@goodandready/dsh-image-gen 0.11.29 → 0.11.31

package/lib/providers.js

@@ -536,19 +536,7 @@ export async function blendImageBuffers(images, weights) {
     const bytes = await sharp(out, { raw: { width: targetW, height: targetH, channels: 4 } }).png().toBuffer()
     return { bytes, width: targetW, height: targetH, mediaType: 'image/png', weights: normWeights }
   } catch (err) {
-    // Fallback for mock or synthetic test buffers that cannot be parsed by sharp
-    const maxLen = Math.max(...images.map(img => (img.bytes?.length || 0)))
-    const out = Buffer.alloc(maxLen)
-    for (let p = 0; p < maxLen; p++) {
-      let val = 0
-      for (let i = 0; i < images.length; i++) {
-        const b = images[i].bytes || Buffer.alloc(0)
-        const byteVal = p < b.length ? b[p] : 0
-        val += byteVal * normWeights[i]
-      }
-      out[p] = Math.round(Math.min(255, Math.max(0, val)))
-    }
-    return { bytes: out, width: 0, height: 0, mediaType: 'image/png', weights: normWeights }
+    throw new Error(`Failed to decode image buffer for blending: ${err.message}`)
   }
 }
 
@@ -791,7 +779,8 @@ export async function testProviderConnection(deps, providerId) {
     if (p === 'gemini') {
       const key = await resolveKey(cfg?.geminiKeyEnv || 'GEMINI_API_KEY')
       if (!key) return { ok: false, state: 'not_configured', latencyMs: 0, message: 'Google Gemini API key not configured' }
-      const model = cfg?.geminiModel || 'gemini-2.0-flash-exp-image-generation'
+      let model = cfg?.geminiModel || 'gemini-3.1-flash-image'
+      if (model.startsWith('imagen-') || model.startsWith('gemini-2.0-flash-exp')) model = 'gemini-3.1-flash-image'
       let res
       try {
         res = await fetchImpl(`https://generativelanguage.googleapis.com/v1beta/models/${model}?key=${encodeURIComponent(key)}`, {

package/lib/quality-gate.js

@@ -1,5 +1,5 @@
 // lib/quality-gate.js
-// Automatic Quality Gate and Silent Re-roll Engine (#166)
+// Automatic Quality Gate and Silent Re-roll Engine (#166, #353)
 
 import { estimateSharpnessAndVariance } from './providers.js'
 
@@ -97,7 +97,24 @@ export async function executeWithQualityGate(generateFn, {
   const attempts = enabled ? Math.max(1, maxRerolls + 1) : 1
 
   for (let attempt = 0; attempt < attempts; attempt++) {
-    lastGenerated = await generateFn(currentSeed)
+    try {
+      lastGenerated = await generateFn(currentSeed)
+    } catch (err) {
+      // #353: If a subsequent re-roll exceeds budget, abort further retries and preserve the initial output
+      if (lastGenerated && (err.message?.includes('budget exceeded') || err.message?.includes('Daily image generation budget'))) {
+        logger(`Quality gate re-roll halted by budget limit: ${err.message}`)
+        return {
+          generated: lastGenerated,
+          qualityReport: {
+            ...lastReport,
+            rerolls,
+            exhausted: true,
+            budgetExceeded: true,
+          },
+        }
+      }
+      throw err
+    }
 
