Möbius
A local-first framework for portable data apps and transformation pipelines, designed for humans and agents.
Möbius is an execution engine and development framework designed for humans and agents to turn single-use data analysis pipelines and applications into auditible, re-usable ones.
To do so, the engine packages workflows, applications and their associated runtime specifications into singlular sharable artifacts .
It also it abstracts the data ingest and run-time configuration processes to a series of plaintext-declared parameters and virtual environments. As a result, it elimiates installation overheads and JS framework bloat, while improving sharability, stability and output consistency.
How it works:
When Möbius package is shared with another user, the recipient can author a binding that maps the the schema of their dataset (or sets) onto the package’s expectations and run it. The engine absorbs the overhead costs of installation and system configuration by assembling a purpose-built virtual environment to the package’s own pinned specification.
Workflow results are emitted into a result bundle, a single DuckDB file with a generated Frictionless datapackage.json containing each result table’s inferred schema and the relations between them beside it. Result bundles themselves are valid Möbius inputs, enabling the chaining of workflows and applications.