Exploring interoperability & composability for local-first software (Groundmist post 3)

The 3rd post of the series focuses "on interoperability and composability—how we can combine data from multiple local-first applications, privately and seamlessly".

The problem: by their nature, "Local-first software applications typically store data separately and independently", so it's fragmented. The lack of "common mechanism or standard for defining how data should be structured or discovered ... significantly limits the possibility of composing data from multiple sources".

As in post 2, she tackles the issue of AI: "While it’s possible to simply store all local-first data on a single device and point an AI model at it, this solution doesn’t scale well across multiple devices and environments... Clearly defined schemas and discoverable data locations enable AI agents to reliably access and manipulate data across multiple devices and applications... [while] Structured interoperability allows AI to effortlessly generate insights, summaries, or recommendations without manual data export and aggregation."

Personal sync servers

Analogous to ATProto's PDS, a PSS is a "user-owned server that syncs your data across multiple local-first apps into a centralized store, making it easy to reuse and compose your data in new contexts."

Using PSSs require "legibility and discoverability", which is provided by the Lexicon schema system. The way data is stored also "sets the stage for a granular capability-based authorization model for access control", while AT Protocol's identity system is "portable, user-controlled ... well-suited for ownership of the personal sync server".

So she built "Groundmist Sync, a personal sync server prototype linked to your AT Protocol identity and located at the domain you configure... [so] your Groundmist apps automatically find it... After authenticating via your atproto account, apps seamlessly sync data using structured paths based on Lexicons".

She concludes by outlining the work to be done.