     if (!enabled) {
       return {

package/lib/tools/character-sheet.js

@@ -15,7 +15,7 @@ import {
   
 } from '../providers.js'
 import { resolveFallbackChain, executeWithFallback } from '../fallback-router.js'
-import { assertBudgetAvailable, calculateGenerationCost, recordSpend } from '../cost-meter.js'
+import { assertBudgetAvailable, calculateGenerationCost, reserveSpend, recordSpend } from '../cost-meter.js'
 import { renderToolOutput } from '../attachment-helper.js'
 import {
   CHARACTER_SHEET_LAYOUTS,
@@ -152,15 +152,20 @@ export function registerCharacterSheetTools(ctx, deps) {
             size: cfg.defaultSize || '1024x1024',
             count: 1,
           })
-          assertBudgetAvailable(estimatedCost, cfg.dailyBudgetUsd)
-
-          const gen = await executeWithFallback(
-            providers,
-            fallbackChain,
-            seed,
-            sheetPlan.prompt,
-            { logger: ctx.logger },
-          )
+          const reservation = reserveSpend(estimatedCost, cfg.dailyBudgetUsd)
+          let gen
+          try {
+            gen = await executeWithFallback(
+              providers,
+              fallbackChain,
+              seed,
+              sheetPlan.prompt,
+              { logger: ctx.logger },
+            )
+          } catch (err) {
+            reservation.release()
+            throw err
+          }
 
           const usedProvider = gen._fallback?.providerUsed || provider
           const finalBytes = gen.bytes
@@ -196,7 +201,7 @@ export function registerCharacterSheetTools(ctx, deps) {
           }
 
           const effectiveCost = Number(gen.cost) || estimatedCost
-          recordSpend(effectiveCost, {
+          reservation.commit(effectiveCost, {
             ctx,
             meta: { provider: usedProvider, model: cfg.model || cfg.customModel, prompt: sheetPlan.prompt, seed },
           })

package/lib/tools/diagram.js

@@ -13,7 +13,7 @@ import {
 } from '../providers.js'
 import { buildDualOutputMarkdown } from '../resolve-image.js'
 import { resolveFallbackChain, executeWithFallback } from '../fallback-router.js'
-import { assertBudgetAvailable, calculateGenerationCost, recordSpend } from '../cost-meter.js'
+import { assertBudgetAvailable, calculateGenerationCost, reserveSpend, recordSpend } from '../cost-meter.js'
 import { renderToolOutput } from '../attachment-helper.js'
 import {
   DIAGRAM_3D_STYLES,
@@ -148,15 +148,20 @@ export function registerDiagramTools(ctx, deps) {
             size: cfg.defaultSize || '1024x1024',
             count: 1,
           })
-          assertBudgetAvailable(estimatedCost, cfg.dailyBudgetUsd)
-
-          const gen = await executeWithFallback(
-            providers,
-            fallbackChain,
-            seed,
-            plan.prompt,
-            { logger: ctx.logger },
-          )
+          const reservation = reserveSpend(estimatedCost, cfg.dailyBudgetUsd)
+          let gen
+          try {
+            gen = await executeWithFallback(
+              providers,
+              fallbackChain,
+              seed,
+              plan.prompt,
+              { logger: ctx.logger },
+            )
+          } catch (err) {
+            reservation.release()
+            throw err
+          }
 
           const usedProvider = gen._fallback?.providerUsed || provider
           const finalBytes = gen.bytes
@@ -192,7 +197,7 @@ export function registerDiagramTools(ctx, deps) {
           }
 
           const effectiveCost = Number(gen.cost) || estimatedCost
-          recordSpend(effectiveCost, {
+          reservation.commit(effectiveCost, {
             ctx,
             meta: { provider: usedProvider, model: cfg.model || cfg.customModel, prompt: plan.prompt, seed },
           })

