Skip to content
CAI
Software that uses CAICheck a score

Anil-matcha/Open-Generative-AI

31.3

Weak · 15 November 2026

79.7k

lines of production code

JavaScript

primary language

5

measurements over time

CAI band scale
CAI trend line
CAI lens gauges

What this system is

Features

Add Chinese (zh) localization support

Users can now access the application in Chinese via the /zh route prefix. This change introduces a new locale registry in lib/locales.js that manages message bundles and fallback logic, merging Chinese translations with English defaults for any missing keys. The app/zh directory contains route wrappers that redirect home and assistant pages to the studio, while the studio page itself renders the existing interface with the 'zh' locale prop, enabling translated UI copy without duplicating component logic.

app/zh, lib · high confidence

Added new design agent and workflow submodules

The packages directory now includes three new Git submodules: Open-AI-Design-Agent, Open-Poe-AI, and Vibe-Workflow. These additions introduce new components for design agent functionality and workflow management into the Studio environment.

packages · high confidence

Introduce Electron desktop app with local AI inference and Wan2GP support

The application now includes an Electron-based desktop client for macOS, Windows, and Linux (Ubuntu), featuring a secure main process configuration with web security and context isolation enabled. Users can now perform local model inference via sd.cpp and utilize the Wan2GP HTTP provider as a second local engine, with the preload script exposing IPC bridges for model management, generation, and progress tracking. The desktop build includes AppImage and .deb packaging targets, along with specific Linux kernel sandbox workarounds for Ubuntu 24.04+.

electron · high confidence

Introduce local AI inference engines (sd.cpp and Wan2GP)

Users can now run image and video generation locally without external API keys. The app bundles the sd.cpp binary for local image models (including Z-Image, Dreamshaper, and SDXL variants) with platform-specific asset selection and progress reporting. Additionally, a new Wan2GP HTTP provider allows users to connect to their own Wan2GP Gradio server for Flux, Qwen, and video models, featuring automatic endpoint discovery to handle API name changes across releases.

electron/lib · high confidence

Introduction of Studio Shell and API Key Modal components

The application now includes a central \StandaloneShell\ component that serves as the main navigation container for the various AI studios (such as Image, Video, Audio, and Marketing). This shell manages the tabbed interface, allowing users to switch between different creative tools. Additionally, a new \ApiKeyModal\ component has been added to handle user authentication, providing a dedicated UI for entering and managing API keys required to access the studio features.

components · high confidence

New AI Studios and Expanded Studio Catalog

This update introduces several new studio interfaces—Agent Studio for AI agent conversations, AI Influencer Studio for detailed character customization, Apps Studio for browsing and requesting SaaS templates, Audio Studio for music generation, and Cinema Studio for cinematic image creation with camera/lens controls. It also adds the Clipping Studio for video clipping. These new components are integrated into the existing studio ecosystem alongside established tools like Image, Video, and Marketing studios, expanding the platform's generative capabilities.

packages/studio/src/components · high confidence

New Linux packaging script and local AI binary staging utilities

Added \scripts/package-linux-deb.js\ to automate the creation of Debian packages for Linux, handling dependency resolution, desktop integration, and executable wrapping. Introduced \scripts/stage-local-ai-binary.js\ to standardize the staging of local AI inference binaries (such as \sd-cli\ and associated libraries) across macOS, Linux, and Windows platforms. Also added \scripts/test\_minimax\_provider.js\ to verify the registration and live functionality of the MiniMax Image 01 model integration.

scripts · high confidence

New local inference client and i18n infrastructure

The application now supports local AI model inference via a new \localInferenceClient\ that bridges to the desktop app's \window.localAI\ interface, enabling the use of bundled \sd.cpp\ models (such as Z-Image Turbo, Dreamshaper 8, and Realistic Vision v5.1) and remote Wan2GP providers (including Flux.1 Dev, Wan 2.2, and LTX Video) directly from the client. Additionally, a new \i18n.js\ module introduces English and Simplified Chinese language support with automatic browser locale detection and migration, while \uploadProxyTarget.js\ adds strict validation for upload endpoints to block unsafe hosts and file types.

src/lib · high confidence

New studio components and UI infrastructure

This change introduces several new UI components for the application's studio features. It adds AgentStudio, AuthModal, CameraControls, CinemaStudio, Header, ImageStudio, LipSyncStudio, and LocalModelManager. The Header component now includes navigation links for image, video, lipsync, cinema, workflows, agents, and mcp-cli tabs, along with a language switcher and settings button. The AuthModal provides a UI for entering API keys. CinemaStudio includes camera controls for selecting camera, lens, focal length, and aperture. ImageStudio supports text-to-image and image-to-image generation with local AI inference. LipSyncStudio allows for lip-sync generation from images or videos. LocalModelManager handles the installation and status of local AI models. These components form the basis for the new studio interfaces.

