Changelog

Sankey charts are here

You can now visualize flows—funnels, transitions, attribution paths, anything with a source-to-target shape—as proper Sankey diagrams. Pick source, target, and value columns from any SQL result and Basedash renders a responsive SVG flow chart with hover tooltips, drilldown variables, and optional per-node color overrides.

The AI chat agent knows how to build Sankeys directly from edge-list SQL results, so you can just describe the flow you want and Basedash picks the right shape and wiring on its own. Sankey charts work end-to-end across dashboards, Insights, and public renders.

Insights run on your schedule

Each organization can now choose how often Insights are generated: daily, weekdays only, weekly on a chosen day, or monthly on a specific date. Pick the cadence that matches how your team actually checks in instead of having Insights arrive every morning whether you need them or not.

The new controls live in Insights settings and apply across the org. We also brushed up surrounding copy that previously described Insights as daily-only.

Insights notifications now show the chart

When an Insight publishes to Slack or email, the post now includes a 700x500 image of the actual chart rendered server-side, with a clean text-only fallback if rendering fails. You get the visual context inline without having to click through to Basedash to see what changed.

We also rolled out a new public Insights API so you can list, generate, fetch (with raw chart data), and delete Insights programmatically—useful for piping insight outputs into your own workflows or pinning specific runs into your own tooling.

Chat can build and manage automations for you

The chat agent now has full read-and-write access to automations: it can list, create, update, and delete them, and successful tool calls render as clickable automation cards in the thread that take you straight to the automation detail page. Ask in natural language to spin up a recurring report or tweak a schedule and the agent does it inline with sensible defaults, instead of peppering you with follow-up questions.

The agent is also better at proactively suggesting automations when it sees a good fit, and it now anchors new chats to the conversation initiator and their browser time zone—so the schedules and recipients it picks default to the right person and the right local time.

Fixes and improvements

• Sorted connected data sources alphabetically by display name across navigation, command menus, and selectors so they appear in a predictable order.
• Unified active members, pending invites, and deactivated members into a single Members settings list with status filtering and live name/email search.
• Surfaced pending invites in the group add-member picker so you can add invitees to groups before they accept.
• Added dashboard folders to favorites—favorited folders show up as collapsible groups in the sidebar.
• Let members with edit or manage access on a folder create dashboards inside it, instead of restricting folder-scoped dashboard creation to admins.
• Refined chart card header layout so titles and action buttons no longer compete for space, and tightened hover/loader behavior.
• Standardized secondary metric label styling across line, vertical bar, funnel, table, and number charts for a more consistent look at any width.
• Fixed chart time zone edge cases so day and month buckets and date-only values no longer shift by a day depending on locale.
• Fixed number chart label truncation so long secondary labels shrink and ellipsize while keeping the metric value fully visible.
• Rendered mutating-query thinking steps in chat with status text and the full SQL block, so you can inspect exactly what the agent is about to run.
• Switched generated Slack messages to Slack markdown blocks so tables, headers, dividers, task lists, and fenced code render cleanly instead of as raw text.
• Tightened AI guidance for number chart labels so the secondary label stays short and unit-like instead of repeating the chart title.
• Split the chat assistant job pool into user-initiated and automation queues so a fan-out of scheduled reports can’t slow down interactive chats.

Past chats are now a tap away on mobile

The mobile chat experience now has a dedicated chat history menu in the /mobile/chats header. Tap the new button to open a dropdown of your accessible past chats grouped by recency, and jump straight into any of them.

Each entry also shows the avatar of whoever created the chat, so it’s faster to spot your own threads versus ones a teammate kicked off—matching the behavior of the chat list on desktop.

Smoother line charts and a richer metric layout

Line charts now render with rounded SVG paths, a halo stroke, and a padded clip region, so lines look cleaner and don’t get cut off at the chart edges. Legends are smarter too: when metrics already convey the segment information, the legend won’t redundantly repeat it.