package/lib/tools/editing.js

@@ -87,7 +87,8 @@ export function registerEditingTools(ctx, deps, options = {}) {
             if (cfg && cfg.enabled === false) {
               throw new Error('Image generation is disabled in settings (dsh-image-gen.enabled is false); enable it in Settings → Image generation.')
             }
-            trackAndAssertLoopGuard(exec?.agent?.session?.id || 'default', { limit: cfg.loopGuardLimit })
+            const sessionId = exec?.agent?.session?.id || exec?.agent?.session?.header?.id || 'session'
+            trackAndAssertLoopGuard(sessionId, { limit: cfg.loopGuardLimit, prompt: args.prompt || '', exec })
             const estimatedCostEdit = calculateGenerationCost({
               provider: cfg.provider || 'fal',
               model: cfg.model,
@@ -242,7 +243,8 @@ export function registerEditingTools(ctx, deps, options = {}) {
             if (cfg && cfg.enabled === false) {
               throw new Error('Image generation is disabled in settings (dsh-image-gen.enabled is false); enable it in Settings → Image generation.')
             }
-            trackAndAssertLoopGuard(exec?.agent?.session?.id || 'default', { limit: cfg.loopGuardLimit })
+            const sessionId = exec?.agent?.session?.id || exec?.agent?.session?.header?.id || 'session'
+            trackAndAssertLoopGuard(sessionId, { limit: cfg.loopGuardLimit, prompt: args.prompt || '', exec })
             const count = normalizeCount(args.count || 1)
             const estimatedCostVary = calculateGenerationCost({
               provider: cfg.provider || 'fal',
@@ -383,7 +385,8 @@ export function registerEditingTools(ctx, deps, options = {}) {
             if (cfg && cfg.enabled === false) {
               throw new Error('Image generation is disabled in settings (dsh-image-gen.enabled is false); enable it in Settings → Image generation.')
             }
-            trackAndAssertLoopGuard(exec?.agent?.session?.id || 'default', { limit: cfg.loopGuardLimit })
+            const sessionId = exec?.agent?.session?.id || exec?.agent?.session?.header?.id || 'session'
+            trackAndAssertLoopGuard(sessionId, { limit: cfg.loopGuardLimit, prompt: args.prompt || '', exec })
             const count = normalizeCount(args.count || 1)
             const estimatedCost = calculateGenerationCost({
               provider: cfg.provider || 'fal',

package/lib/tools/generation.js

@@ -15,7 +15,7 @@ import {
 } from '../providers.js'
 import { buildDualOutputMarkdown } from '../resolve-image.js'
 import { executeWithQualityGate } from '../quality-gate.js'
-import { calculateGenerationCost, assertBudgetAvailable } from '../cost-meter.js'
+import { calculateGenerationCost, assertBudgetAvailable, reserveSpend } from '../cost-meter.js'
 import { trackAndAssertLoopGuard } from '../loop-guard.js'
 import { getCachedGeneration } from '../generation-cache.js'
 import { resolveFallbackChain, executeWithFallback } from '../fallback-router.js'
@@ -378,14 +378,25 @@ export function registerGenerationTools(ctx, deps, options = {}) {
           async (s, p) => {
             let totalAttempts = 0
             let accumulatedCost = 0
-            // #353: Pipeline evaluates raw gen.bytes first; saves artifact ONCE after acceptance
+            // #353, #355: Pipeline reserves budget per attempt, evaluates raw gen.bytes, and commits spend per attempt
+            const perAttemptCost = Number(estimatedCost / totalCount) || 0.02
             const rawGenerate = async (seedVal) => {
+              const reservation = reserveSpend(perAttemptCost, cfg.dailyBudgetUsd)
               totalAttempts++
-              const gen = await providers[k](seedVal, p)
-              guard(gen)
-              const attemptCost = Number(gen.cost) || estimatedCost
-              accumulatedCost += attemptCost
-              return gen
+              try {
+                const gen = await providers[k](seedVal, p)
+                guard(gen)
+                const attemptCost = Number(gen.cost) || perAttemptCost
+                accumulatedCost += attemptCost
+                reservation.commit(attemptCost, {
+                  ctx,
+                  meta: { provider: k, model: cfg.model || cfg.customModel, prompt: p, seed: seedVal },
+                })
+                return gen
+              } catch (err) {
+                reservation.release()
+                throw err
+              }
             }
 
             const { generated: acceptedGen, qualityReport } = await executeWithQualityGate(
@@ -398,7 +409,7 @@ export function registerGenerationTools(ctx, deps, options = {}) {
               }
             )
 
