# mostly-ai/mostlyai-engine architecture

> Architecture diagram of the GitHub repository mostly-ai/mostlyai-engine, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-01). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/mostly-ai/mostlyai-engine
- Repository: https://github.com/mostly-ai/mostlyai-engine

## Overview

MOSTLY AI Engine is a Python library for fitting tabular or language generative models and reusing them for synthetic sampling and downstream tasks. The principal workflow is data preparation and analysis, model training, then generation or inference through the model interface; tabular models additionally support probabilities, prediction, imputation, likelihood scoring, and fairness controls. Edges are limited to README statements and relationships named in the SOURCE INDEX; unsampled implementation details are not inferred.

Main entry points: [`mostlyai/engine/_tabular/interface.py`](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_tabular/interface.py) (TabularARGN API), [`mostlyai/engine/_workspace.py`](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_workspace.py) (Workspace files), [`mostlyai/engine/_language/interface.py`](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_language/interface.py) (LanguageModel API), [`mostlyai/engine/_tabular/generation.py`](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_tabular/generation.py) (Tabular generation)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_api["Public API"]
  node_tab_api["TabularARGN API<br/>[interface.py]"]
  node_lang_api["LanguageModel API<br/>[interface.py]"]
end

subgraph group_prep["Data Preparation"]
  node_split["Split data<br/>[splitting.py]"]
  node_analysis["Analyze data<br/>[analysis.py]"]
end

subgraph group_tabular["Tabular Models"]
  node_tab_train["Tabular training<br/>[training.py]"]
  node_argn["ARGN model<br/>[argn.py]"]
  node_tab_encode["Tabular encoding<br/>[encoding.py]"]
  node_tab_generate["Tabular generation<br/>[generation.py]"]
  node_tab_probability["Probability queries<br/>[probability.py]"]
  node_fairness["Fairness controls<br/>[fairness.py]"]
end

subgraph group_language["Language Models"]
  node_language_train["Language training<br/>[training.py]"]
  node_language_encode["Language encoding<br/>[encoding.py]"]
  node_language_generate["Language generation<br/>[generation.py]"]
  node_language_engine["Model backends"]
end

subgraph group_shared["Shared Engine"]
  node_workspace[("Workspace files<br/>[_workspace.py]")]
  node_type_encode["Encoding types"]
  node_common["Shared utilities<br/>[_common.py]"]
  node_domain["Model domain<br/>[domain.py]"]
end

node_user(("Python user"))

node_user -->|"uses"| node_tab_api
node_user -->|"uses"| node_lang_api
node_tab_api -->|"fits model"| node_tab_train
node_tab_api -->|"requests samples"| node_tab_generate
node_tab_api -->|"queries probabilities"| node_tab_probability
node_tab_api -->|"imports analysis"| node_analysis
node_tab_api -->|"imports splitting"| node_split
node_lang_api -->|"fits model"| node_language_train
node_lang_api -->|"requests samples"| node_language_generate
node_lang_api -->|"imports analysis"| node_analysis
node_lang_api -->|"imports splitting"| node_split
node_tab_train -->|"uses model"| node_argn
node_tab_train -->|"uses encoding"| node_tab_encode
node_tab_generate -->|"uses model"| node_argn
node_tab_generate -->|"uses encoding"| node_tab_encode
node_tab_generate -->|"uses controls"| node_fairness
node_tab_api -->|"supports fairness"| node_fairness
node_tab_api -->|"supports prediction and likelihood"| node_tab_probability
node_language_train -->|"uses encoding"| node_language_encode
node_language_generate -->|"uses encoding"| node_language_encode
node_language_generate -->|"dispatches backend"| node_language_engine
node_analysis -->|"uses encoders"| node_type_encode
node_language_encode -->|"uses encoders"| node_type_encode
node_analysis -->|"reads and writes"| node_workspace
node_split -->|"writes partitions"| node_workspace
node_tab_train -->|"reads and writes"| node_workspace
node_language_train -->|"reads and writes"| node_workspace
node_tab_generate -->|"reads model data"| node_workspace
node_language_generate -->|"reads model data"| node_workspace
node_tab_encode -->|"imports utilities"| node_common
node_language_encode -->|"imports utilities"| node_common
node_analysis -->|"uses model types"| node_domain

