+
+        const count = normalizeCount(args.count)
+        const seedBase = args.seed ?? Math.floor(Math.random() * 100000)
+        const images = []
+        for (let i = 0; i < count; i += 1) {
+          images.push(await one(seedBase + i))
         }
+        const first = images[0]
+        return { ...first, images }
       },
     }),
   )

package/lib/providers.js

@@ -9,6 +9,13 @@
 
 export const PROVIDER_KEYS = ['fal', 'custom', 'codex', 'grok']
 
+/** Привести count из аргумента инструмента к диапазону 1..4. */
+export function normalizeCount(count) {
+  const n = Math.floor(Number(count))
+  if (!Number.isFinite(n)) return 1
+  return Math.max(1, Math.min(4, n))
+}
+
 // Размеры, которыми оперируют подписки, — свой набор, не похожий на именованные
 // размеры FAL. Перевод один к одному по смыслу: квадрат, вертикаль, горизонталь.
 export const SUBSCRIPTION_SIZES = {
@@ -130,10 +137,10 @@ export function makeProviders(deps, job) {
   const { fetchImpl, resolveKey, cfg } = deps
   const { prompt, size, format, seed, signal } = job
 
-  async function fal() {
+  async function fal(seedArg = seed) {
     const key = await resolveKey(cfg.apiKeyEnv)
     const body = { prompt, image_size: size, num_images: 1 }
-    if (seed !== undefined) body.seed = seed
+    if (seedArg !== undefined) body.seed = seedArg
     if (format !== 'png') body.output_format = format
 
     const submit = await submitJob(fetchImpl, cfg.baseURL, cfg.model, key, body, signal)
@@ -158,13 +165,13 @@ export function makeProviders(deps, job) {
       mediaType: normalizeMediaType(image.content_type, format),
       width: image.width ?? 0,
       height: image.height ?? 0,
-      seed: result.seed ?? seed ?? 0,
+      seed: result.seed ?? seedArg ?? 0,
       sourceUrl: image.url,
     }
   }
 
   // Любой OpenAI-совместимый API картинок. Один запрос вместо очереди FAL.
-  async function custom() {
+  async function custom(seedArg = seed) {
     const base = String(cfg.customBaseURL || '').replace(/\/+$/, '')
     if (!base) throw new Error('Custom image provider: base URL is not configured (Settings → Image generation)')
     if (!cfg.customModel) throw new Error('Custom image provider: model is not configured')
@@ -204,7 +211,7 @@ export function makeProviders(deps, job) {
         mediaType: normalizeMediaType(data.output_format || '', format),
         width: 0,
         height: 0,
-        seed: seed ?? 0,
+        seed: seedArg ?? 0,
         sourceUrl: '',
       }
     }
@@ -223,7 +230,7 @@ export function makeProviders(deps, job) {
       mediaType: normalizeMediaType(contentType, format),
       width: 0,
       height: 0,
-      seed: seed ?? 0,
+      seed: seedArg ?? 0,
       sourceUrl: item.url,
     }
   }
@@ -232,7 +239,7 @@ export function makeProviders(deps, job) {
   // только запрос и разбор ответа. Токен сюда не попадает вовсе — служба
   // отдаёт готовую картинку, а не ключ доступа.
   function subscription(provider) {
-    return async function generate() {
+    return async function generate(seedArg = seed) {
       const images = deps.subscriptionImages
       if (!images || typeof images.generate !== 'function') {
         return {
@@ -263,7 +270,7 @@ export function makeProviders(deps, job) {
         mediaType: 'image/png',
         width: 0,
         height: 0,
-        seed: seed ?? 0,
+        seed: seedArg ?? 0,
         sourceUrl: '',
         revisedPrompt: first.revisedPrompt || '',
       }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@goodandready/dsh-image-gen",
-  "version": "0.8.1",
+  "version": "0.8.2",
   "description": "Image generation for DeepSeek Harness: a generate_image tool with pluggable providers — 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",