-            // Save the accepted artifact ONCE
+            // Save the accepted artifact ONCE (spend already committed per attempt)
             const saved = await saveGeneratedImageArtifact({
               ctx,
               exec,
@@ -417,6 +428,7 @@ export function registerGenerationTools(ctx, deps, options = {}) {
               slugify,
               estimatedCost: accumulatedCost || estimatedCost,
               totalCount,
+              recordSpend: false,
             })
 
             return {
@@ -500,12 +512,17 @@ export function registerGenerationTools(ctx, deps, options = {}) {
           const hPrompt = h
           const found = findCachedGeneration(historyEntries, hPrompt)
           if (found) return await cachedResult(found)
+          const requestedCriteria = {
+            format: args.output_format || cfg.defaultFormat || 'png',
+            aspectRatio: args.aspect_ratio || cfg.defaultAspectRatio,
+            size: args.aspect_ratio ? 'custom' : (args.image_size || cfg.defaultSize),
+          }
           if (cfg.cacheBySeed) {
-            const bySeed = await findCached(historyEntries, seedVal, promptVal)
+            const bySeed = await findCached(historyEntries, seedVal, promptVal, requestedCriteria)
             if (bySeed) return await cachedResult(bySeed)
           }
           if (cfg.cacheByPrompt) {
-            const byPrompt = await findCachedByPrompt(historyEntries, promptVal)
+            const byPrompt = await findCachedByPrompt(historyEntries, promptVal, requestedCriteria)
             if (byPrompt) return await cachedResult(byPrompt)
           }
           return undefined

package/lib/tools/pattern.js

@@ -67,7 +67,8 @@ export function registerPatternTools(ctx, deps) {
         async execute(args, exec) {
           try {
             const cfg = live()
-            trackAndAssertLoopGuard(exec?.agent?.session?.id || 'default', { limit: cfg.loopGuardLimit })
+            const sessionId = exec?.agent?.session?.id || exec?.agent?.session?.header?.id || 'session'
+            trackAndAssertLoopGuard(sessionId, { limit: cfg.loopGuardLimit, prompt: args.motif || '', exec })
             const density = normalizeDensity(args.density)
             const prompt = buildSeamlessPrompt(args.motif, args.style, density)
             const provider = cfg.provider || 'fal'

package/lib/tools/processing-basic.js

@@ -15,7 +15,7 @@ import {
   saveAttachmentSafe,
 } from '../providers.js'
 import { buildDualOutputMarkdown } from '../resolve-image.js'
-import { assertBudgetAvailable, recordSpend } from '../cost-meter.js'
+import { assertBudgetAvailable, reserveSpend, recordSpend } from '../cost-meter.js'
 import { trackAndAssertLoopGuard } from '../loop-guard.js'
 import { sanitizeErrorAndLogs } from '../security.js'
 
