Registry

Subscribes to lab lifecycle events, collects per-student behavior

Issues and validates student session tokens for the WolfCTF platform.

Runs per-student lab containers on a single Docker host and exposes a

The Go backend server responsible for orchestrating real-time sessions, managing characters, knowledge bases, and tasks. It acts as a gateway, routing requests to various AI inference services and integrating with LiveKit for WebRTC.

Provides adapters for various Automatic Speech Recognition (ASR) models, allowing the core inference service to use different ASR providers (e.g., Qwen ASR, Whisper) through a unified interface.

Provides adapters for real-time avatar generation models, specifically FlashHead and SoulX-LiveAct. This capsule integrates these complex model implementations into the core inference service.

Provides Dockerfiles, Docker Compose configurations, and Nginx configurations for deploying the CyberVerse application. This capsule serves as a starting point for setting up the production environment.

The Vue.js single-page application (SPA) that provides the user interface for interacting with digital humans. It includes character management, chat UI, settings, and WebRTC integration. It's a template because the UI/UX, character data, and specific landing pages are bespoke.

The core Python inference service that provides a unified gRPC interface for various AI capabilities. It dynamically registers and dispatches requests to different AI model plugins (ASR, LLM, TTS, Avatar, VoiceLLM, RAG).

Provides Go-based integration with LiveKit, a WebRTC platform, for managing real-time rooms, generating access tokens, and handling media streams. This capsule abstracts LiveKit interactions for the Go orchestration server.

Provides adapters for various Large Language Model (LLM) providers, enabling the core inference service to interact with different LLMs (e.g., OpenAI, Qwen) through a consistent interface.

Defines the gRPC service contracts and generates client/server stubs for communication between the Go orchestration server and the Python inference services. This capsule provides a standardized interface for AI capabilities.

Provides a Retrieval Augmented Generation (RAG) engine adapter, allowing the core inference service to integrate knowledge retrieval capabilities for enhancing LLM responses.

Provides adapters for various Text-to-Speech (TTS) models, allowing the core inference service to generate speech using different providers (e.g., OpenAI TTS, Qwen TTS) through a unified interface.

Provides adapters for voice-enabled Large Language Models, including specific integrations for Doubao and Qwen Omni, and a Persona Agent framework for complex conversational flows. This allows the core inference service to handle real-time voice interactions.

Manages direct WebRTC peer-to-peer connections within the Go backend, including TURN server setup, RTP packet handling, and media processing. This capsule provides the low-level WebRTC capabilities for direct streaming.

Defines standard GitHub Actions workflows for automating continuous integration, dependency updates, code quality analysis (CodeQL), and supply-chain security scanning (Scorecard) for a Node.js project. This capsule provides a template for setting up a robust CI/CD pipeline.

Provides a flexible HTTP request logging middleware for Node.js applications, supporting various predefined and custom logging formats, and a mechanism for defining custom tokens to extract request/response data. This capsule encapsulates the core functionality of the `morgan` library.

Provides the development and testing environment for the `ext-http-request-logger` capsule, including linting configurations, test suites, and test fixtures. This capsule serves as a starting point for setting up a local development workflow.

Manages per-room AI chat sessions, including message history persistence and hourly rate limiting, using Cloudflare Durable Objects and Cloudflare AI Gateway. It provides an API endpoint for external services to interact with the bot.

Provides the necessary configuration and scripts for building, deploying, and running the Next.js application as a Cloudflare Worker using OpenNext.js and Wrangler. It includes GitHub Actions workflows and Cloudflare-specific configuration files.

Provides consistent code style and quality checks using ESLint and Prettier, integrated with Husky pre-commit hooks to enforce standards before commits.

A Next.js application template providing the full user interface for the chat application, integrating various backend services and reusable UI components. It includes all pages, main UI layouts, styling, and branding.

Provides core real-time communication logic and generic UI components for audio/video calls and participant management using Cloudflare RealtimeKit. It encapsulates the integration with the RealtimeKit SDK and offers hooks and types for building real-time features.

