LLM Reference
Sovrium provides machine-readable documentation files designed for Large Language Models (LLMs) and AI coding assistants. Use these files to give AI tools full context about the Sovrium app schema, CLI, API, and configuration format.
Download LLM Documentation. Plain-text files optimized for AI tools like Claude, ChatGPT, GitHub Copilot, and Cursor.
- llms.txt ↗ — compact quick reference
- llms-full.txt ↗ — comprehensive single-file reference
What is llms.txt?
An emerging standard for providing documentation to AI assistants in a machine-friendly format.
The llms.txt convention is a growing standard where websites serve plain-text documentation files at well-known URLs (/llms.txt and /llms-full.txt). These files are specifically formatted for Large Language Models to consume, providing structured information about the project, its APIs, configuration format, and capabilities. Sovrium publishes two files: a compact quick reference for fast lookups, and a comprehensive full reference for generating configurations and answering detailed questions.
The Files
Two documentation files optimized for different use cases.
llms.txt
https://sovrium.com/llms.txt
Compact overview with links to all documentation sections. Best for quick context and navigation.
- Size: About 40 lines — fits easily in any LLM context window.
- Use case: Use when you need a quick summary or want to point an AI tool to specific documentation sections.
llms-full.txt
https://sovrium.com/llms-full.txt
Complete documentation covering the entire Sovrium platform. Best for deep AI assistance.
- Size: About 3000 lines — comprehensive, single-file reference.
- Use case: Use when you want an AI assistant to have full context about Sovrium, generate configurations, or answer detailed questions.
The full reference covers:
- Getting Started guides
- Schema Overview & Root Properties
- All 49 Field Types
- All Component Types
- Authentication configuration
- Theme & design tokens
- Pages & interactions
- Languages & i18n
- Analytics configuration
- API Reference (records, auth, admin, health)
- Complete YAML examples
- JSON Schema reference
Auto-synced with releases. Both files are automatically updated with every Sovrium release to reflect the latest schema version, field types, component types, and API endpoints. The current version is always accurate.
Usage with AI Tools
How to feed Sovrium documentation to your AI assistant of choice.
Fetch the documentation directly:
# Quick reference (~40 lines)
curl https://sovrium.com/llms.txt
# Complete documentation (~2700 lines)
curl https://sovrium.com/llms-full.txt
Paste the URL or content into your AI assistant with a prompt like:
Read the Sovrium documentation from https://sovrium.com/llms-full.txt
and help me create an app.yaml configuration for a blog with:
- A posts table with title, content, and published_at fields
- Email/password authentication
- A homepage with a list of recent posts
Which file to use? Start with llms.txt for quick questions about what Sovrium can do. Use llms-full.txt when you need the AI to generate complete configurations, understand all field types, or work with the full API.
Building Sovrium apps with AI
Because a Sovrium app is a single declarative config file, it is an ideal target for AI-assisted authoring. A practical loop:
- Give the AI context. Point it at
llms-full.txtso it knows the full schema — all 49 field types, component types, auth options, and the REST API surface. - Author in TypeScript for type safety. Have the AI generate an
app.tsusingdefineConfig()from@sovrium/types. Your editor then validates the AI's output as it writes, catching invalid field types and misshaped sections inline. - Validate before running. Run
sovrium validate app.ts(see the CLI Reference) to confirm the config decodes againstAppSchema, surfacing unknown field types and structural errors with exit code1. - Iterate against the running app. Start with
sovrium start app.ts --watch; the AI can refine the config and the server hot-reloads on save.
Type-checked output beats free-text guesses. Pairing llms-full.txt (context) with @sovrium/types (compile-time validation) and sovrium validate (runtime decode) gives the AI a tight feedback loop — generated configs are checked at author time and validate time, before they ever boot.