@@ -55,7 +55,7 @@ export function registerProcessingTools(ctx, deps) {
             const source = await resolveSource(ctx, exec, args.image)
             if (!source || !source.bytes) throw new Error('Source image not found: ' + args.image)
           const cfg = live()
-          assertBudgetAvailable(0.02, cfg.dailyBudgetUsd)
+          const resvNobg = reserveSpend(0.02, cfg.dailyBudgetUsd)
           const res = await removeBackgroundFal(
             { fetchImpl: fetch, resolveKey: (ref) => resolveApiKey(ctx, ref), cfg },
             { imageBytes: source.bytes, mediaType: source.mediaType, model: args.model, signal: exec?.signal }
@@ -68,7 +68,7 @@ export function registerProcessingTools(ctx, deps) {
           await mkdir(outDir, { recursive: true })
           const filePath = path.join(outDir, name)
           await writeFile(filePath, res.bytes)
-          recordSpend(0.02, { ctx, meta: { tool: 'remove_background', model: args.model || 'fal-ai/birefnet' } })
+          resvNobg.commit(0.02, { ctx, meta: { tool: 'remove_background', model: args.model || 'fal-ai/birefnet' } })
           const summary = buildDualOutputMarkdown({
             action: 'background removed',
             filePath,
@@ -126,7 +126,7 @@ export function registerProcessingTools(ctx, deps) {
             const source = await resolveSource(ctx, exec, args.image)
             if (!source || !source.bytes) throw new Error('Source image not found: ' + args.image)
           const cfg = live()
-          assertBudgetAvailable(0.025, cfg.dailyBudgetUsd)
+          const resvUpscale = reserveSpend(0.025, cfg.dailyBudgetUsd)
           const res = await upscaleImageFal(
             { fetchImpl: fetch, resolveKey: (ref) => resolveApiKey(ctx, ref), cfg },
             { imageBytes: source.bytes, mediaType: source.mediaType, scale: args.scale, prompt: args.prompt, creativity: args.creativity, signal: exec?.signal }
@@ -139,7 +139,7 @@ export function registerProcessingTools(ctx, deps) {
           await mkdir(outDir, { recursive: true })
           const filePath = path.join(outDir, name)
           await writeFile(filePath, res.bytes)
-          recordSpend(0.025, { ctx, meta: { tool: 'upscale_image', scale: args.scale } })
+          resvUpscale.commit(0.025, { ctx, meta: { tool: 'upscale_image', scale: args.scale } })
           const summary = buildDualOutputMarkdown({
             action: `upscaled (${args.scale || 2}x)`,
             filePath,
@@ -252,8 +252,9 @@ export function registerProcessingTools(ctx, deps) {
         async execute(args, exec) {
           try {
             const cfg = live()
-            trackAndAssertLoopGuard(exec?.agent?.session?.id || 'default', { limit: cfg.loopGuardLimit })
-            assertBudgetAvailable(0.05, cfg.dailyBudgetUsd)
+            const sessionId = exec?.agent?.session?.id || exec?.agent?.session?.header?.id || 'session'
+            trackAndAssertLoopGuard(sessionId, { limit: cfg.loopGuardLimit, prompt: args.prompt || '', exec })
+            const resvBlend = reserveSpend(0.05, cfg.dailyBudgetUsd)
             const refs = Array.isArray(args.images) ? args.images : [args.images].filter(Boolean)
           if (refs.length < 2) {
             throw new Error('At least two images are required for blending')
@@ -278,7 +279,7 @@ export function registerProcessingTools(ctx, deps) {
           await mkdir(outDir, { recursive: true })
           const filePath = path.join(outDir, name)
           await writeFile(filePath, res.bytes)
-          recordSpend(0.05, { ctx, meta: { tool: 'blend_images' } })
+          resvBlend.commit(0.05, { ctx, meta: { tool: 'blend_images' } })
           const summary = buildDualOutputMarkdown({
             action: 'blended composition',
             filePath,

package/lib/tools/responsive.js

@@ -101,7 +101,8 @@ export function registerResponsiveTools(ctx, deps) {
         async execute(args, exec) {
           try {
             const cfg = live()
-            trackAndAssertLoopGuard(exec?.agent?.session?.id || 'default', { limit: cfg.loopGuardLimit })
+            const sessionId = exec?.agent?.session?.id || exec?.agent?.session?.header?.id || 'session'
+            trackAndAssertLoopGuard(sessionId, { limit: cfg.loopGuardLimit, prompt: args.prompt || '', exec })
             if (!args.prompt || !String(args.prompt).trim()) {
               throw new Error('prompt is required for generate_responsive_mockups')
             }

package/lib/tools/sketch.js

@@ -87,7 +87,8 @@ export function registerSketchTools(ctx, deps) {
         async execute(args, exec) {
           try {
             const cfg = live()
-            trackAndAssertLoopGuard(exec?.agent?.session?.id || 'default', { limit: cfg.loopGuardLimit })
+            const sessionId = exec?.agent?.session?.id || exec?.agent?.session?.header?.id || 'session'
+            trackAndAssertLoopGuard(sessionId, { limit: cfg.loopGuardLimit, prompt: args.prompt || '', exec })
             const estimated = calculateGenerationCost({
               provider: cfg.provider || 'fal',
               model: cfg.model,

