User Guide — Mirabelle (ForecastAI 2026.01)¶
Welcome to the Mirabelle user guide. Mirabelle is a supply chain forecasting and inventory optimisation platform for demand planners. It runs as a web application: the API listens on port 8002 and the browser interface on port 5173.
This guide covers every screen in the application, written for planners who are new to the system. You do not need to understand the underlying algorithms to use the software effectively.
Screens at a Glance¶
| Screen | URL | One-line description |
|---|---|---|
| Overview / Dashboard | / |
Series table — every item-site combination with metrics and sparklines |
| Time Series Viewer | /series/:id |
Drill-down for one series: all method forecasts, ridge chart, metric comparison, override editor |
| Exceptions | /exceptions |
Triage screen — items with forecast, IO, or supply issues requiring planner action |
| Pipelines | /pipelines |
Manage forecast scenarios, EOQ curves, MEIO runs, supply scenarios, and pipeline workflows |
| Segments | /segments |
Define named groups of items with rule-based criteria |
| Classifications | /abc |
ABC/XYZ classification — assign velocity and variability classes |
| Settings | /settings |
Pipeline parameters, forecasting method weights, locale, dark mode |
| Process Runner | /processes |
Run individual pipeline steps (ETL, Forecasting, MEIO, etc.) |
| Alert Rules | /alert-ontology |
View and seed the 32-rule alert ontology |
| Causal Forecasting | (via Pipelines) | Asset-driven demand model for MRO / maintenance organisations |
| NPI Manager | (via Pipelines) | Copy donor forecasts to new parts with a scale factor |
| Supply Planning | (via Dashboard or Detail) | Inventory projections, orders, lead times |
| Firm Supply Orders | (via Supply panels) | ERP-originated firm orders — input to the supply engine |
| Replay Engine | /replay |
Re-run supply + allocation over a historical horizon to analyse fill rate and stock evolution |
| Mirabelle AI Assistant | (right panel) | Natural-language Q&A, exceptions, and help — fully on-premise |
| Forecasting Guide | — | Non-technical guide to reading forecast results |
Navigation¶
The left sidebar is always visible. It collapses to icon-only mode by clicking the double-arrow button at the top. When expanded it shows:
Action group
- Overview — the main Dashboard
- Exceptions — triage queue
- Detail — the last series you visited (remembers across page loads)
- Scenario Review — frozen pipeline snapshots and comparison
- Replay — historical replay of supply + allocation planning
Configuration group
- Segments — named item groups
- Classifications — ABC/XYZ
- Settings — system parameters
Admin group
- Process Runner — run pipeline steps
- Audit Log — change history
- Users — user management (admin only)
- Alert Rules — rule ontology (admin only)
- Tenant — multi-tenant provisioning (superadmin only)
Right toolbar — Mirabelle AI Assistant¶
On the right edge of the screen, the 
sparkle button opens the Mirabelle AI panel — a conversational APS assistant that translates natural-language questions into read-only ClickHouse queries. The panel has three tabs:
| Tab | Purpose |
|---|---|
| Chat | Ask questions about demand, supply, forecasts, inventory — results appear as charts, tables, or KPI cards |
| Exceptions | Approve or dismiss flagged alerts with severity filters |
| Help | Built-in guide with sample prompts and privacy notes |
The panel is resizable (drag the left edge) and fully on-premise — no data leaves the network. See the Mirabelle AI Assistant page for full details.
The Pipeline Selector¶
Just below the logo in the sidebar is the Pipeline Selector dropdown. This is a critical control: everything you see in the application — forecasts, supply projections, MEIO results, exceptions — belongs to the currently selected pipeline.
Always check your pipeline first
Before reviewing exceptions or supply projections, confirm the correct pipeline is shown in the selector. Running processes on the wrong pipeline is the most common source of confusion.
Pipelines are named run contexts (for example "Production Q2 2026" or "Scenario — High Demand"). Each pipeline links to one or more scenarios. Switching the pipeline changes the data shown across all screens simultaneously.
Master data (items, sites, demand actuals, firm orders) is not scoped to a pipeline — it is shared across all pipelines.
Dark Mode and Locale¶
At the bottom of the sidebar you can toggle between light and dark mode using the sun/moon button. Number and date formatting (decimal separator, date order) can be changed in Settings.
Getting Help¶
- Use the Full Tour button at the bottom of the sidebar for a guided walkthrough of every page in the application.
- Use the Demo Tour button for a presales-focused walkthrough that highlights management KPIs, probabilistic forecasting, and planner simplicity across 9 stops.
- Hover over any column header or field label to read a tooltip.
- The Mirabelle panel (right edge of the screen) can answer natural-language questions about your data.
- The Swagger API documentation is available at
http://localhost:8002/docs.