Packages Overview

OpenAdapt v1.0+ uses a modular meta-package architecture. The main openadapt package provides a unified CLI and depends on focused sub-packages via PyPI.

Architecture

graph TD
    OA[openadapt<br/>Meta-package]

    OA -->|capture| CAP[openadapt-capture]
    OA -->|ml| MLP[openadapt-ml]
    OA -->|evals| EVL[openadapt-evals]
    OA -->|viewer| VWR[openadapt-viewer]
    OA -->|grounding| GRD[openadapt-grounding]
    OA -->|retrieval| RET[openadapt-retrieval]
    OA -->|privacy| PRV[openadapt-privacy]

    OA -->|core| CORE[Core Bundle]
    CORE --> CAP
    CORE --> MLP
    CORE --> EVL
    CORE --> VWR

    OA -->|all| ALL[Full Bundle]
    ALL --> CORE
    ALL --> GRD
    ALL --> RET
    ALL --> PRV

    classDef meta fill:#2C3E50,stroke:#1A252F,color:#fff
    classDef core fill:#27AE60,stroke:#1E8449,color:#fff
    classDef optional fill:#E67E22,stroke:#A04000,color:#fff
    classDef bundle fill:#8E44AD,stroke:#5B2C6F,color:#fff

    class OA meta
    class CAP,MLP,EVL,VWR core
    class GRD,RET,PRV optional
    class CORE,ALL bundle

Core Packages

These packages provide the essential functionality for recording, training, evaluating, and visualizing.

Package	Description	Install Extra
openadapt-capture	GUI recording, event capture, storage	capture
openadapt-ml	ML engine, training, inference	ml
openadapt-evals	Benchmark evaluation infrastructure	evals
openadapt-viewer	HTML visualization components	viewer

Install all core packages:

pip install openadapt[core]

Optional Packages

These packages provide enhanced functionality for specific use cases.

Package	Description	Install Extra
openadapt-grounding	UI element localization	grounding
openadapt-retrieval	Multimodal demonstration retrieval	retrieval
openadapt-privacy	PII/PHI scrubbing	privacy

Install all packages:

pip install openadapt[all]

Installation Options

Individual Packages

pip install openadapt[capture]     # GUI capture/recording
pip install openadapt[ml]          # ML training and inference
pip install openadapt[evals]       # Benchmark evaluation
pip install openadapt[viewer]      # HTML visualization
pip install openadapt[grounding]   # UI element localization
pip install openadapt[retrieval]   # Demo search/retrieval
pip install openadapt[privacy]     # PII/PHI scrubbing

Multiple Packages

pip install openadapt[capture,ml,evals]

Bundles

pip install openadapt[core]        # capture + ml + evals + viewer
pip install openadapt[all]         # Everything

Data Flow

flowchart LR
    subgraph Record["1. Record"]
        A[User Demo] --> B[Capture Session]
        B --> C[Screenshots + Events]
    end

    subgraph Store["2. Store"]
        C --> D[JSON/Parquet Files]
        D --> E[Demo Library]
    end

    subgraph Train["3. Train"]
        E --> F[Data Loading]
        F --> G[Model Training]
        G --> H[Checkpoint]
    end

    subgraph Deploy["4. Deploy"]
        H --> I[Agent Policy]
        I --> J[Inference]
        J --> K[Action Replay]
    end

    subgraph Evaluate["5. Evaluate"]
        I --> L[Benchmark Runner]
        L --> M[Metrics]
        M --> N[Results Report]
    end

    GROUND[Grounding] -.-> J
    RETRIEVE[Retrieval] -.-> F
    PRIV[Privacy] -.-> C

Package Repositories

Each package is maintained in its own repository:

Package	Repository
openadapt	OpenAdaptAI/OpenAdapt
openadapt-capture	OpenAdaptAI/openadapt-capture
openadapt-ml	OpenAdaptAI/openadapt-ml
openadapt-evals	OpenAdaptAI/openadapt-evals
openadapt-viewer	OpenAdaptAI/openadapt-viewer
openadapt-grounding	OpenAdaptAI/openadapt-grounding
openadapt-retrieval	OpenAdaptAI/openadapt-retrieval
openadapt-privacy	OpenAdaptAI/openadapt-privacy

Contributing

To contribute to a specific package:

1. Fork and clone the package repository
2. Install in development mode: pip install -e ".[dev]"
3. Make your changes
4. Submit a pull request

See Contributing for more details.