src/components · high confidence

Behavioural changes

Application restructured with client-side routing and modular studio components

The application has transitioned from a single-page counter demo to a multi-studio interface featuring client-side navigation. The main entry point now implements a router that dynamically loads and mounts specific studio components (Image, Video, Cinema, Lip Sync, Workflows, Agents, and MCP/CLI) into a persistent content area, preserving state during tab switches. The styling has been refactored to support a full-viewport layout with flexbox, replacing the previous centered, single-column design with modular CSS imports for global and studio-specific themes.

src · high confidence

Introduce standalone shell layout for Studio and Workflow pages

The Studio and Workflow application areas now render their content within a \StandaloneShell\ component. This change establishes a consistent layout structure for the Studio interface and the Workflow detail and tab views, ensuring they share the same standalone presentation mode rather than inheriting a different global layout.

app/studio, app/workflow · high confidence

Linux desktop build support and installer improvements

This change introduces Linux (Ubuntu) desktop build support, including AppImage and .deb targets, alongside a new NSIS installer script that customizes the installation directory to LocalAppData and forces the application to close before installation. It also adds an AppArmor profile to address Ubuntu 24.04+ user namespace sandboxing issues and provides documentation for staging local inference binaries for bundling.

build · medium confidence

Migration to Next.js App Router with new visual identity and localization support

The application has been restructured to use the Next.js App Router, introducing a new root layout that handles locale detection via headers and applies a global dark theme with a cyan primary color. This update includes a rebranding of the project title to 'Open Generative AI' and adds support for Chinese (zh) localization, alongside a new glassmorphism UI style and custom scrollbar aesthetics defined in the global styles.

app · high confidence

New Agent Studio pages and hardened API proxy routes

This change introduces the Agent Studio interface in the app/api area by adding server-side pages for agent chat, creation, and editing (under /agents) that fetch agent details and user data from the backend, and by adding new API proxy routes (/api/agents, /api/app, /api/workflow, /api/api/v1, /api/v1/creative-agent, /api/v1/get\_upload\_url, and /api/v1/upload-binary) that forward requests to the external MuAPI service. The proxies now enforce API-key-based authentication via the x-api-key header, remove cookie-based auth, strip sensitive headers (including cookies and Authorization) before forwarding, and block credential logging to prevent leakage. File uploads are secured through a binary proxy that validates the S3 target URL and blocks dangerous file types before uploading. A home page now redirects to /studio.

app/api · high confidence

New design system with cyan branding and interactive UI components

The application introduces a new visual identity and component library. The primary brand color has been updated to electric cyan (\#22d3ee) with a strict dark-mode background palette. New CSS classes provide glassmorphism effects (glass, glass-panel), neon borders, and interactive glow animations for elements. A built-in tooltip system is now available via the \[data-tooltip\] attribute, supporting hover states and bottom-positioning variants. Additionally, specific styles for the image generation studio have been added, including layout and interaction styles for style cards, prompt bars, and canvas toolbars.

src/styles · high confidence

Project rebrand to Open Generative AI with enhanced security and localization

The application has been rebranded from Open-Higgsfield AI to Open Generative AI, including updated metadata, titles, and documentation. This release introduces a Next.js middleware that enforces security headers (X-Content-Type-Options, X-Frame-Options, CSP) and adds Chinese (zh) localization support. Additionally, the project now includes Docker support for containerized deployment and configures Electron builds to perform ad-hoc code signing on macOS.

(repo-wide) · high confidence

Studio adopts Tailwind CSS with new design tokens and Inter font

The studio now uses Tailwind CSS for styling, introducing a specific design system that sets the primary color to cyan (\#22d3ee) and defines secondary and muted palette tokens. The default sans-serif font family is configured to use Inter, and new utility classes for width and height (18 and 22) are added to ensure image placeholders reserve space before thumbnails load.

packages/studio · high confidence

Unified video model registry and parameter handling across all studio providers

The Studio now uses a centralized \groupedVideoRegistry\ and \groupedVideoParameters\ system to manage video model selection, variants, and input parameters for all providers (Alibaba, Kling, MiniMax, Seedance, Veo, Happy Horse, Vidu, PixVerse, LTX, Sora, and xAI). This change replaces fragmented provider-specific logic with a unified registry that handles facet options (like profile and speed), workflow variant resolution, and parameter normalization, ensuring consistent behavior and easier maintenance across the entire video generation suite.

packages/studio/src · high confidence

Fixes

Improved error handling, image download formatting, and generation polling

This update introduces three new utility modules to enhance the Studio's reliability and user experience. The new \formatError.js\ utility cleans up raw API and studio error messages by extracting readable details from JSON payloads, mapping specific HTTP codes (like 402, 401, 403, 429) to user-friendly text, and truncating long messages to prevent UI overflow. The \downloadImage.js\ utility ensures downloaded images retain the correct file extension based on their MIME type, falling back to the original filename if the type is unrecognized. Finally, \generationLifecycle.js\ provides robust polling for async generation results, handling timeouts, network errors, and authentication failures, while also appending refund notices to error messages when applicable.

