Release 0.12.0: Tables that scroll Dbt Labs released dbt Charts 0.12.0, which makes wide tables scroll horizontally with pinned leading columns via style.column_layout.freeze and adds a unified style.key setting for legends and direct labels. The release also adds style.conditional value-based mark styling across bar, line, area, scatter, pie, heatmap, map and point-map charts, a new funnel chart type, and warehouse-sourced column types; dbt Labs says boards now compile nearly twice as fast and render about twice as fast in dct render, dct serve and the MCP tools. dbt Charts 0.12.0 is about the details that decide whether a board gets read. Tables now hold wide and long results without squeezing them, charts can restyle marks by their values, and legends and direct labels share one style.key setting. Funnels show the stages of a process, while automatic outlier bars keep one huge value from squeezing everything else. This release also makes results more trustworthy. Column types come from the warehouse, queries run in the board’s time zone, and line charts stop drawing across gaps in the data. Boards compile nearly twice as fast and render about twice as fast in dct render , dct serve and the MCP tools. Here’s what’s new. Tables that scroll A table wider than its card now scrolls sideways instead of squeezing every column, and style.column layout.freeze pins its leading columns so row labels stay in view. Long tables scroll their rows under a pinned header, with a pager below that follows the scroll. Thanks to Tracy Chow https://github.com/fivetran-tracychow for requesting horizontal scroll https://github.com/dbt-labs/dbt-charts/issues/83 and frozen columns https://github.com/dbt-labs/dbt-charts/issues/84 . Sample account data, shown in an HTML export from dbt Charts 0.12.0. Reading a table is easier, too: - Hover a cell cut short with an ellipsis to read its full value, and click to keep it open for copying request 93 https://github.com/dbt-labs/dbt-charts/issues/93 . - Columns marked list: true show each value on its own line, and live boards open a row in place from its chevron. - Readers can sort and filter pivot tables from the column header menu request 101 https://github.com/dbt-labs/dbt-charts/issues/101 . - Query columns whose names start with an underscore stay out of the table, so a query can carry helper columns for sorting or styling. - A result cut by the row or size limit says so in the pager. - HTML exports sort, filter and page their tables in the browser, like a live board. Thanks also to Tracy for requesting full cell values https://github.com/dbt-labs/dbt-charts/issues/93 and pivot-table sorting https://github.com/dbt-labs/dbt-charts/issues/101 . See tables and text https://docs.dbtcharts.com/charts/tables/ for scrolling, frozen columns and table styling. Legends and direct labels style.key is now the one place to choose how a chart identifies its series. Set key.type: legend for a legend, key.type: direct for labels beside the series, or key.visible: false to hide the key. Existing legend and endpoint labels settings migrate automatically when a board loads. Conditional styling style.conditional restyles a mark by the value in its own row, with no color channel to set up. It works on bar, line, area, scatter, pie, heatmap, map and point-map charts. On a line or area, a run of matching points restyles that stretch of the stroke or fill. See the YAMLa example queries: margin: columns: region, margin values: - North, 0.18 - South, 0.07 - East, -0.04 - West, 0.12 - Central, -0.09 formats: margin: percent charts: margin: type: bar title: Operating margin by region query: margin x: region y: margin style: orientation: vertical conditional: margin: - lt: 0 marks: bar: color: " c62828" rows: - margin Tables get the same idea with style.row.conditional : any column’s value can restyle its whole row or another cell in it, and a column’s own rules move to its conditional: entry under style.columns . KPI cards move their conditional formatting to style.conditional as well. Existing boards migrate on load. See conditional styling https://docs.dbtcharts.com/charts/bar-charts/ conditional-styling for the rules each chart type accepts. Funnel charts type: funnel draws the stages of a process as centered bars, first to last, with ribbons joining each stage to the next. Each bar shows what remains at that stage, and each ribbon shows the share lost since the stage before. Label templates can switch either one to a share of the first stage. See the YAMLa example queries: stages: columns: stage order, stage, accounts values: - 1, Visited pricing, 4200 - 2, Started trial, 3100 - 3, Built a board, 1900 - 4, Shared a board, 1240 - 5, Upgraded, 740 charts: trial: type: funnel title: Trial accounts by stage query: stages x: stage y: accounts sort: by: