storytold/artcraft-services

This repository contains ArtCraft's backend, HTTP API, background workers, and web frontends. It is a Rust and TypeScript monorepo with shared API definitions, provider clients, database queries, and development tooling.

The Tauri desktop application is maintained in storytold/artcraft, along with the product overview, feature demos, and desktop downloads.

storyteller-web is the main HTTP service, built with Rust, Actix Web, and Tokio. It handles authentication, accounts, media uploads and libraries, generation requests, job status, credits, and Stripe billing. The service retains the storyteller-web name, and its hosted API uses https://api.storyteller.ai.

Generation spans the HTTP service, provider integrations, and asynchronous workers:

flowchart LR
  Clients[Web, desktop, and API clients] --> API[storyteller-web]
  API --> Router[artcraft_router]
  Router --> Providers[Generation providers]
  Providers -->|Webhooks| API
  Workers[Background workers] -->|Poll jobs| Providers
  API --> DB[(MySQL)]
  Workers --> DB
  API --> Storage[(Object storage)]
  Workers --> Storage

A typical generation request follows this path:

1. The API authenticates the caller, validates the request and input media, and checks the user's access and credits.
2. The generation pipeline calculates the cost, bills the wallet, and uses artcraft_router to build and submit a provider-specific request. The API records inference jobs in MySQL and returns job tokens to the client.
3. Completion is handled through provider webhooks or polling workers, depending on the integration. Results are downloaded into object storage, registered as media files, and associated with the completed jobs. Separate workers handle follow-up processing such as video thumbnails.
4. Clients poll job-status endpoints and load the resulting media through CDN URLs.

Storage responsibilities are split across these components:

HTTP routes and handlers live in storyteller_web/src/http_server. MySQL queries belong in the shared mysql_queries crate so that handlers, workers, and CLI tools use the same data access layer. Reusable billing components live under crates/service/plugins, and background services live under crates/service/job.

The API exposes two generation interfaces:

• Application endpoints: /v1/omni_gen/generate/* uses user sessions for image, video, audio, mesh, and splat generation. /v1/omni_gen/models/* and /v1/omni_gen/cost/* expose model discovery and cost estimation without requiring a user session. Application job status is available under /v1/jobs.
• Programmatic endpoints: /v1/omni_api uses an API key in the Authorization header. It provides image and video generation, image/video/audio uploads, and job-status polling. Use Authorization: Bearer <api-key>; API access must be enabled for the account.

For example, a programmatic video request goes to POST /v1/omni_api/generate/video. The response contains an inference_job_token, which the caller polls with GET /v1/omni_api/job_status/job/{token}. The Omni API guide covers authentication, request and response bodies, URL inputs, and runnable examples.

Start with these sources when adding or tracing an endpoint:

• Route registration maps URLs and HTTP methods to handlers for generation, media, users, billing, API keys, and other service areas.
• API definitions contain shared Rust request, response, and error types.
• Rust API client and TypeScript API library provide client implementations.
• Provider clients implement upstream HTTP integrations; the generation router adapts generation requests and cost estimates across providers.

frontend contains the Nx workspace for React and TypeScript apps and shared libraries. The main web frontends use Vite, with Zustand and signals for state, Three.js for 3D scenes, and shared UI and generation tools.

The web apps call the backend through the shared API library, which handles JSON and multipart requests and session credentials. Libraries such as tauri-api and tauri-utils remain because shared web components still import their types, helpers, and browser-compatible behavior. The native desktop app and libraries used only by that app live in the separate desktop repository.

artcraft-services/
├── crates/
│   ├── service/web/       # HTTP services, including storyteller_web
│   ├── service/job/       # Provider workers, media processing, analytics
│   ├── service/plugins/   # Shared billing and service components
│   ├── api_clients/       # ArtCraft API types, clients, router, provider clients
│   ├── schema/            # Database access, public tokens/enums, bucket paths
│   ├── lib/               # Shared Rust utilities
│   └── cli/               # Development and operations tools
├── frontend/
│   ├── apps/              # Web frontends
│   └── libs/              # Shared TypeScript libraries
├── _database/             # SQL migrations, materialized schemas, search schemas
├── _docs/                 # Setup guides and technical documentation
├── _tools/postman/        # HTTP request collections
├── build/                 # Service Dockerfiles
├── script/                # Development, build, and database tooling
└── Cargo.toml             # Rust workspace

Use Rust/Cargo for backend work and Node.js/npm for the main frontend workspace. See the development setup guide for toolchain setup and frontend README for dependency installation and Nx usage.

The server needs a migrated MySQL database, Redis, Elasticsearch configuration, object storage, and credentials for the integrations being exercised. The server setup guide covers local MySQL and Redis; the remaining configuration is defined in the server config directory and startup code.

In development, the server loads storyteller-web.common.env, storyteller-web.development.env, and storyteller-web.development-secrets.env from its configuration search paths: the repository root, ./config, and the server's config directory. Its bootstrap skips the root .env file.

With the toolchain and service configuration in place, run from the repository root:

SQLX_OFFLINE=true cargo check -p storyteller-web
SQLX_OFFLINE=true cargo run -p storyteller-web

The default bind address is 0.0.0.0:12345, configurable through BIND_ADDRESS. GET /_status exposes the service health check. Provider polling and thumbnail processing require their corresponding worker processes and configuration.

SQLX_OFFLINE=true uses the checked-in .sqlx query metadata during compilation; the running server still needs its databases. When changing SQLx queries, use script/rust/sqlx_codegen_database.sh to regenerate metadata against migrated development databases.

To run the browser app against a local backend:

cd frontend
npm install
VITE_USE_LOCAL_API=true npx nx dev artcraft-webapp

The browser app runs at http://localhost:4201. VITE_USE_LOCAL_API=true selects http://localhost:12345; without that override, its Vite development proxy targets the hosted API. API host selection lives in StorytellerApiHostStore.

From frontend, build the web app with npx nx build artcraft-webapp. The product website is a standalone Next.js app (not part of the Nx workspace); run it from the repository root with ./script/website/unix_frontend_website_dev.sh.

The website runs at http://localhost:3000. Repository-root launchers are in script/website. For desktop development, use the ArtCraft desktop repository.

• Technical documentation
• Code conventions
• Database schemas and migrations
• Postman collections
• Build and development scripts
• License