Lore DB | Docs

Tips for Great Content

The quality of your AI answers depends directly on the quality of your documents. Here’s how to get the best results.

Write clear, descriptive titles

Titles are used as context when the AI searches. “Q4 2024 Incident Response Playbook” is much more useful than “Playbook v2 (final)”. A good title tells the AI what the document is about before it reads the content.

Use headings and structure

Documents are split into chunks for search. Well-structured content with headings helps the chunking process create more meaningful pieces. Use Markdown headings (## Section Name), bullet lists, and numbered steps.

One topic per document vs. long reference docs

Both approaches work, but for different purposes. Focused, single-topic documents give more precise search results. Long reference docs work well when the information is interconnected. A good rule of thumb: if two sections would never be relevant to the same question, they should probably be separate documents.

Keep documents up to date

Stale information is worse than no information — the AI will confidently cite outdated docs. Review important documents periodically and click the verify button to signal they’re still accurate. Delete documents that are no longer relevant.

Organize with libraries

Group documents by team, project, or topic. This makes it easy to toggle specific knowledge sets on and off, share the right docs with the right people, and export focused backups.

Example library structure

LibraryContents
EngineeringArchitecture decisions, deployment runbooks, API documentation
ProductFeature specs, user research findings, release notes
OperationsIncident playbooks, onboarding guides, vendor contracts
PersonalMeeting notes, draft proposals, quick references

Use the naming convention [System/Service] — [Specific Topic] for document titles. Examples: “Payment Service — Stripe Webhook Configuration”, “Onboarding — New Engineer First Week Checklist”.