Spaces:

---

nolanzandi

/

virtual-data-analyst

like 23

Running

Feel free to request features that would be helpful to you or report issues with released features.

Recent Bugfixes:

• Fixed chart iframes being hardcoded to 1000px wide -- charts are now fully responsive and resize with the chat window 5/31/2026
• Fixed chart URLs containing 'None' when ROOT_URL environment variable was not set, causing all charts to display as empty boxes 5/31/2026
• Fixed SQL data source being incorrectly identified as MongoDB in the example question generator 5/31/2026
• Showing errors on data file upload so user isn't stuck with a nonfunctional chat window 3/4/2025
• Released refactor to get more accurate LLM responses (especially when generating charts/graphs) and fewer errored responses 2/25/2025
• Fixed bug where conversations between sessions were mixing together and messages from old sessions were still being processed by llm 2/21/2025
• Fixed bug where missing package prevented XLSX file upload 2/17/2025
• Fixed bug where datetime fields were written as strings and were not queryable 2/15/2025

Releases:

• Added rolling statistics / moving average chart: overlay rolling mean, std, min, or max on any time series with a configurable window 5/31/2026
• Added K-Means clustering: groups data into k clusters on numeric feature columns, returns a scatter plot coloured by cluster and a centroid summary table 5/31/2026
• Added hypothesis testing: independent-samples t-test, one-sample t-test, and chi-square test of independence with formatted results table 5/31/2026
• Added box plot for distribution visualisation and outlier detection, optionally grouped by a category 5/31/2026
• Added correlation heatmap: pairwise Pearson correlation matrix rendered as a colour-coded heatmap 5/31/2026
• Added descriptive statistics summary table (count, mean, std, min, quartiles, max) for numeric columns 5/31/2026
• UX and design improvements across the interface 5/30/2026
• Added user-supplied API keys with model selection -- supports OpenAI (gpt-4.1, gpt-4o, etc.) and Anthropic (Claude Sonnet, Opus, Haiku) 5/29/2026
• Added an integration for GraphQL API endpoints and queries in a live setting - 5/10/2025
• Added an integration for NoSQL document databases (MongoDB for now, can add more if requested) to be able to connect to JSON data in a live data source - 5/4/2025
• Added an integration for SQL databases (PostgreSQL for now, can add more if requested) to be able to speak to a live data source - 4/18/2025
• Added new size option to scatter plots to generate bubble charts 3/11/2025
• Added pie charts and histograms. Added new options to scatter plots to allow for trendlines, and marginal analysis. 3/9/2025
• Released a regression function to run (currently linear) regressions on data and answer questions about relationships to variables 3/3/2025
• Released a refactor to allow much larger data queries, to prevent the results of queries from passing through llm and creating API errors. Should increase speed of responses as well 3/3/2025
• Allowed for tabular formatting 2/25/2025
• Updated charts and graphs to use Plotly and be locally hosted 2/25/2025
• Json files are flattened and normalized so nested json can be uploaded to Sqlite. Useful to process log data although this instance likely doesn't have enough compute to process large logfiles (which was my usecase for building this feature) 2/20/2025
• Charts and graphs are dynamic and link to Quickchart where they can be edited and shared 2/18/2025
• Ability to generate charts and graphs (images) 2/17/2025
• Generate recommended sample questions on data upload that could uncover data insights 2/16/2025
• Ability to upload new data file formats (CSV, TSV, TXT, XLS, XLSX, XML, and JSON) 2/15/2025
• Allow for simultaneous users (no more shared data file) 2/13/2025
• Example datasets to make it easier to demo 2/14/2025

Upcoming:

• Talk to live data: Allow for REST API queries and the ability to connect to an API instead of a data file (low structure, hardest)
• Talk to live data: Add additional SQL database connectors
• Talk to live data: Add additional document database connectors
• Allow for tables to be downloaded in CSV once generated
• Adding and enhancing visualization types and options (subplots, custom chart controls, etc)
• Enable voice transcription to allow user to speak to Virtual Data Analyst
• Allow for created DB files to be downloaded and/or exposed with URL

Medium Term:

• Setup unit tests to reduce bug incidences
• Add additional models and setup evaluators to improve response quality
• Be able to upload data to a NoSQL instead of SQLite database if appropriate (more db types, better than flattening files for nested Json)

Long Term:

• Add user profiles and permanent document storage tied to users
• Multiple file evaluation that can be compared to each other
• Data editing and manipulation with export