Looking for the commercial version? Visit www.ambientsystems.ai

Ambient Core

English

    Ambient Core

    Governed data products — open-source foundation anyone can rebuild.

    What it is

    Ambient Core is the governed data-product substrate for teams that want versioned YAML contracts, a reference metrics catalog, and pipeline governance in one MIT-licensed, fully open-source tree. It is not an agent framework and does not provide customer-facing AI, model hosting, or inference orchestration.

    What you get

    • Contracts — versioned data-product YAML under contracts/, validated and bundled for libraries.
    • Catalog — industries, metrics, and benchmarks with manifest generation for runtime reference data.
    • Pipeline — pandas Bronze→Silver→Gold plus optional Spark/Delta helpers in lib/ambient_pipeline/.
    • Libraries — ambient_contracts, ambient_cli, and ambient_calc for validation, catalog tooling, and formulas.
    • A hosted notebook entry (notebooks/00_colab_run_me.ipynb) and CI-ready packaging — no sample data or hosted storage is included in the repository.

    Scope

    You run pipelines and store data on your own infrastructure. The commercial Ambient Systems product (see the banner above) provides hosted storage, tenancy, billing, and managed product surfaces — none of that lives in this open-source core. Integrators should pin a release tag (currently v0.3.11) rather than floating main.

    Start

    Package version 0.3.11 (tag v0.3.11). Use Python 3.10–3.12. Spark tests require Java 17+ with JAVA_HOME set when running AMBIENT_SPARK_TESTS=1. 

    git clone https://github.com/Ambient-Team/ambient-core.git
cd ambient-core
python3.12 -m venv .venv
source .venv/bin/activate
pip install -e ".[all]"
cp contracts/*.yaml lib/ambient_contracts/bundled/
validate-contracts
AMBIENT_SPARK_TESTS=1 pytest

    Hosted notebook path: open notebooks/00_colab_run_me.ipynb from the tagged release in a notebook runtime, supply your own CSVs under data/raw/, and run Bronze→Silver→Gold with pandas — no paid Databricks or cloud hosting required for the core demo.

    For local Spark runs outside pytest, set AMBIENT_PII_SALT to a unique non-empty value (see .env.example); the pseudonymizer blocks missing or placeholder salts in production paths.

    Releases

    This project follows Keep a Changelog and Semantic Versioning. Highlights from recent tags:

    • v0.3.11

      Spark pipeline E2E golden tests, AMBIENT_PII_SALT runtime policy, structured pipeline logging, expanded CI quality gates, and lib/ambient_pipeline/ coverage floors — current recommended pin.

    • v0.3.8

      Tag-driven release workflow, maintainer cut-release automation, and releasing documentation for downstream integrators.

    • v0.3.0

      Baseline semver release: governed contracts, reference catalog, ambient_pipeline, ambient_calc, ambient_cli, and CI validation for contracts, catalog, and Spark tests.

    License

    Ambient Core is released under the MIT License. You may use, copy, modify, merge, publish, and distribute the software, including commercially, provided you include the copyright notice and license text shipped with the source.