We also reworked how metrics show up across line, vertical bar, funnel, table, and number charts. Layouts are more consistent, secondary labels are standardized, and grouped vertical bar breakdowns now show per-segment metrics that light up the linked row when you hover. The legacy “breakdown legend” switch has been replaced by a more flexible “Label only” metric option. As a small but welcome bonus, extreme or invalid percent-change values are now capped at ±999% so KPIs never display a runaway number.

New Security setting to require Google sign-in

Admins can now require their organization’s members to sign in with Google. Flip the new toggle in Security settings and OTP and magic-link sign-in attempts will be transparently redirected back to the login page, where Google sign-in starts automatically with the user’s email pre-filled.

The setting plays nicely with your existing auth: SAML SSO still takes precedence when it’s configured, and the Google-required toggle is automatically disabled if you turn SSO on, so you can’t end up in a conflicting state.

Slack AI replies are simpler and quieter

When you mention the Basedash app in Slack, the assistant now stays on a single “Thinking” status while it works—no more noisy stream of intermediate reasoning or tool-call updates inside the thread. Once the answer is ready, Basedash posts the final response as a regular Slack message and clears the assistant status, so threads stay focused on the actual answer instead of mid-generation churn.

Fixes and improvements

• Improved truncation and scrolling behavior for metrics in number, table, and funnel charts so long values stay readable.
• Standardized secondary metric label styling across chart types for a more consistent look.
• Smoothed the sign-in flow for invited users on Google-required orgs: OTP and magic-link entry points now route them straight into Google sign-in.
• Tightened legend behavior for stacked vertical bar charts so they retain their legends where it makes sense.

AI skills are now reusable across the product

You can now create organization-level AI skills—reusable bundles of context and instructions—and have them automatically available across every AI surface in Basedash. Skills are picked up by chat, chart generation, dashboards, automations, Insights, and tasks, so you can teach Basedash about your business once instead of repeating the same context in every prompt.

Skills are managed in a dedicated section under settings, with a clean list-based UI for creating, editing, and organizing them. You can right-click any skill or use its overflow menu to rename, edit, or remove it, and only admins can create or modify skills so the shared context stays consistent for everyone in the org.

Insights got smarter and more shareable

Basedash now defaults to GPT-5.5 across the product, which makes chart generation and chat answers a noticeable step better—especially on more involved analytical questions. Insights are the most obvious place you’ll feel it, but the upgrade applies broadly to every AI-powered surface.

Insights also work harder to stay fresh. The Insights agent now sees richer context about what you’ve already received—up to 10 full recent insights and 40 older titles—so it actively avoids repeating prior topics, framings, metrics, or chart choices. You’ll get genuinely new angles on your data instead of variations on the same story.

Finally, you can do more with individual insights. The detail page has a new share popover, breadcrumb-adjacent overflow menu, and right-click context menu, and admins can now delete insights directly from the list or the detail view—so curating what shows up in your feed is much faster.

More control over team invites

Pending invites are now first-class citizens in your settings. You can change a pending invite’s role or revoke it before it’s accepted, without having to wait for the person to log in first. Pending invites are also filtered out of active member surfaces like sidebar avatars, so the people you actually work with are the ones you see, and last-admin protections only count verified active admins.

Sign-in for invited users is also cleaner. Invalid or expired magic links now redirect to the regular login page (with the email prefilled when we know it) instead of dead-ending on an error, and the underlying invite verification flow more reliably refreshes pending invites into full members for already-open clients.

Fixes and improvements

• Polished the styling of expanded thinking steps in chat so the text reads more consistently.
• Streamlined the chart generation agent’s reasoning loop so chart creation does less unnecessary work and finishes more directly.
• Tightened guidance for when chat asks for clarification versus picking up tools and acting, for a snappier feel on simple requests.

Data source setup is smoother across the board

Connecting data feels a little friendlier this week. ClickHouse now uses the modern clickhouse:// URI scheme for connection strings, with older jdbc:clickhouse:// URIs still supported so existing connections keep working without changes.

