Early-stage research on compositional worst-case resource bounds, with a linear session-IR checker, occupancy-indexed state experiments in Idris 2, and a prototype stack-depth certificate checker. The long-term goal is predictable memory and network-buffer usage.
Stepwise resource-bounding types: deterministic memory first, network resources second. This checkout hosts the Session IR calculus, checker and Idris spike (ULTRAPLAN R0-B) and the stackcert certificate checker (R1) on top of the RSR scaffold.
-
Plan of record (thesis, rung contract, kill criteria, decision log):
ULTRAPLAN.md -
Operational model (what the grade means):
docs/OPERATIONAL-MODEL.md -
Claims-to-files map (with verification status):
docs/EXPLAINME.adoc -
Vocabulary:
docs/glossary.adoc· Retractions & blocked runs:docs/retraction-ledger.adoc -
The gate:
just check(= proofs + tests + expected-rejection controls
checker on fixtures;just check-irruns the toolchain-free slice)
This repo is one node of the hyperpolymath type-theory family. The shared
map — vocabulary, connection obligations, ownership routing — lives in
nextgen-typing:
TYPE-CONNECTIONS.adoc.
-
Question this repo asks — how do live-state footprints compose, and when are they reclaimed? (Occupancy/HWM grades; protocol frontier = reclamation.)
-
Boundary — an occupancy bound is not a cost bound (HWM monoid ≠ max-plus); a measured live-byte count is not an occupancy grade. Cost grades live in tropical-types (the ULTRAPLAN name
tropical-resource-typingresolves there). -
Not this repo — echo indices, warrants, residue measures (shared glossary in TYPE-CONNECTIONS separates all of these from resource grades).
Quick start (Julia 1.10, standard library only):
julia --project=src/julia src/julia/bin/sessionir.jl test examples/session_ir/manifest.json
julia --project=src/julia src/julia/bin/sessionir.jl check examples/session_ir/accept_seq_add.sir|
Important
|
This file still contains RSR template material. You are reading |
A new repository scaffolded from the Rhodium Standard Repository (RSR) template: a batteries-included starting point that ships with CI/CD, machine-readable project metadata, an AI-agent gatekeeper protocol, a formally-typed ABI/FFI seam (Idris2 + Zig), container and reproducible-build scaffolding, and governance infrastructure — all wired and passing the RSR validators on day one.
Replace this section with a description of your project once initialised.
# 1. Create a repo from this template (or clone it), then from the repo root:
mise trust # allow this repository's mise.toml (needed without a TTY)
mise install # the toolchain, including the yq repo-init needs
mise exec -- just repo-init # interactive bootstrap: fills every {{PLACEHOLDER}}
# 2. See the available tasks:
mise exec -- just # lists all phases (build, test, validate, audit, ...)
# 3. Check the repo still satisfies the RSR shape:
mise exec -- just validate # structure + metadata checksmise exec -- puts the toolchain on PATH for that one command. Once mise is
activated in your shell (eval "$(mise activate bash)", or your shell’s
equivalent), plain just works and the prefix can be dropped.
just repo-init prompts for the project name, owner, author, licence contact, and the
other values listed in .machine_readable/ai/PLACEHOLDERS.adoc, substitutes them
across the tree, validates the result, and (if available) runs the k9-svc checks.
Say this to any AI assistant that can read a URL and run commands (or give you commands to paste):
Create a new repository from https://github.com/hyperpolymath/occupancy-typesThe URL is the key: it leads the assistant to
docs/AI_INSTALLATION_GUIDE.adoc, the
complete step-by-step recipe. The assistant checks your system, installs the
prerequisites, instantiates the template with just repo-init, and verifies the
result with just doctor.
-
"Set up occupancy-types so I can work on the template itself — https://github.com/hyperpolymath/occupancy-types"
-
./launcher.sh --ai-setup(orjust ai-setup) prints the line and the guide path.
-
Your operating system (Linux, macOS, or Windows via WSL2).
