🚀 Updates v0.10.8: Lossless JSON Unification, Dual-Output & Strict Validation (#199)

Lossless JSON & Dual-Output Unification: upscale_image, remove_background, blend_images, and vectorize_image now consistently return toLosslessJson and formatted markdown summary for text-only LLMs.
Strict Input Image Validation: extract_design_tokens, image_to_css_gradient, and check_image_contrast explicitly validate source image readability instead of silent fallback.
Unit Testing Suite: Added test/tool-consistency.test.mjs covering tokenization, gradient generator, and PWA suite.

📦 @goodandready/dsh-image-gen

Comprehensive Visual Generation & Image Processing Suite for DeepSeek Harness

  ⭐ If you like this plugin, please star it on GitHub — it shows me that the plugin is useful to you and motivates me to keep developing it.
  

  🐛 If you find a bug or would like to request a feature, open a GitHub issue in any language — I will review your proposal and implement useful suggestions in a future plugin version.

🚀 Updates v0.10.26: Architecture Decomposition & Repository Sanitization (#235, #238, #239)

Architecture & Modular Decomposition (#239): All server modules in lib/ now strictly meet the )`.
Sensitive File Hardening: History, cache meta/data, and spend-meter writes now enforce 0600 permissions after write.
Dead Code Removal & Module Split: Removed unused listCuratedStyles; split oversized server modules — providers.js utils extracted to provider-utils.js, processing.js into processing-basic.js + processing-advanced.js. Client file split is tracked separately (#244).

🚀 Updates v0.10.23: In-App Auto-Updater, Settings Synchronization, Active Probes & Language Purity (#232)

One-Click In-App Auto-Updater: Direct 1-click update support inside the plugin Settings Card (UpdaterSection) backed by /api/dsh-image-gen/update. Fetches the npm registry for release manifests, verifies semver differences, and securely triggers dsh plugin add @goodandready/dsh-image-gen@latest restricted to loopback and private LAN connections.
1:1 Settings Synchronization: Fully synchronized host Config schema in lib/index.js with client UI fields (lib/client.js), exposing autoEnhancePrompt and defaultStylePreset under the ✨ Enhancer tab.
Active Diagnostic Probes: Added real HTTP network probes in testProviderConnection for Replicate (/v1/models), Google Gemini (/v1beta/models), and ByteDance Seedream with timeout protection and live roundtrip latency tracking.
Strict Multi-Language Compliance: Pure standard en and zh localization bundles embedded in core package; modularized Russian language support via goodandready/dsh-russian-lang per DSH plugin guidelines. Exactly 0 Cyrillic characters in product code (lib/).
WAI-ARIA Accessibility Hardening: Enhanced tab navigation and input form validation with role="tablist", role="tab", aria-selected, aria-controls, role="tabpanel", aria-describedby, and role="alert".

🚀 Updates v0.10.22: Referential Stability & React Error #185 Infinite Loop Fix (#230)

Settings Card Referential Stability: Resolved React Error #185 (Maximum update depth exceeded) in FalSettingsCardController and CardForm.prototype.bind. Store snapshots are referentially cached during render cycles (Object.is(prev, next) === true), completely preventing infinite re-render loops in React 18 / useSyncExternalStore.
Deep DSH Store Integration: Added seamless runtime integration with @deepseek-ai/dsh-client-store (runtime.createSnapshotStore) paired with a robust standalone snapshot caching fallback.
Automated Regression Test Suite: Added dedicated referential identity test in test/client-syntax.test.mjs verifying snapshot immutability across repeated reads, single-notification publish triggers, and clean listener lifecycle.

🚀 Updates v0.10.21: Variations Workbench, Smart Aspect Ratio Crop & Brand Asset Pack (#228)

Variations & Style Remix Workbench: Interactive visual drawer embedded directly in FalImageCard and Gallery. Fine-tune variations with a continuous creativity / denoising strength slider (0.05 to 0.95), preset quick buttons ("Subtle 0.25", "Creative 0.65"), and prompt modifier guidance.
Intelligent Aspect-Ratio Canvas Cropper & Resizer (smart_crop_image): Instant lossless aspect ratio framing for social and responsive surfaces (1:1, 16:9, 9:16, 4:3, 3:2, 2:3) supporting focal heuristics (auto_focus, center, rule_of_thirds, letterbox) via zero-dependency pure Node.js SVG vector raster wrapper.
Project Asset Pack & Social Card Exporter (export_asset_pack): Production-ready automated brand asset generation producing SVG favicons, PWA application icon suite (icon-192.svg, icon-512.svg), manifest.webmanifest, high-resolution OpenGraph social banner preview (og-card.svg at 1200×630), and an asset manifest catalog (catalog.json).
Dedicated Toolviews for 12 Visual Tools: Direct interactive action buttons and visual renderers in conversation sessions for all generation, editing, inpainting, remix, crop, export, and collage tools.
Strict Localization Standard: Complete zh and en UI dictionary integration; Russian locale registered via goodandready/dsh-russian-lang.