stage order rows: - trial See funnels https://docs.dbtcharts.com/charts/bar-charts/ funnels for label templates and theme styling. Off-the-chart bars One huge bar can make every other value look the same. style.outliers now detects a vertical bar at least 12 times the median bar and lets it run off the card, keeping the smaller bars readable. Set style.outliers.median multiplier to change the threshold, or set it to none to turn off automatic detection. It also works with a fixed axis. The fuel chart below uses a 0 to 60 scale in megajoules per kilogram. Hydrogen goes past that maximum and runs up through the header to the chart’s top edge, ending in a ragged paper tear with its value set upright inside the bar. Hover over hydrogen to see the bar at its true length, straight if it fits and wound into a spiral if it doesn’t. See the YAMLa example queries: fuel: type: values columns: item, megajoules values: - Wood, 18 - Coal, 24 - Ethanol, 30 - Diesel, 46 - Gasoline, 46 - Methane, 56 - Hydrogen, 142 charts: fuel: title: Energy per kilogram subtitle: Hover over hydrogen to see how far it really goes. query: fuel height: 460 type: bar x: item y: megajoules style: orientation: vertical axis y: scale: continuous: domain: 0, 60 rows: - fuel With style.outliers.variant: torn with cap , a long bar is cut at a break instead and its labeled cap floats above the plot, so the smaller bars keep a readable scale. Line charts whose y-axis leaves out zero can say so with style.axis y.grid.torn baseline , which tears the chart’s bottom edge like paper. Filter controls - Multiselect pills: selected values appear as pills you can remove one at a time. - Button rows: input: radio draws every option as a row of buttons, and style.variables.presets display: buttons does the same for date-range presets. - Slider preview: hovering or dragging a slider shows the value it would pick. - Sort by clicking: a filter’s sort button cycles the order with each click and shows the order in force, instead of opening a menu requested by Tracy Chow https://github.com/dbt-labs/dbt-charts/issues/99 . Results you can trust - Warehouse types: charts take date, time and number column types from the warehouse’s declared result schema instead of guessing them from the values. - Time zones in SQL: warehouse queries run in the board’s regional.time zone , so SQL date truncation matches the times the board prints. - Gaps in lines: line charts break across a large gap in the data instead of drawing a straight line through it. - Column case: y: revenue reads a Revenue column, and two columns that differ only in case raise an error instead of guessing. Warehouse sources keep a quoted alias’s spelling, such as Q3-25 . - Unit checks: a chart warns when its title names a unit that contradicts its compacted K, M, B or T value axis. - Empty results: a chart with no rows shows a quiet “No results” card, and a KPI with no value shows a dash. Faster boards Compiling a board is nearly twice as fast, and boards render about twice as fast in dct render , dct serve and the MCP tools. Thanks to Francisco Calderón https://github.com/frcalder for reporting the warm-load slowdown https://github.com/dbt-labs/dbt-charts/issues/88 . Large tables render faster and no longer make a board lag on hover, and boards with several KPI cards render faster. dct serve shows a board’s navigation and a loading state at once while slow queries run. Others - Board sidebar: dct serve shows a collapsible sidebar of the boards in a folder and keeps your filters when you switch between them. Thanks to Greg https://github.com/gskoff for requesting the sidebar https://github.com/dbt-labs/dbt-charts/issues/97 . - Positional channels: a chart can leave out x , y and its other channels. The query’s columns bind by position, dimensions first and the measure last. - Axis labels above gridlines: axis y.labels.position: above gridline sets value labels just above their lines; the paper theme uses it. - Overlay legend warning: line and area charts warn when an overlay legend covers a line or point. - Layers on multi-measure charts: charts with y: ... accept layers: . - Layer labels and tooltips: line, area and scatter layers can label just their first, last, lowest or highest points. A layered chart also accepts tooltip: columns on its base marks. - Bubble sizes: style.size sets the mark-area range and scale type on scatter and bubble map charts. Thanks to Tracy Chow https://github.com/fivetran-tracychow for requesting it https://github.com/dbt-labs/dbt-charts/issues/98 . - Spark bars: spark.value adds a cell’s value beside its bar, and spark.background adds a background track. - Links and band labels: chart links can use any column in the clicked row, and style.marks.text.max width controls how a band’s label wraps. - Comment highlights: themes set