Codex Config Checker

A static, browser-only configuration validator, linter, formatter, and converter. The unified workbench checks JSON, YAML, or TOML against a tracked Codex schema or a JSON Schema you upload.

Features

• Validates TOML with Taplo WebAssembly and the selected Codex release schema
• Gets the current stable Codex schema from OpenAI's documentation and release schemas from exact GitHub release assets named config-schema.json
• Selects the latest stable or alpha schema, or searches a scrollable archive of older release versions
• Validates JSON, YAML, and TOML against JSON Schema drafts 4, 7, 2019-09, and 2020-12
• Supports internal $ref values or an uploaded local schema bundle, without network fetching
• Runs configurable format-specific lint rules with Off, Info, Warning, and Error severities
• Explains every finding with its precise location, reason, suggested fix, actual value, expected value, data path, and schema path when available
• Highlights unknown keys in red, keys under the wrong table in orange, and wrong value types in green
• Offers one-click migration fixes for supported removed or renamed Codex keys
• Highlights complete editor lines and filters the Problems view by severity or source
• Formats JSON and YAML with Prettier and TOML with Taplo
• Includes 20 curated Rainglow editor colour themes, split evenly between dark and light presets
• Imports and exports lint settings as JSON
• Upload, paste, copy, download, format, validate, and clear controls
• Downloads a successfully validated configuration as JSON, YAML, or TOML
• Responsive layout with System, Light, and Dark themes
• Keeps configuration and schema text entirely inside the browser

Using the workbench

1. Choose a program and use Select Version to load its latest stable, latest alpha, or an older tracked schema.
2. Paste a configuration or upload a .json, .yaml, .yml, or .toml file.
3. To use your own schema instead, upload a primary JSON Schema. Version selection remains locked until that custom schema is removed. For multi-file schemas, upload dependency files and select Uploaded local bundle.
4. Choose Validate. Ordinary typing does not trigger validation; Enter, pointer movement, and editor blur do.
5. After a successful validation, open Download and choose JSON, YAML, or TOML. Editing the configuration requires validating it again before download.

Remote schema references are deliberately blocked. Local dependencies are matched by filename or declared $id. Configuration files are limited to 2 MiB; schema bundles are limited to 50 files and 10 MiB.

Local development

Requires Node.js 24 or newer.

npm ci
npm run dev

Run all checks:

npm run check
npm run build

Schema updates

scripts/sync-schemas.mjs downloads the current stable schema from https://learn.chatgpt.com/docs/config-schema.json. It also reads Codex GitHub releases and downloads every exact release asset named config-schema.json, including alpha releases and older versions. It writes only verified JSON and records each source, release version, SHA256 digest, and successful synchronization time in public/schemas/manifest.json.

The Sync Codex schemas workflow runs this check every 30 minutes. Meaningful schema or release metadata changes are committed to main, which starts the separate tested GitHub Pages deployment workflow.

Each manifest entry records its own syncedAt timestamp. The website shows this as the last time that exact schema was successfully copied from OpenAI and merged into this repository.

Optional visitor analytics

The site supports Google Analytics 4 for visitor and page view metrics. Create a GA4 property and web data stream, then copy its measurement ID, such as G-AB12CD34.

In this GitHub repository, open Settings, Secrets and variables, Actions, then Variables. Add a repository variable named GA_MEASUREMENT_ID containing that measurement ID and manually run the Pages workflow, or push a new commit.

The measurement ID is public configuration rather than a secret. When the variable is absent or malformed, analytics is disabled. When it is configured, the Google tag is not loaded until a visitor explicitly allows analytics. Only standard page visit data is sent. Configuration text, uploaded filenames, schemas, diagnostics, validation results, and formatting results remain local.

Deployment

Every push to main runs linting, type checks, browser tests, workflow tests, and the production build before deploying dist to GitHub Pages.

The production site is configured for:

https://superpauly.github.io/Configurex/