package/lib/tools/spritesheet.js

@@ -74,7 +74,8 @@ export function registerSpritesheetTools(ctx, deps) {
         async execute(args, exec) {
           try {
             const cfg = live()
-            trackAndAssertLoopGuard(exec?.agent?.session?.id || 'default', { limit: cfg.loopGuardLimit })
+            const sessionId = exec?.agent?.session?.id || exec?.agent?.session?.header?.id || 'session'
+            trackAndAssertLoopGuard(sessionId, { limit: cfg.loopGuardLimit, prompt: args.prompt || '', exec })
             const animation = String(args.animation || 'idle').toLowerCase()
             const animKey = ANIMATION_PRESETS[animation] ? animation : 'idle'
             const frameCount = clampFrames(args.frames, animKey)

package/lib/tools/style-matrix.js

@@ -14,7 +14,7 @@ import {
   saveAttachmentSafe,
 } from '../providers.js'
 import { resolveFallbackChain, executeWithFallback } from '../fallback-router.js'
-import { assertBudgetAvailable, calculateGenerationCost, recordSpend } from '../cost-meter.js'
+import { assertBudgetAvailable, calculateGenerationCost, reserveSpend, recordSpend } from '../cost-meter.js'
 import {
   normalizeMatrixStyles,
   buildMatrixCellPrompt,
@@ -145,7 +145,7 @@ export function registerStyleMatrixTools(ctx, deps) {
             size: cfg.defaultSize || '1024x1024',
             count: cellCount,
           })
-          assertBudgetAvailable(estimatedCost, cfg.dailyBudgetUsd)
+          const batchReservation = reserveSpend(estimatedCost, cfg.dailyBudgetUsd)
 
           const chain = resolveFallbackChain(provider, cfg.fallbackProviders, PROVIDER_KEYS)
           const quadrantIds = ['A', 'B', 'C', 'D']
@@ -187,7 +187,7 @@ export function registerStyleMatrixTools(ctx, deps) {
             })
 
             const cellCost = Number(gen.cost) || +(estimatedCost / cellCount).toFixed(4)
-            recordSpend(cellCost, {
+            batchReservation.commit(cellCost, {
               ctx,
               meta: { provider, model: cfg.model || cfg.customModel, prompt: cellPrompt, seed: cellSeed },
             })

package/lib/tools/theme-pair.js

@@ -13,7 +13,7 @@ import {
   saveAttachmentSafe,
 } from '../providers.js'
 import { resolveFallbackChain, executeWithFallback } from '../fallback-router.js'
-import { assertBudgetAvailable, calculateGenerationCost, recordSpend } from '../cost-meter.js'
+import { assertBudgetAvailable, calculateGenerationCost, reserveSpend, recordSpend } from '../cost-meter.js'
 import { renderToolOutput } from '../attachment-helper.js'
 import {
   THEME_STYLES,
@@ -153,7 +153,7 @@ export function registerThemePairTools(ctx, deps) {
             size: cfg.defaultSize || '1024x1024',
             count: 2,
           })
-          assertBudgetAvailable(estimatedCost, cfg.dailyBudgetUsd)
+          const batchReservation = reserveSpend(estimatedCost, cfg.dailyBudgetUsd)
 
           const chain = resolveFallbackChain(primaryProvider, cfg.fallbackProviders, PROVIDER_KEYS)
           const baseSeed = typeof args.seed === 'number' ? args.seed : Math.floor(Math.random() * 1000000)
@@ -202,7 +202,7 @@ export function registerThemePairTools(ctx, deps) {
             height: lightGen.height || aspectPixels[1],
           })
           const halfCost = +(estimatedCost / 2).toFixed(4)
-          recordSpend(Number(lightGen.cost) || halfCost, {
+          batchReservation.commit(Number(lightGen.cost) || halfCost, {
             ctx,
             meta: { provider: primaryProvider, model: cfg.model || cfg.customModel, prompt: lightPrompt },
           })
@@ -245,7 +245,7 @@ export function registerThemePairTools(ctx, deps) {
             width: darkGen.width || aspectPixels[0],
             height: darkGen.height || aspectPixels[1],
           })
-          recordSpend(Number(darkGen.cost) || halfCost, {
+          batchReservation.commit(Number(darkGen.cost) || halfCost, {
             ctx,
             meta: { provider: primaryProvider, model: cfg.model || cfg.customModel, prompt: darkPrompt },
           })

