This repository contains a documentation-first specification for a compact binary format for structured data.

The format is intended to remain easy to use in schema-less workflows while becoming materially smaller than plain MessagePack when repeated structure, repeated strings, homogeneous arrays, batching, or session reuse is present.

• reduce repeated object-key overhead
• support schema-aware compact encoding when schemas are available
• support learned structure in dynamic mode
• support row-wise, columnar, and stateful compression strategies
• keep deterministic wire behavior within a fixed profile

• replacing every existing JSON or binary protocol
• defining application semantics
• mandating a single transport handshake for every deployment

recurram/
├ README.md
├ LICENSE
├ CONTRIBUTING.md
├ SPEC.md
├ docs/
│  ├ format.md
│  ├ encoding.md
│  └ transport.md
├ versions/
│  └ v1.md
├ examples/
│  ├ README.md
│  ├ basic.json
│  └ schema-example.json
├ diagrams/
│  ├ protocol-flow.md
│  └ encoding-structure.md
├ .editorconfig
├ .gitignore
├ .markdownlint.jsonc
├ .prettierrc.json
└ package.json

1. SPEC.md for the core model and format overview.
2. docs/format.md for top-level kinds, object forms, batches, patches, and resets.
3. docs/encoding.md for scalar rules, vector codecs, string modes, and compression.
4. docs/transport.md for session-scoped state and transport assumptions.
5. versions/v1.md for the reference interoperability profile.
6. examples/ and diagrams/ for small concrete artifacts.

This repository includes a reference profile in versions/v1.md.

That profile fixes:

• supported kinds
• required reset operations
• default deterministic heuristics
• recommended metadata and codec rules
• constraints for stateful operation

This repository is distributed under CC-BY-4.0. See LICENSE for the full license text.