click node_tab_api "https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_tabular/interface.py"
click node_lang_api "https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_language/interface.py"
click node_split "https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/splitting.py"
click node_analysis "https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/analysis.py"
click node_workspace "https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_workspace.py"
click node_tab_train "https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_tabular/training.py"
click node_argn "https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_tabular/argn.py"
click node_tab_encode "https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_tabular/encoding.py"
click node_tab_generate "https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_tabular/generation.py"
click node_tab_probability "https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_tabular/probability.py"
click node_fairness "https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_tabular/fairness.py"
click node_language_train "https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_language/training.py"
click node_language_encode "https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_language/encoding.py"
click node_language_generate "https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_language/generation.py"
click node_language_engine "https://github.com/mostly-ai/mostlyai-engine/tree/main/mostlyai/engine/_language/engine"
click node_type_encode "https://github.com/mostly-ai/mostlyai-engine/tree/main/mostlyai/engine/_encoding_types"
click node_common "https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_common.py"
click node_domain "https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/domain.py"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_tab_api,node_lang_api,node_user toneBlue
class node_split,node_analysis toneAmber
class node_tab_train,node_argn,node_tab_encode,node_tab_generate,node_tab_probability,node_fairness toneMint
class node_language_train,node_language_encode,node_language_generate,node_language_engine toneRose
class node_workspace,node_type_encode,node_common,node_domain toneIndigo
```

## Components

### Public API

- [TabularARGN API](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_tabular/interface.py): `mostlyai/engine/_tabular/interface.py`
- [LanguageModel API](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_language/interface.py): `mostlyai/engine/_language/interface.py`

### Data Preparation

- [Split data](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/splitting.py): `mostlyai/engine/splitting.py`
- [Analyze data](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/analysis.py): `mostlyai/engine/analysis.py`

### Tabular Models

- [Tabular training](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_tabular/training.py): `mostlyai/engine/_tabular/training.py`
- [ARGN model](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_tabular/argn.py): `mostlyai/engine/_tabular/argn.py`
- [Tabular encoding](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_tabular/encoding.py): `mostlyai/engine/_tabular/encoding.py`
- [Tabular generation](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_tabular/generation.py): `mostlyai/engine/_tabular/generation.py`
- [Probability queries](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_tabular/probability.py): `mostlyai/engine/_tabular/probability.py`
- [Fairness controls](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_tabular/fairness.py): `mostlyai/engine/_tabular/fairness.py`

### Language Models

- [Language training](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_language/training.py): `mostlyai/engine/_language/training.py`
- [Language encoding](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_language/encoding.py): `mostlyai/engine/_language/encoding.py`
- [Language generation](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_language/generation.py): `mostlyai/engine/_language/generation.py`
- [Model backends](https://github.com/mostly-ai/mostlyai-engine/tree/main/mostlyai/engine/_language/engine): `mostlyai/engine/_language/engine`

### Shared Engine

- [Workspace files](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_workspace.py): `mostlyai/engine/_workspace.py`
- [Encoding types](https://github.com/mostly-ai/mostlyai-engine/tree/main/mostlyai/engine/_encoding_types): `mostlyai/engine/_encoding_types`
- [Shared utilities](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/_common.py): `mostlyai/engine/_common.py`
- [Model domain](https://github.com/mostly-ai/mostlyai-engine/blob/main/mostlyai/engine/domain.py): `mostlyai/engine/domain.py`

### Other components

- **Python user**

## Connections

- Python user → TabularARGN API: uses
- Python user → LanguageModel API: uses
- TabularARGN API → Tabular training: fits model (evidence: `mostlyai/engine/_tabular/interface.py`)
- TabularARGN API → Tabular generation: requests samples (evidence: `mostlyai/engine/_tabular/interface.py`)
- TabularARGN API → Probability queries: queries probabilities (evidence: `mostlyai/engine/_tabular/interface.py`)
- TabularARGN API → Analyze data: imports analysis (evidence: `mostlyai/engine/_tabular/interface.py`)
- TabularARGN API → Split data: imports splitting (evidence: `mostlyai/engine/_tabular/interface.py`)
- LanguageModel API → Language training: fits model (evidence: `mostlyai/engine/_language/interface.py`)
- LanguageModel API → Language generation: requests samples (evidence: `mostlyai/engine/_language/interface.py`)
- LanguageModel API → Analyze data: imports analysis (evidence: `mostlyai/engine/_language/interface.py`)
- LanguageModel API → Split data: imports splitting (evidence: `mostlyai/engine/_language/interface.py`)
- Tabular training → ARGN model: uses model (evidence: `mostlyai/engine/_tabular/training.py`)
- Tabular training → Tabular encoding: uses encoding (evidence: `mostlyai/engine/_tabular/training.py`)
- Tabular generation → ARGN model: uses model (evidence: `mostlyai/engine/_tabular/generation.py`)
- Tabular generation → Tabular encoding: uses encoding (evidence: `mostlyai/engine/_tabular/generation.py`)
- Tabular generation → Fairness controls: uses controls (evidence: `mostlyai/engine/_tabular/generation.py`)
- TabularARGN API → Fairness controls: supports fairness
- TabularARGN API → Probability queries: supports prediction and likelihood (evidence: `mostlyai/engine/_tabular/interface.py`)
- Language training → Language encoding: uses encoding (evidence: `mostlyai/engine/_language/training.py`)
- Language generation → Language encoding: uses encoding (evidence: `mostlyai/engine/_language/generation.py`)
- Language generation → Model backends: dispatches backend (evidence: `mostlyai/engine/_language/generation.py`)
- Analyze data → Encoding types: uses encoders (evidence: `mostlyai/engine/analysis.py`)
- Language encoding → Encoding types: uses encoders (evidence: `mostlyai/engine/_language/encoding.py`)
- Analyze data → Workspace files: reads and writes (evidence: `mostlyai/engine/analysis.py`)
- Split data → Workspace files: writes partitions (evidence: `mostlyai/engine/splitting.py`)
- Tabular training → Workspace files: reads and writes (evidence: `mostlyai/engine/_tabular/training.py`)
- Language training → Workspace files: reads and writes (evidence: `mostlyai/engine/_language/training.py`)
- Tabular generation → Workspace files: reads model data (evidence: `mostlyai/engine/_tabular/generation.py`)
- Language generation → Workspace files: reads model data (evidence: `mostlyai/engine/_language/generation.py`)
- Tabular encoding → Shared utilities: imports utilities (evidence: `mostlyai/engine/_tabular/common.py`)
- Language encoding → Shared utilities: imports utilities (evidence: `mostlyai/engine/_language/encoding.py`)
- Analyze data → Model domain: uses model types (evidence: `mostlyai/engine/analysis.py`)