-
The new project’s name, owner, author and licence contact (the values listed in
.machine_readable/ai/PLACEHOLDERS.adoc). -
To confirm the privacy notice below.
|
Important
|
Setting up occupancy-types clones this repository and installs its developer
tools per-user with mise (no administrator rights, nothing sent anywhere).
|
Agents and CI can mint non-interactively: set RSR_NON_INTERACTIVE=1 and pass
the answers as environment variables named like the prompts' variables in
build/just/repo-init.just (PROJECT_NAME, REPO, OWNER, AUTHOR and
AUTHOR_EMAIL are required; any other answer left unset is blank or takes its
default). Piping the answers to stdin in prompt order also works.
Run just doctor; each FAIL carries a code explained in
docs/SETUP.adoc §Troubleshooting, and
just heal applies the automatic fixes. Prefer to do it yourself? The full
manual route is docs/SETUP.adoc; developers and
maintainers can brief an AI with just ai-warmup dev / just ai-warmup maintainer.
-
Machine-readable metadata (
.machine_readable/descriptiles/) —STATE,META,ECOSYSTEM,PLAYBOOK,AGENTIC,NEUROSYM,CLADE, andanchors/ANCHOR, in deed, so tools and agents can read the project’s state and boundaries. -
AI gatekeeper protocol —
occupancy-types_chora.deed(the repo deed) is the universal entry point that tells an AI agent how to work in this repo before it touches anything; it carries the AI allocation policy and the ply directory tree. -
Typed ABI/FFI seam —
src/interface/Abi/(Idris2 type + layout proofs) oversrc/interface/ffi/(Zig implementation), with generated C headers. -
CI/CD — GitHub Actions for quality, security (CodeQL, Scorecard, secret scanning), multi-forge mirroring, and RSR anti-pattern enforcement.
-
Supply-chain & reproducibility — container layering (stapeln), Guix shells, SBOM, and signing hooks.
-
Governance —
GOVERNANCE.adoc,MAINTAINERS.adoc,.github/community health files, and a releaseAUDIT.adocgate.
The K9 conformance badge in the row at the top becomes live after
just repo-init replaces the hyperpolymath and occupancy-types placeholders in its URL
and an eligible push to main publishes it. It is published by
.github/workflows/dogfood-gate.yml to a generated orphan branch called
governance-badges, and read back through the shields endpoint.
| Badge | What it means |
|---|---|
green |
The K9 lane ran for that commit on |
red |
The K9 lane ran for that commit on |
grey |
That commit produced no trustworthy K9 result. Grey is not "pending" and not a stale green: a skipped, cancelled or unavailable lane replaces an earlier green badge with grey. |
-
governance-badgesis generated output. It deliberately has no common ancestor withmain, it has no pull request, and it must never be merged. Recent pushes to it are ordinary badge refreshes. -
A run that would publish the same bytes makes no commit at all.
-
Nothing in the badge pipeline can turn a failing validation green: only a lane that ran and passed, with its artifact present, renders brightgreen.
The full operational contract - the exact state table, how API errors are
classified, how concurrent publications are serialised, and how to repair a bad
file - is in .github/workflows/README.adoc.
The authoritative map is generated, so it cannot drift from the tree:
docs/architecture/REPOSITORY-MAP.adoc
(regenerate with just repo-map; CI fails if it is stale).
The short version:
| Path | What lives there |
|---|---|
|
Start here - humans and AI agents respectively. |
|
The code and its tests. |
|
Human documentation, including the full map above. |
|
Manifests, contractiles and policies that tools read. |
|
Every task runs through |
|
CI configuration; GitHub reads |
-
The repository map — generated; what every directory is for.
-
EXPLAINME — the engineering deep-dive: how the pieces actually work.
-
AFFIRMATION — the dated, signed honesty snapshot of the repo’s true state.
-
AUDIT — the release audit gate.
-
AI installation guide — for agents instantiating this template.
-
PLACEHOLDERS — the full placeholder reference.
Code, configuration and scripts are Mozilla Public License 2.0
(MPL-2.0); prose documentation is CC-BY-SA-4.0. Both texts live in
LICENSES/, and per-file SPDX-License-Identifier headers are authoritative.
The GitHub-detected licence is MPL-2.0 (the root LICENSE). Long-term
attribution uses Quantum-Safe Provenance — see
the Quantum-Safe Provenance exhibit.