Going AI-native to enhance how humans/agents access ScalarDB and ScalarDL docs
ScalarDB and ScalarDL documentation sites now feature an "Ask AI" conversational search interface powered by Google AI Mode, scoped to their documentation Both sites generate llms.txt and llms-full.txt files automatically on every build, following the llmstxt.org standard for AI tool consumption A "Copy page as Markdown" feature and "Chat with page" integration with ChatGPT, Claude, and Perplexity enable direct page-level AI interactions Context7 library is kept automatically synchronized via a
Analysis
TL;DR
- ScalarDB and ScalarDL documentation sites now feature an "Ask AI" conversational search interface powered by Google AI Mode, scoped to their documentation
- Both sites generate llms.txt and llms-full.txt files automatically on every build, following the llmstxt.org standard for AI tool consumption
- A "Copy page as Markdown" feature and "Chat with page" integration with ChatGPT, Claude, and Perplexity enable direct page-level AI interactions
- Context7 library is kept automatically synchronized via a weekly GitHub Actions workflow triggered from llms-full.txt
- The updates reflect a broader industry shift toward making technical documentation AI-agent-friendly alongside human-readable formats
Why It Matters
This represents a practical implementation of the growing trend toward AI-native documentation, where technical docs are structured not just for human readers but also for seamless consumption by LLMs and AI agents. For AI practitioners and developer tool builders, these patterns — particularly llms.txt generation, Context7 synchronization, and direct AI service integrations — serve as a replicable blueprint for making any documentation site agent-ready.
Technical Details
- Ask AI Interface: A navbar button opens a modal that launches Google AI Mode with queries scoped to the relevant documentation site. Supports both English and Japanese, including localized pre-written questions. Accessible via direct URL parameter (?ask-ai) for shareable links.
- llms.txt and llms-full.txt: Auto-generated on every build. llms.txt follows the llmstxt.org standard as a lightweight index of titles and links. llms-full.txt is a richer custom companion containing consolidated latest-version documentation content, suitable for LLM prompt context and AI indexing.
- Copy page as Markdown: Available on all docs pages (except home), this button copies the current page as structured Markdown including page hierarchy and links, enabling direct pasting into AI tools without manual reformatting.
- Chat with page: Integrates with ChatGPT, Claude, and Perplexity, automatically injecting the current page's content as conversation context so users can start AI-assisted troubleshooting or code generation without building prompts manually.
- Context7 automation: A GitHub Actions workflow refreshes the Context7 library source from llms-full.txt on a weekly schedule, with manual trigger support for one-off updates, ensuring AI tools with Context7 access stay current.
Industry Insight
- The llms.txt/llms-full.txt dual-file approach demonstrates an emerging best practice: providing both a lightweight index for discovery and a consolidated content dump for context injection, which other documentation platforms should adopt.
- Direct integrations with third-party AI services (ChatGPT, Claude, Perplexity) rather than building proprietary chat interfaces reduce friction and meet users where they already work, a strategy that could accelerate AI tool adoption across developer documentation.
- Automated synchronization pipelines like the Context7 workflow highlight the importance of keeping AI-accessible documentation versions tightly coupled with source docs, as stale AI context is a common failure point in agent-based workflows.
Disclaimer: The above content is generated by AI and is for reference only.