Provides API endpoints for creating and joining chat rooms, issuing RealtimeKit participant tokens, and persisting room metadata in Cloudflare KV. It handles room type negotiation, expiry, and initial rate limiting.

A Cloudflare Scheduled Worker that periodically cleans up expired chat rooms from Cloudflare KV and deactivates corresponding RealtimeKit meetings to prevent resource leakage.

Provides a React component to integrate Cloudflare Turnstile for bot verification, wrapping the application content. It handles the client-side rendering of the Turnstile widget and stores the token in session storage.

Bundles all development and build configuration files for a Next.js project, including ESLint, Prettier, TypeScript, package dependencies, and environment variable examples. This capsule provides a standardized setup for developing and building a LiveKit-enabled Next.js application.

Contains various image assets primarily used for GitHub documentation, README files, and potentially other project branding. These assets are specific to the LiveKit Meet project's visual identity.

Defines GitHub Actions workflows for continuous integration (CI) and continuous deployment (CD). This includes workflows for running tests, linting, formatting, and syncing branches to production environments.

Implements the application's landing page and room entry points, providing UI for users to start a demo meeting, connect to a custom LiveKit server, or join a specific room. This capsule is designed as a starting point for a video conferencing application's user flow.

Provides a backend API endpoint for generating LiveKit participant access tokens. This capsule handles authentication logic and token creation, which is a critical part of any LiveKit application.

Provides a collection of client-side utility functions and types essential for interacting with LiveKit, including E2EE passphrase handling, random string generation, LiveKit URL manipulation, and performance optimization hooks.

Provides backend API endpoints to start and stop LiveKit room recordings. It integrates with LiveKit's Egress functionality to capture room content and store it in an S3-compatible bucket.

Offers a suite of reusable React UI components and associated styles for common LiveKit features, such as camera and microphone settings, keyboard shortcuts for media control, a recording indicator, and a debug overlay.

Provides the foundational Next.js application structure, including the global layout, metadata, and base styles. This capsule serves as a starting point for a Next.js project, defining the overall look and feel.

A collection of static image files intended for use as virtual backgrounds in video conferencing applications. These are generic stock images that can be reused.

Encapsulates the core client-side logic and UI components for a LiveKit video conference. This includes handling room connection, E2EE setup, and rendering the main conference view, providing a robust foundation for any LiveKit video application.

Registers all backend API routes for the SceneMint application, connecting specific URL paths to their respective handlers from other backend capsules. It acts as the central routing configuration for the HTTP server.

Manages the overall Go application lifecycle, configuration loading, logger setup, and serves static frontend assets. It acts as the orchestrator, integrating various backend services and exposing the main HTTP server.

Provides core functionality for AI image generation and prompt enhancement. It interacts with an upstream AI API, handles image uploads, task submission, and polling for results, as well as prompt enhancement via streaming.

Manages user credit quotas for image generation. It provides functionalities for checking current balance, daily check-ins to grant credits, and spending credits, using a local bbolt database for persistence.

Provides a set of reusable security middleware for an Echo HTTP server, including secure headers, body limits, source origin validation, and CSRF protection. It ensures that API endpoints are protected against common web vulnerabilities.

The main frontend application for SceneMint, built with Astro and React. It provides the user interface for image generation, prompt enhancement, quota management, and displays image history. It integrates with the backend API services.

Provides a collection of generic, reusable UI components based on shadcn/ui and Tailwind CSS. These components form the visual building blocks for the frontend application, offering a consistent design system.

Tiny capsule used to smoke-test the L3 `capsule push` flow. Lives in

Single source of truth for "is this request the site admin?". Implements

Password-protected single-page admin panel for editing all site

Cloudflare Pages deployment adapter. Owns the wrangler.toml template,

Single source of truth for all editable site content of yingjieliartist.com:

Stores and serves artwork images for yingjieliartist.com. Uploads land

The public, read-only frontend of yingjieliartist.com. Renders the