package/lib/tools/ui-asset.js

@@ -15,7 +15,7 @@ import {
   saveAttachmentSafe,
 } from '../providers.js'
 import { resolveFallbackChain, executeWithFallback } from '../fallback-router.js'
-import { assertBudgetAvailable, calculateGenerationCost, recordSpend } from '../cost-meter.js'
+import { assertBudgetAvailable, calculateGenerationCost, reserveSpend, recordSpend } from '../cost-meter.js'
 import { buildDualOutputMarkdown } from '../resolve-image.js'
 import {
   UI_ASSET_TYPES,
@@ -168,14 +168,19 @@ export function registerUiAssetTools(ctx, deps) {
             size: cfg.defaultSize || '1024x1024',
             count: 1,
           })
-          assertBudgetAvailable(estimatedCost, cfg.dailyBudgetUsd)
-
+          const reservation = reserveSpend(estimatedCost, cfg.dailyBudgetUsd)
           const chain = resolveFallbackChain(provider, cfg.fallbackProviders, PROVIDER_KEYS)
           const seedVal = args.seed ?? Math.floor(Math.random() * 100000)
 
-          const gen = await executeWithFallback(providers, chain, seedVal, synthesizedPrompt, {
-            logger: ctx.logger,
-          })
+          let gen
+          try {
+            gen = await executeWithFallback(providers, chain, seedVal, synthesizedPrompt, {
+              logger: ctx.logger,
+            })
+          } catch (err) {
+            reservation.release()
+            throw err
+          }
 
           let finalBytes = gen.bytes
           let mediaType = gen.mediaType || 'image/png'
@@ -211,7 +216,7 @@ export function registerUiAssetTools(ctx, deps) {
           })
 
           const effectiveCost = Number(gen.cost) || estimatedCost
-          recordSpend(effectiveCost, {
+          reservation.commit(effectiveCost, {
             ctx,
             meta: { provider, model: cfg.model || cfg.customModel, prompt: synthesizedPrompt, seed: seedVal },
           })

package/lib/tools/vision-ocr.js

@@ -12,7 +12,7 @@ import {
 } from '../providers.js'
 import { buildDualOutputMarkdown } from '../resolve-image.js'
 import { resolveFallbackChain, executeWithFallback } from '../fallback-router.js'
-import { assertBudgetAvailable, calculateGenerationCost, recordSpend } from '../cost-meter.js'
+import { assertBudgetAvailable, calculateGenerationCost, reserveSpend, recordSpend } from '../cost-meter.js'
 import { renderToolOutput } from '../attachment-helper.js'
 import {
   normalizeBbox,
@@ -150,8 +150,18 @@ export function registerVisionOcrTools(ctx, deps) {
             throw new Error(`Unable to resolve source image for text replacement: ${args.image}`)
           }
 
-          const imgWidth = source.width || 1024
-          const imgHeight = source.height || 1024
+          let imgWidth = 1024
+          let imgHeight = 1024
+          try {
+            const sharpMod = await import('sharp')
+            const sharp = sharpMod.default || sharpMod
+            const meta = await sharp(source.bytes).metadata()
+            imgWidth = meta.width || source.width || 1024
+            imgHeight = meta.height || source.height || 1024
+          } catch (_err) {
+            imgWidth = source.width || 1024
+            imgHeight = source.height || 1024
+          }
 