comment highlight tints with style.comments.highlight and style.comments.highlight new . - Markdown pages: the Markdown extension takes a variables option, no longer prints a page’s frontmatter, and its charts show tooltips and hover emphasis. dct runtime-js writes the hover script for other static sites. Markdown boards default to a wider 90-character line. - Query caches: a query that reads {{ queries.X.cache }} can set dialect: and be written in your warehouse’s SQL. dct query --cache PATH shares a cache file across runs, so upstream queries execute once. - Date filters on UTC columns: filter date range takes is naive utc=true for a naive column that stores UTC, replacing the timestamp= option. - Platform from the terminal: dct cloud comments lists, adds, replies to, edits, resolves and reopens board comment threads. Use --review for a review’s threads and --mention to notify people and teams. dct cloud project settings and dct cloud project set read and change project settings, and dct cloud project connect github connects a repository without a browser step when gh is signed in. - Import and migration: the migration skill finds a tile’s query in the warehouse’s query history, screenshots a dashboard page the import could not render, and walks you through putting API keys in a .env file. The Omni importer https://docs.dbtcharts.com/import/omni/ keeps going past tiles it cannot plan, and Sigma imports https://docs.dbtcharts.com/import/sigma/ carry grouping and tooltip columns. - Diagnostics: dct doctor warns when profiles.yml is newer than the dbt manifest, and a missing-relation error on a ref or source query suggests running dbt parse . Warehouse permission and credential failures say to fix access rather than the SQL. A few community-reported fixes are worth calling out, too: - Wide unpinned tables: columns keep readable widths and cell text no longer paints over the next column. Thanks to georgea-tk https://github.com/georgea-tk for the report https://github.com/dbt-labs/dbt-charts/issues/78 . - YAML features: the docs and the chart skill say which YAML features board files accept, and which unquoted values YAML reads as booleans or numbers. Thanks to Josh Andrews https://github.com/jrandrews for asking https://github.com/dbt-labs/dbt-charts/issues/90 . - Text rules and spark bars: a --- rule spans the whole card, and a spark bar with a value of 0 draws an empty cell. Thanks to Tracy Chow https://github.com/fivetran-tracychow for reports 91 https://github.com/dbt-labs/dbt-charts/issues/91 and 95 https://github.com/dbt-labs/dbt-charts/issues/95 . The full release notes https://github.com/dbt-labs/dbt-charts/releases/tag/dbt-charts-v0.12.0 include the remaining fixes and technical changes. Upgrade uv tool upgrade dbt-charts Older YAML syntax migrates automatically when a board loads, and dct migrate saves those updates to your files. The main changes that need a manual edit are: - KPI conditional formatting moves to style.conditional and its background to style.background . A default rule now applies only when no other rule in its column matched. A rule that set a glyph or a headline tone no longer migrates; remove it and let good: pick the glyph. - Nonlinear scales: a pow , sqrt or symlog axis on a line, area, bar or histogram chart is now an error instead of quietly drawing linear. Use a scatter, or remove the scale type. - Date filter option: filter date range ..., timestamp=... now raises an error. It lives in SQL, so no migration rewrites it: use is naive utc=true for a naive column that stores UTC, and otherwise drop the argument. The moves to style.key , table column conditional: rules, and constant colors such as style.marks.bar.color migrate automatically. If you maintain custom templates that read charts.series label.font or endpoint labels.label offset , update them to read key.direct . Previously emitted board artifacts must be regenerated with dct artifact emit . A few defaults changed, each with a setting to keep the old behavior: tables scroll sideways style.column layout.overflow: fit , table rows scroll under a pager style.pagination.mode: pages , underscore columns are hidden style.hide underscore columns: false , line charts break at gaps style.marks.line.connect: true , and boards no longer draw a rule above the footer style.footer.divider.visible: true . The removed options.label column setting is no longer used. For query-backed dropdowns, options.column selects the value column and options.count column explicitly enables counts. Try 0.12.0 https://github.com/dbt-labs/dbt-charts/releases/tag/dbt-charts-v0.12.0 , bring a board to office hours, start a thread in the dbt Community Slack https://getdbt.slack.com , or open an issue https://github.com/dbt-labs/dbt-charts/issues . We look forward to seeing what you make with it.