🚀 Updates v0.10.19: Sidebar Gallery, Diagnostics & Style Engine (#224)

Native Sidebar Gallery & History Drawer: Integrated directly into DSH's native right sidebar pane (sidebarRightTabs and sidebar.right.pane.tab), betterSidebar, and a quick-access utility chip in the conversation header (conversation.session.header.utilities). Browse past generations, inspect metadata, copy prompts, and export links without cluttering the chat.
Interactive Provider Diagnostics: Real-time "Test Connection" button in the Settings Card under the Providers tab. Ping Fal.ai, ComfyUI, Automatic1111, or Custom OpenAI-compatible endpoints with live roundtrip latency tracking and status badges (Ready, High Latency, Unreachable).
Curated Style Presets & Smart Prompt Polisher: Built-in artistic engine with 10 curated styles (cinematic, photorealistic, anime, minimalist_vector, isometric_3d, analog_film, cyberpunk, pixel_art, oil_painting, claymation), optional autoEnhancePrompt setting, and style_preset tool parameter.
Collage & Comparison Grid Assembler (assemble_image_grid): Stitch 2 to 4 images into clean side-by-side, 2x2 grid, or vertical collage comparison graphics via pure Node.js SVG raster rendering, zero heavy external binaries ( 1, promptsarray) andgenerate_image_pack multi-ratio workflows now execute concurrently with a concurrency limit (concurrency: 3) via asyncPool`, reducing generation wait times by up to 3x.
Zero-Delay Initial Polling: Eliminated artificial initial delay in pollStatus and queue runners (Fal.ai, ComfyUI), inspecting generation status immediately on first attempt and shaving 1–1.5s off fast generative workflows.
Sub-Millisecond L1 In-Memory Cache: Built-in 32-entry in-memory LRU cache atop disk cache L2 provides instantaneous (`) for clean unload/reload disposal.
Project meta: AGENTS.md and index.md document architecture, constraints, and the test matrix (internal workflow files — not shipped in the npm package).
Design contract: docs/design/DESIGN.md no longer advertises the removed HistoryGallery UI (#203).
Regression lock: test/lifecycle-structure.test.mjs asserts modular layout, labeled effects, and meta files. Suite: 132 unit tests.

⚡ Overview

@goodandready/dsh-image-gen is a premier graphic generation and visual processing suite for DeepSeek Harness. It equips autonomous agents with an extensible set of tools for image generation, transformation, background removal, upscaling, vectorization, multi-reference blending, and quality inspection across 8 generative backends.

🛡️ Reliability, Performance & Quality (v0.10.0)

Exponential Backoff with Jitter: Adaptive polling for FAL, Replicate, and ComfyUI queues protects against HTTP 429 rate limits.
Error Classification in Fallback Cascade: Client-side errors (Content Policy, 400 Bad Request, NSFW) fail fast without wasting API credits on other providers.
Deterministic Hash Caching: Exact matches of prompt, model, and seed return instantly from local storage with zero API expense.
ComfyUI & Automatic1111 Drag-and-Drop: Metadata is packed into PNG Parameters chunks in standard format.
Dimension Snapping: Automatic normalization to multiples of 64 guarantees VAE bucket compatibility.
Enhanced Style Presets: Built-in styles include tailor-made negative prompts and optimal guidance scale settings.

🛠️ Complete Tools Reference

generate_image: Generate images with pluggable providers, seeds, aspect ratios, and style presets.
edit_image: Targeted inpainting and modification with automatic session reference resolution (#142, #144, #145).
vary_image: Controlled variation generation preserving composition (#143, #144).
remove_background: Extract subject with transparent PNG output (FAL BiRefNet / Rembg).
upscale_image: 2x / 4x super-resolution with clarity reconstruction.
vectorize_image: Convert raster graphics to clean scalable SVG vectors with palette quantization.
blend_images: Multi-reference composition mixing.
generate_image_pack: Simultaneous multi-aspect ratio rendering with graceful partial recovery.
generate_responsive_mockups (#173): One shared product prompt rendered as Mobile (9:16), Tablet (3:4) and Desktop (16:9) UI mockups in parallel. Keeps style/palette coherent across viewports, saves mockup-mobile.png / mockup-tablet.png / mockup-desktop.png, and shows a tabbed viewport preview in chat.
compare_images: Pixel-level visual difference ratio comparison.
inspect_image_quality: Automated visual audit, Laplacian sharpness scoring, and defect detection.
extract_design_tokens: Ex

@goodandready/dsh-image-gen

インストール

ドキュメント