The BigQuery connection form also includes inline guidance on where to create and download a service account JSON key in Google Cloud Console, so getting your credentials in the right place is less of a scavenger hunt.

A clearer experience when a trial ends

If your organization’s trial has ended, the product no longer locks you out of places where you actually need to take action. You can still reach Profile and Billing settings, which makes upgrading, updating payment info, or managing your account straightforward instead of frustrating.

The navigation also adapts to the trial-over state. A dedicated “Trial over” item in the sidebar links directly back to the upgrade page, so the next step is always one click away even when the main nav is hidden.

Fixes and improvements

• Added a Move to folder submenu in the dashboard header’s overflow menu, right below Duplicate, so you can reorganize dashboards without opening the sidebar.

Dashboards are much easier to create with AI

The Dashboards page now gives AI dashboard creation a much more prominent starting point. Instead of feeling like a secondary workflow, you can now begin from a dedicated input right on the page, describe the dashboard you want, and jump straight into editing the new dashboard once it is created.

We also made that flow feel more guided instead of more mysterious. Empty dashboards now show a clear generating state while Basedash AI is building them, and the page includes curated dashboard templates you can use as a faster starting point when you want ideas instead of writing a prompt from scratch.

Insights are more reliable and less confusing

Insights should now show up more consistently in the background. We fixed a scheduling issue that could silently prevent many organizations from receiving their scheduled insights, and we also fixed retry behavior that could occasionally send duplicate Slack or email notifications for the same insight.

We cleaned up the Insights UI too. Header actions that only make sense on the main Insights page no longer appear on individual insight detail pages, which makes the detail view feel more focused and less cluttered.

Fixes and improvements

• Improved automation editing by giving the instructions field much more room.
• Improved automation detail pages so the active toggle and label stay aligned and easier to scan.
• Fixed dashboard creation flows so duplicate submissions are less likely while AI is still working.
• Improved chart stability when working with charts that have long version histories.

Chats are easier to share and collaborate on

Chats now use the same kind of grant-based sharing model as the rest of Basedash. That makes it much easier to control who can see or manage a chat, and it makes chat sharing feel more consistent with the way dashboards and other resources already work.

Admins can also set default access for new chats, and new organizations now start with chats shared to everyone with full access by default. For teams that use chat as a shared workspace instead of a private scratchpad, this should make collaboration feel much more natural from the start.

Insights are easier to generate on demand

You can now manually generate an insight from the Insights page whenever you want one, instead of waiting for the next scheduled run. That makes it easier to pull a fresh insight right after your data changes or when you want to actively explore a question.

We also increased the reasoning effort behind Insights. In practice, that should make Insights feel more thoughtful on harder questions and more useful when you want something that goes beyond a quick summary.

Connector setup feels smoother end-to-end

The connector setup flow got a broad UX cleanup. Forms are easier to read, descriptions render more cleanly, keyboard submission feels better, and loading states feel more polished instead of looking like unfinished placeholder UI.

We also fixed some rough edges after setup. New connectors now take you to the connector page directly, brand-new warehouses no longer look broken before their first sync, and SQL autocomplete handles names with special characters more cleanly.

Fixes and improvements

• Improved dashboard auto-refresh so large dashboards do less unnecessary work and stay more stable under frequent refreshes.
• Fixed mobile dashboards so charts render reliably on smaller screens.
• Fixed the chat composer getting stuck in “Generating…” after an AI response had already finished.
• Preserved chart version history when reopening charts from dashboards and made reverted AI versions show up correctly in the timeline.
• Added character counters and sensible limits across AI context fields so it is easier to tune context without guessing.
• Improved MCP connector OAuth setup so supported servers request better scopes and launch authorization more reliably.
• Enabled automations in embedded sidebars, with an option to hide them when they do not belong in the embed.
• Improved number chart sizing so large metric cards render more consistently.
• Improved the billing AI usage breakdown so system-generated usage is visible alongside user-attributed usage.