Show HN: OM Core – multidimensional models without spreadsheet cell formulas
Summary
OM Core is an open-source multidimensional modeling engine that separates model structure from layout and presentation, using dimensions, cubes, and rules instead of spreadsheet cell formulas.
View Cached Full Text
Cached at: 07/03/26, 02:14 PM
cloudcell/om-core
Source: https://github.com/cloudcell/om-core
OM Core
OM Core is an open-source reference implementation of a multidimensional modeling engine for structured financial, operational, and analytical models.
Instead of treating the spreadsheet grid as the model, OM Core represents the model using dimensions, cubes, groups, and rules. Grids and views are projections of that model, not the source of truth.
Alpha software: OM Core is under active development. APIs, command names, file formats, GUI behavior, and module boundaries may change before v1.0.
Documentation
The main documentation site is here:
https://cloudcell.github.io/om-docs/
Start with:
- What is OM Core: https://cloudcell.github.io/om-docs/start/what-is-om-core/
- Why not spreadsheets: https://cloudcell.github.io/om-docs/start/why-not-spreadsheets/
- Installation: https://cloudcell.github.io/om-docs/start/installation/
- Quickstart: https://cloudcell.github.io/om-docs/start/quickstart/
Core idea
Most spreadsheet models mix several things together:
- model structure
- business logic
- layout
- presentation
- calculation flow
- user interaction
OM Core separates these concerns.
The model is built from:
- Dimensions — business axes such as Time, Account, Region, Product, Scenario, or Line Item.
- Cubes — data stored over one or more dimensions.
- Groups and hierarchies — structured collections and rollups.
- Rules — calculations expressed over semantic model addresses instead of spreadsheet coordinates.
- Views — grids and interfaces for inspecting and interacting with the model.
A rule should describe the business relationship, not the cell location. For example, a model should be able to express a relationship such as gross margin being derived from revenue and cost without making that relationship depend on a particular row, column, or copied formula.
Repository scope
This repository currently contains the full alpha application stack required to run OM Core.
That includes the modeling engine and supporting command, REPL, GUI/TUI, runtime, timeline, scripting, plugin, storage-adapter, examples, and test layers.
Some modules are internal implementation layers. They are included because the current alpha application depends on them. They should not yet be treated as stable public extension APIs.
Repository configuration
The .om/ directory is intentionally committed.
It contains default OM Core application configuration, including toolbar settings required by the current alpha application. It is not a temporary cache directory or private local state.
Do not delete .om/ unless the application configuration system has been
changed to load these defaults from another documented location.
Installation
OM Core currently runs from source.
Linux / macOS
git clone https://github.com/cloudcell/om-core.git
cd om-core
./start.sh
Windows
Use PowerShell:
git clone https://github.com/cloudcell/om-core.git
cd om-core
.\start.ps1
./start.sh or .\start.ps1 starts the GUI and asks whether to open a TUI in a
separate terminal.
OM Core uses uv to manage its Python environment.
Install uv, then run uv sync in the project root to create .venv and install
dependencies. After that, the start scripts and test scripts use uv run
automatically.
You can also start specific runtime modes:
# Linux / macOS
./start.sh --gui # graphical interface only
./start.sh --tui # terminal interface in the current terminal
./start.sh --runtime # headless runtime only
./start.sh --repl # REPL command shell in the current terminal
# Windows
.\start.ps1 --gui # graphical interface only
.\start.ps1 --tui # terminal interface in the current terminal
.\start.ps1 --runtime # headless runtime only
.\start.ps1 --repl # REPL command shell in the current terminal
For more detail, see the installation guide:
https://cloudcell.github.io/om-docs/start/installation/
Quickstart
Running ./start.sh (Linux / macOS) or .\start.ps1 (Windows) launches the GUI.
You will be prompted to open a TUI in a separate terminal; accepting the default
(Y) gives you a command shell alongside the GUI, as shown below.
Note: The first time you run
./start.shor.\start.ps1,uvwill create a Python virtual environment in the project folder (.venv) and install dependencies fromuv.lock.