packages/studio/src/utils · high confidence

Test coverage

Added unit tests for local AI inference and model availability logic

Added comprehensive test coverage for the local AI inference subsystem and Wan2GP model availability checks. The new tests verify platform-specific binary asset selection and bundled resource path resolution for Linux ARM64, path resolution logic including custom storage directory overrides and whitespace trimming, generation progress parsing with de-duplication and ANSI sequence stripping, and model availability status handling for Wan2GP endpoints including fallback to default API names when metadata is missing.

tests · high confidence

Dependencies

Project restructured as a Next.js monorepo with new studio and desktop build capabilities

The project has been renamed from 'open-higgsfield-ai' to 'open-generative-ai' and upgraded to version 2.0.0, migrating from a Vite-based setup to a Next.js monorepo architecture. This change introduces a new 'studio' package containing shared UI components (such as ImageStudio, VideoStudio, and AgentStudio) and integrates local AI agent and design agent workspaces. Additionally, Electron desktop build support is added for macOS, Windows, and Linux, enabling users to install the application as a native desktop client.

(dependencies) · high confidence

Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

How this codebase got here

This is the PUBLIC form of this artifact. Findings are listed in full, but the details of SECURITY findings — which rule fired, in which file, on which line, and how to fix it — are deliberately withheld, and any secret-scanner results are excluded entirely. Where detail is absent here it was REMOVED FOR PUBLICATION; it is not missing from the analysis. The complete artifact is available from the repository owner.

Score

  • CAI 32 → 31 (-0.6)
  • Rubric changed (rubric-2026.10.5 → rubric-2026.11.4) — scores are not directly comparable.

Lenses

  • Code Health 34 → 32 (-1.9)
  • Architecture 86 → 87 (+1.0)
  • Maturity 57 → 58 (+0.3)
  • Readiness 18 → 17 (-1.3)
  • Security 69 → 72 (+3.4)
  • Accessibility 39 → 42 (+2.3)
  • Performance 60 → 60 (+0.0)

Resolved (3)

  • No ADRs found
  • Off-boarding risk: anonymized user #1
  • Off-boarding risk: anonymized user #2

New (49)

  • Banned license: @img/sharp-libvips-darwin-arm64 (package-lock.json)
  • Banned license: @img/sharp-libvips-darwin-x64 (package-lock.json)
  • Banned license: @img/sharp-libvips-linux-arm (package-lock.json)
  • Banned license: @img/sharp-libvips-linux-arm64 (package-lock.json)
  • Banned license: @img/sharp-libvips-linux-ppc64 (package-lock.json)
  • Banned license: @img/sharp-libvips-linux-riscv64 (package-lock.json)
  • Banned license: @img/sharp-libvips-linux-s390x (package-lock.json)
  • Banned license: @img/sharp-libvips-linux-x64 (package-lock.json)
  • Banned license: @img/sharp-libvips-linuxmusl-arm64 (package-lock.json)
  • Banned license: @img/sharp-libvips-linuxmusl-x64 (package-lock.json)
  • Banned license: @img/sharp-wasm32 (package-lock.json)
  • Banned license: @img/sharp-win32-arm64 (package-lock.json)
  • Banned license: @img/sharp-win32-ia32 (package-lock.json)
  • Banned license: @img/sharp-win32-x64 (package-lock.json)
  • Medium CVE: [GHSA redacted] (package-lock.json)
  • Medium CVE: [GHSA redacted] (package-lock.json)
  • Medium CVE: [GHSA redacted] (package-lock.json)
  • Medium: security finding (details withheld)
  • Medium: security finding (details withheld)
  • Medium: security finding (details withheld)
  • …and 29 more

Changes since last survey

  • 16 commits — 15 feature/other, 1 fixes

By area

  • (root) — 11 commits
  • packages/studio — 3 commits
  • (repo) — 1 commit
  • electron/lib — 1 commit

Notable commits

  • fix: Merge pull request #395 from Darkgoatie/fix-vae-download-redirect
  • change: Add MiniMax H3 Max and Max Turbo Spicy models to studio catalog
  • change: Add new video links for Seedance and GPT Image
  • change: Promote Open Duet in README
  • change: Replace old video links with new API guide links
  • change: Replace old video links with new ones
  • change: Replace video links and titles in README
  • change: Resolve relative redirect Location in downloadFile
  • change: Update README.md
  • change: Update README.md
  • change: Update README.md
  • change: Update README.md
  • change: Update video links and descriptions in README
  • change: Update video links in README.md
  • change: feat(studio): sync enabled Muapi models
  • change: heygen-video-translate: note 120 second max input duration

Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

Survey your own repository

Anil-matcha/Open-Generative-AI was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.

About this page

  • The score is its most recent published measurement, taken on 15 November 2026 at a pinned commit. It is not a live figure and does not change until the project is measured again.
  • Measured at commit 7dd778614075f220f37bfa419a798304dcd2d2eb — the exact code this score is about.
  • Scored under rubric-2026.11.4 — the same rubric and the same method as every other entry in this index.
  • Measured by Canine Code Analyzer (cca.canine.dev) using cca measure-2026.11.65+preprod-3c33c1882dc1.