Going AI-native to enhance how humans and agents access ScalarDB and ScalarDL docs

Good documentation isn’t just about content — it’s also about how easily readers and their tools can find what they need. Over the past several months, we’ve shipped several AI-focused improvements to the ScalarDB and ScalarDL docs sites. This post walks through those updates.

• Ask AI interface. Conversational search powered by Google AI Mode, with pre-written questions and direct URL access.
• llms.txt and llms-full.txt. “llms.txt” follows the llmstxt.org standard, and “llms-full.txt” is a richer custom companion for AI tools and LLM integrations.
• Copy page as Markdown. Export any documentation page as Markdown for easy use in AI tools and editors.
• Chat with page. Start conversations in ChatGPT, Claude, or Perplexity with page context automatically included.
• Context7 automation. Weekly synchronization keeps the Context7 library current with the latest documentation.

Ask AI interface

Finding the right answer in technical documentation can take time, especially when you’re not sure which page to start with or are in the middle of developing an application that uses ScalarDB or ScalarDL. To help, we added an Ask AI button to the navbar of both docs sites.

Selecting the button opens a modal we built on the docs site. When you submit a question, it opens Google AI Mode with the query scoped to the relevant documentation, and the response is generated there. You can either pick from a set of pre-written questions — covering topics like getting started, developing applications, and deploying ScalarDB or ScalarDL — or type your own question in the text field.

Ask AI interface

A few details worth knowing:

• The button appears only when you’re reading the latest version of the docs.
• Both English and Japanese are supported, including localized pre-written queries for English and Japanese readers.
• You can open the modal directly by adding “?ask-ai” to any docs URL — for example: scalardb.scalar-labs.com/docs/latest/?ask-ai or scalardl.scalar-labs.com/docs/latest/?ask-ai. This makes it easy to share a link that drops someone straight into the AI search experience.

llms.txt and llms-full.txt for AI tools

As AI-powered developer tools have become more common, we wanted to make sure that the docs are easy for those tools to consume — not just for humans reading in a browser.

Both docs sites now automatically generate two files on every build:

• llms.txt: A lightweight index of titles and links for the latest version of the docs, following the llmstxt.org standard.
• llms-full.txt: A richer custom companion file with consolidated latest docs content, suitable for LLM prompt context and AI indexing.

Both files are available at the root of each site:

ScalarDB:

• scalardb.scalar-labs.com/llms.txt
• scalardb.scalar-labs.com/llms-full.txt

ScalarDL:

• scalardl.scalar-labs.com/llms.txt
• scalardl.scalar-labs.com/llms-full.txt

On the docs home page of the latest version, you can also copy llms.txt and llms-full.txt directly.

Because both files include only the latest docs version, these copy buttons are available only on the top page of the latest version.

If you’re building tooling or AI agents that need to reference Scalar product documentation, these files give you a reliable, always-current starting point.

llms.txt and llms-full.txt for AI tools

Copy page content and chat with page

To make page-level AI actions more useful, we added two related capabilities to both the ScalarDB and ScalarDL docs sites. These are available on all versions of the documentation.

Copy page as Markdown

On every docs page (except the home page), you can use a Copy page as Markdown button in the breadcrumb bar to copy the current page as Markdown, including the page structure and links.

This makes it easy to paste documentation directly into AI tools or editors for tasks like summarization, analysis, code generation, and internal notes, without manually reformatting content into Markdown.

Copy page as Markdown

Chat with page

We also recently added a Chat with page button that lets you select an AI service (currently ChatGPT, Claude, and Perplexity). After selecting an AI service, you can start a conversation through the AI service’s interface, using the page that you’re currently viewing as the starting context.

This feature is useful when you want quick, page-specific answers, troubleshooting guidance, or draft code examples without manually building prompts from scratch. As with any AI output, verify important details against the original documentation page.

Chat with page

Keeping Context7 in sync automatically

We also added a GitHub Actions workflow to refresh the Context7 library source based on llms-full.txt, with both a weekly scheduled run and a manual trigger available for one-off refreshes.

The workflow is currently configured to run once a week because we determined that a weekly refresh is sufficient. However, the schedule is fully customizable through the workflow, allowing us to increase or decrease the refresh frequency as our documentation or operational needs change.

This automation keeps the Context7 library in sync with the latest documentation on a regular schedule without requiring manual updates.

Wrap-up

These updates are part of our ongoing effort to make the Scalar docs sites more useful for both humans and AI agents. We’ll keep iterating on the experience as we get feedback and discover new ways to improve.

Try these features on the ScalarDB and ScalarDL docs sites, and let us know how they work for your AI workflows.

To learn more about ScalarDB and ScalarDL, please take a look at our other posts on the Scalar Engineering Blog and at the ScalarDB documentation and ScalarDL documentation.

If you are considering using ScalarDB or ScalarDL, please contact us.