After starting OM Core, try the built-in help command:
om> help
You can ask for help on specific topics:
om> help rule
om> help calc
A minimal OM Core script looks like this:
# Dimensions
dim Month Jan Feb Mar
# Cube
cube Sales Month
# View
view SalesView = Sales::Month
# Rules
rule Sales::Month.Jan = 100
rule Sales::Month.Feb = Sales::Month.Jan * 1.1
rule Sales::Month.Mar = Sales::Month.Feb * 1.1
# Calculate
calc
Save that as hello.openm, then source it from the REPL or TUI:
om> source hello.openm
For the full walkthrough, see:
https://cloudcell.github.io/om-docs/start/quickstart/
Testing
Run the test suite with the platform-specific script:
# Linux / macOS
./test.sh
# Windows
.\test.ps1
Architecture
OM Core is split into a session-scoped runtime layer, a command/query service layer, and multiple clients. The engine owns the canonical workspace state; the GUI, TUI, REPL, and CLI are clients that communicate through the message bus.

Why not just spreadsheets?
Spreadsheets are fast and flexible, but large models often become fragile because business logic is encoded in cell addresses, copied formulas, linked tabs, and implicit layout conventions.
OM Core uses a different level of abstraction. It makes the model explicit: dimensions describe the axes, cubes hold values, groups organize structure, rules define calculations, and views display the result.
The tradeoff is deliberate: you define more structure up front, and in return the model becomes easier to audit, extend, test, and maintain as it grows.
For the longer explanation, see:
https://cloudcell.github.io/om-docs/start/why-not-spreadsheets/
Project status
OM Core is currently alpha software.
Not yet promised before v1.0:
- stable public API
- stable plugin API
- stable scripting API
- stable file format
- packaged desktop installer
- production readiness for critical business use without independent validation
See also:
KNOWN_ISSUES.mdCHANGELOG.mdSECURITY.md
Legal
OM Core is distributed under the GNU Affero General Public License v3.0 unless otherwise stated.
See:
LICENSENOTICElegal/THIRD-PARTY-NOTICES.mdlegal/CONTRIBUTOR-CLA.mdlegal/CONTRIBUTOR-SIGNOFFS.md
The OM Core name is governed separately from the software license.
See:
legal/TRADEMARKS.md
Contributing
Contributions are welcome, especially small, focused improvements to examples, documentation, and clearly scoped engine behavior.
Please read:
CONTRIBUTING.mdlegal/CODE_OF_CONDUCT.mdSECURITY.md
Security issues should not be reported through public GitHub issues. See
SECURITY.md.
Feedback
- Bugs: open a GitHub issue.
- Discussion: join the Discord.
Similar Articles
Show HN: Visualize Model Spikiness in 3D
ModelMap is a web-based 3D visualization tool that displays AI model test scores as spiky shapes, allowing users to explore and compare model performance through interactive 3D graphics.
Show HN: Geomatic – a command-driven geometry studio enabled with autodiff
Geomatic is a command-driven geometry studio that leverages automatic differentiation for interactive geometric operations.
Show HN: OpenKnowledge – open source AI-first alternative to Obsidian/Notion
OpenKnowledge is an open source, local-first markdown editor and LLM wiki that integrates with Claude, Codex, and Cursor, offering real-time collaborative AI editing and team sharing via GitHub.
Orchestra-o1: Omnimodal Agent Orchestration
Orchestra-o1 is an omnimodal agent orchestration framework that supports efficient agent collaboration across text, image, audio, and video. It introduces decision-aligned group relative policy optimization (DA-GRPO) and achieves state-of-the-art performance on the OmniGAIA benchmark.
OneHOI: Unifying Human-Object Interaction Generation and Editing
OneHOI is a unified diffusion transformer framework that consolidates human-object interaction (HOI) generation and editing into a single conditional denoising process using relational modeling and structured attention mechanisms. The approach achieves state-of-the-art results across both HOI generation and editing tasks with support for multiple control modalities.