Version 0.12.0 documentation — covering the rebuilt recipe system, AI-guided analysis, multi-database support, and platform configuration. For background and earlier versions, see the v0.9 documentation.
The Statisticians Workbench is a self-hosted web application that combines database management, AI assistance, and data visualisation in one place — designed for domain experts, not IT specialists.
Read overview →Install on Linux directly. Desktop users can use Workbench Win (Windows) or Workbench Mac (planned). Covers prerequisites, first-run setup, and post-installation checks.
Installation guide →Configure your Workbench after installation — email settings, public URL, AI providers, database connections, user registration, and scheduled tasks.
Configuration guide →Migration notes and breaking changes when upgrading from v0.9. Covers schema changes, recipe format changes, and translation system updates.
Upgrade guide →The recipe system has been completely rebuilt for 0.12.0. Master-detail layout, three-component pipeline editor (Dataset/Transform, Algorithm, Output), and a universal SQL execution engine across five database backends.
Recipe documentation →An AI assistant guides you through building recipes — it sees your actual schema and data, suggests stages and configuration, and records every suggestion for audit. Connect to cloud AI providers or run a local LLM.
AI guide →Connect to PostgreSQL, MySQL/MariaDB, SQL Server, and MS Access. Schema browsing, table preview, and cross-database querying from a single interface. New in 0.12.0: MSSQL and MS Access support.
Connections guide →Import CSV, JSON, and Excel files with automatic type detection. New in 0.12.0: Virtual Datasets — DuckDB federated queries across uploaded files and database tables, with AI-guided creation.
Import guide →Descriptive statistics, correlation, t-tests, ANOVA, chi-square, and linear regression via DuckDB-native aggregates. New chart types in 0.12.0: scatter, area, horizontal bar, donut, histogram, distribution.
Statistics guide →Import Jupyter .ipynb notebook files into your workspace. Notebooks are stored alongside datasets and recipes and can be shared with your group.
Notebook import →Create accounts, self-register with email verification, password policies, and preferred language selection. New in 0.12.0: enhanced password strength enforcement.
User guide →Organise users into groups. Share recipes, database connections, and datasets within a group. New in 0.12.0: recipe sharing with "Shared by" badge and read-only mode.
Groups guide →System administration — user management, global settings, licence management, AI provider control, demo mode, and scheduled tasks. New in 0.12.0: Settings restructured under Admin → Setup.
Admin guide →Evaluate the Workbench in read-only demo mode with sample data. A pre-configured example user account demonstrates working recipes and workflows for learning purposes.
Demo guide →The Workbench supports 9 languages in the application UI, targeting 20. Translation is managed through YAML-based key files loaded into the database. New in 0.12.0: translation compliance audit completed.
Languages guide →Local-first architecture — your data stays on your machine. No cloud required. AI-guided features can use cloud providers or local LLMs. Designed for healthcare, government, and commercial data with privacy requirements.
Security overview →Export recipes as JSON or YAML for portability and backup. Import recipes from other Workbench installations. Planned for 0.12.0.
Export guide →Detailed release notes for each version — what's new, what's fixed, breaking changes, and migration notes from v0.9 through v0.12.0.
View changelog →| Topic | Description |
|---|---|
| SQL Execution Enginenew | Universal ops compiler (451 lines) with 5 backends: DuckDB, PostgreSQL, MySQL, MSSQL, Access. Capability matrix with automatic pandas fallback. |
| Recipe Stages Referencenew | Complete reference for every recipe stage: Filter, Join, Aggregate, Pivot, Impute, Sort, Sample, and Custom SQL. Input/output types and configuration options. |
| AI Provider Configurationupdated | Setup for Anthropic, Ollama, OpenAI-compatible providers (DeepSeek, etc.). Endpoint URLs, API keys, model selection, and per-provider capabilities. |
| Dataset Types & Sourcesupdated | Understanding dataset sources: database tables, uploaded files, virtual datasets, and recipe outputs. How DuckDB federated queries work across sources. |
| Translation Systemupdated | How the YAML-based translation system works. Adding a new language, updating translations, and the delta-mode import process. |
| Licensing & Feature Gatesupdated | Open-source Workbench vs Workbench-BE commercial edition. Feature comparison, licence key activation, and per-backend licence flags (mssql_db, etc.). |
| Audio Input/Output | Speech-to-text and text-to-speech capabilities. Per-user settings for audio input language and TTS voice. Delivered in 0.11.0. |
| Mobile & Responsive UI | Hamburger drawer navigation, responsive sidebars, touch-optimised targets. The Workbench is designed for desktop use but fully functional on tablets and phones. Delivered in 0.11.0. |
| Chart Types Referenceupdated | Complete reference: bar, line, scatter, area, horizontal bar, donut, histogram, distribution, and pie charts. Configuration options and usage in recipes vs sandbox. |
| Algorithm & Output Discoverynew | How the recipe editor dynamically discovers available algorithm types and visualisation types from the backend API. New in 0.12.0. |
| v0.9 Documentation | Archived documentation for version 0.9. Useful for historical context and upgrade planning. Some content may be superseded by 0.12.0. |