           // 1. Gather text items and bounding boxes
           const replacements = Array.isArray(args.replacements) ? args.replacements : []
@@ -162,13 +172,14 @@ export function registerVisionOcrTools(ctx, deps) {
             let bbox = item.bbox
             if (!bbox && item.find && ctx.tools?.get) {
               const groundTool = ctx.tools.get('vision_ground')
-              if (groundTool && typeof groundTool.execute === 'function' && source.ref) {
+              if (groundTool && typeof groundTool.execute === 'function') {
                 try {
                   const gRes = await groundTool.execute({
-                    attachmentId: source.ref.startsWith('sha256:') ? source.ref : undefined,
+                    image: args.image,
+                    attachmentId: source.ref?.startsWith('sha256:') ? source.ref : (args.image?.startsWith('sha256:') ? args.image : undefined),
                     target: `text "${item.find}"`,
                   }, execCtx)
-                  if (gRes?.bbox && gRes.bbox.length === 4) {
+                  if (gRes?.bbox && Array.isArray(gRes.bbox) && gRes.bbox.length === 4) {
                     bbox = gRes.bbox
                   }
                 } catch (_err) {
@@ -241,15 +252,20 @@ export function registerVisionOcrTools(ctx, deps) {
             size: `${imgWidth}x${imgHeight}`,
             count: 1,
           })
-          assertBudgetAvailable(estimatedCost, cfg.dailyBudgetUsd)
-
-          const gen = await executeWithFallback(
-            providers,
-            fallbackChain,
-            seed,
-            inpaintPrompt,
-            { logger: ctx.logger },
-          )
+          const reservation = reserveSpend(estimatedCost, cfg.dailyBudgetUsd)
+          let gen
+          try {
+            gen = await executeWithFallback(
+              providers,
+              fallbackChain,
+              seed,
+              inpaintPrompt,
+              { logger: ctx.logger },
+            )
+          } catch (err) {
+            reservation.release()
+            throw err
+          }
 
           const inpaintedBytes = gen.bytes
           if (!inpaintedBytes) {
@@ -258,13 +274,23 @@ export function registerVisionOcrTools(ctx, deps) {
 
           // 5. Composite typography overlay on inpainted background
           let finalBytes = inpaintedBytes
+          let composited = false
           try {
             const sharpMod = await import('sharp')
             const sharp = sharpMod.default || sharpMod
+            const inpaintMeta = await sharp(inpaintedBytes).metadata()
+            const compW = inpaintMeta.width || imgWidth
+            const compH = inpaintMeta.height || imgHeight
+            let compOverlay = overlayBuffer
+            if (compW !== imgWidth || compH !== imgHeight) {
+              const scaledOverlaySvg = buildTextOverlaySvg(compW, compH, overlayItems, args.font_family || 'sans-serif')
+              compOverlay = Buffer.from(scaledOverlaySvg, 'utf8')
+            }
             finalBytes = await sharp(inpaintedBytes)
-              .composite([{ input: overlayBuffer, top: 0, left: 0 }])
+              .composite([{ input: compOverlay, top: 0, left: 0 }])
               .png()
               .toBuffer()
+            composited = true
           } catch (sharpErr) {
             if (ctx.logger?.warn) ctx.logger.warn(`Sharp compositing skipped: ${sharpErr.message}`)
           }
@@ -302,7 +328,7 @@ export function registerVisionOcrTools(ctx, deps) {
           })
 
           const effectiveCost = Number(gen.cost) || estimatedCost
-          recordSpend(effectiveCost, {
+          reservation.commit(effectiveCost, {
             ctx,
             meta: { provider: usedProvider, model: cfg.model || cfg.customModel, prompt: inpaintPrompt, seed },
           })
@@ -322,7 +348,7 @@ export function registerVisionOcrTools(ctx, deps) {
 
           const response = {
             summary: dualOutput,
-            replacedCount: overlayItems.length,
+            replacedCount: composited ? overlayItems.length : 0,
             bboxes,
             inpaintPrompt,
             targetLanguage: args.target_language || '',

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@goodandready/dsh-image-gen",
-  "version": "0.11.29",
+  "version": "0.11.31",
   "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",