From 07a6582ad7964199dec6e2249afcb68ed0a73062 Mon Sep 17 00:00:00 2001 From: Tori Whaley Date: Fri, 22 May 2026 15:18:07 -0400 Subject: [PATCH 1/2] docs: update verified content page for shipped AI agent features Co-Authored-By: Claude Opus 4.7 --- guides/verified-content.mdx | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) diff --git a/guides/verified-content.mdx b/guides/verified-content.mdx index ccebfd27..b3456469 100644 --- a/guides/verified-content.mdx +++ b/guides/verified-content.mdx @@ -44,6 +44,10 @@ This makes it easy for teams to distinguish between exploratory or work-in-progr Homepage with a 'Verified' tab listing all verified charts and dashboards for quick access +## How the AI agent uses verified content + +The in-product Lightdash AI agent ranks admin-verified charts and dashboards first when retrieving content for context, and its system prompt treats verified items as the canonical source of truth for company metrics and patterns. When you ask the agent a question, it will lean on verified examples and prefer fields that already appear in verified charts. + ## Verification is automatically removed when content changes Verification is tied to the exact state of the content at the time it was verified. If anyone edits a verified chart or dashboard, the verification badge is automatically removed. An admin will need to review the updated version and re-verify it. @@ -99,7 +103,6 @@ This means CI pipelines run by non-admin deployers (for example, with the `devel Today, admin-verified charts and dashboards live in a separate stream from [verified answers](/guides/ai-agents/verified-answers) (the AI-agent-generated content that users mark as verified inside an AI agent conversation). We're working on bringing them together so that any verified piece of content, wherever it was verified, is treated as a single trusted signal across: -- The verified badge in tiles, search, the resource browser, and space overviews -- Content-as-code YAML download/upload -- The [Lightdash MCP](/references/integrations/lightdash-mcp) `list_verified_content` tool -- In-product AI agent retrieval and suggestions +- The `list_verified_content` MCP tool returning AI-verified artifacts alongside admin-verified ones (today it only returns admin-verified) +- Content-as-code YAML round-trip for AI-verified content (admin-verified content already round-trips; AI-verified is the gap) +- Using verified content as templates when the AI agent generates new charts and dashboards (basic retrieval shipped, templates is the next step) From bf670d705ad146321ac049bfedef0b3f108476eb Mon Sep 17 00:00:00 2001 From: Tori Whaley Date: Fri, 22 May 2026 15:19:51 -0400 Subject: [PATCH 2/2] docs: drop parenthetical commentary from coming-soon bullets Co-Authored-By: Claude Opus 4.7 --- guides/verified-content.mdx | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/guides/verified-content.mdx b/guides/verified-content.mdx index b3456469..562a98a6 100644 --- a/guides/verified-content.mdx +++ b/guides/verified-content.mdx @@ -103,6 +103,6 @@ This means CI pipelines run by non-admin deployers (for example, with the `devel Today, admin-verified charts and dashboards live in a separate stream from [verified answers](/guides/ai-agents/verified-answers) (the AI-agent-generated content that users mark as verified inside an AI agent conversation). We're working on bringing them together so that any verified piece of content, wherever it was verified, is treated as a single trusted signal across: -- The `list_verified_content` MCP tool returning AI-verified artifacts alongside admin-verified ones (today it only returns admin-verified) -- Content-as-code YAML round-trip for AI-verified content (admin-verified content already round-trips; AI-verified is the gap) -- Using verified content as templates when the AI agent generates new charts and dashboards (basic retrieval shipped, templates is the next step) +- The `list_verified_content` MCP tool returning AI-verified artifacts alongside admin-verified ones +- Content-as-code YAML round-trip for AI-verified content +- Using verified content as templates when the AI agent generates new charts and dashboards