Changelog

Adds a `category` column to the `_datasets_meta` schema via a new migration,

threaded through the full Rust core → FFI → SDK → Desktop stack. The

ChartDatasetGenerator now tags AI-derived datasets with category "ai-generated"

when calling ImportFromContentAsync. The SDK's IDatasetService and DatasetInfo

types expose the new optional category field with backward-compatible defaults.

Rust changes: schema migration, DatasetMeta.category field, store CRUD/mutations

updated to read/write/copy category, new set_category() helper. FFI request

structs (CreateEmptyRequest, ImportContentRequest) accept optional category.

SDK/Desktop: DatasetInfo gains Category property, IDatasetService methods

CreateEmptyDatasetAsync and ImportFromContentAsync accept optional category

parameter, DatasetService.Mutations threads it to FFI JSON requests.

Version bumps: SDK 1.55.0 → 1.56.0, Desktop 1.55.5 → 1.56.0.

Instead of referencing the original dataset for chart aggregation (which

fails when the AI uses column names that don't match the underlying table),

the orchestrator now creates small generated datasets from sample data.

For each chart the AI suggests:

1. Aggregates the sample rows client-side using the specified x/y/group

columns and aggregation function (sum/avg/count/min/max)

2. Creates a new dataset via ImportFromContentAsync named

"[AI] {ChartTitle} — {DatasetName}"

3. References the generated dataset ID in the chart block

This eliminates column mismatch errors entirely — the generated dataset

has exactly the columns the chart references. Falls back to the original

dataset when the generated one can't be created.

New file: ChartDatasetGenerator.cs (~170 lines) handles client-side

aggregation and CSV generation for single and grouped chart types.

Desktop 1.55.4 → 1.55.5

When insights are generated from a SQL view, the AI often references

computed columns (e.g., billable_pct) that exist in the view result but

not in the underlying dataset. These charts fail at render time with

"column not found" errors.

Now uses a two-pass approach:

1. Parse chart markers against all analysis columns (allows column name

normalization)

2. Validate parsed charts against EffectiveChartColumns (underlying

dataset) — skip charts with invalid columns instead of emitting them

3. Strip any unparsed [CHART:] markers so raw marker text never appears

in the rendered page

Desktop 1.55.3 → 1.55.4

Updated both cloud and local AI system prompts to frame the AI as a

senior data scientist performing exploratory data analysis. The cloud

prompt now suggests specific section structure (Data Quality, Key

Statistics, Patterns & Trends, Anomalies, Recommendations) and instructs

the AI to use sub-headings, bullet lists, and markdown tables for

well-structured output. Also instructs the model to cite specific column

names, row counts, percentages, and values for concrete analysis.

Desktop 1.55.2 → 1.55.3

BuildSectionBlocks now recognizes:

• ### through ###### headings → heading blocks at the correct level
• Bullet lists (-, *, +) → bullet_list blocks with grouped items
• Numbered lists (1., 2.) → numbered_list blocks with grouped items

Also relaxed chart marker column validation back to analysis columns

instead of chart-eligible columns — the prompt already guides the AI

toward using chart-eligible columns, and strict validation was rejecting

valid view-computed columns that the AI referenced in chart suggestions.

Desktop 1.55.1 → 1.55.2