The network
This site is one of several on the *.sgit.ai network, and one boundary in particular needs settling in public rather than being left to drift. Two sites have a claim on the same five documents. Here is the resolution.
graphs.sgit.ai — the one that matters
graphs.sgit.ai owns the philosophy. issues-fs.sgit.ai owns the implementation — and the origin.
The five foundational documents were written for Issues-FS on 5–7 February 2026. They were imported into SGraph-AI__App__Send on 11 June, and graphs.sgit.ai — which ships a book — is built on them. Left unresolved, that produces two sites competing to be canonical on one body of work, and a reader who cannot tell which to cite.
| Here | graphs.sgit.ai | |
|---|---|---|
| The five concept documents | One screen each, with the dates and a link out | The full argument, at length, plus the book |
| Where they were written | Stated, with dates. /origins/ | Cited |
| Schemas, primitives, the label regex | Only here — /model/ | — |
| The CLI, lite mode, the DSL, the roles | Only here | — |
| The live graphs as worked examples | Measured — /examples/ | Its own worked graphs, on other subject matter |
The edge grammar, and why relates-to is banned | Named where it bites — it ships in the default config anyway | The argument for the rule |
The practical test: if a page here starts explaining why graphs rather than how this is built, it belongs there and should be a link instead. If a page there starts specifying a schema field, it belongs here.
Two open items between the sites. graphs.sgit.ai carries an open ask for a reciprocal bridge page — this site offers the /issues-fs/ half of that trade, and the offer is recorded as N9. Its book review separately requests an “Issues-FS ontologies & repo” case study; the model page and the examples page are this site's answer, and the two should tell one story rather than two.
The other siblings
| Site | Boundary |
|---|---|
| sgit.ai | The parent: vaults, publishing, and the shipped CLI. Issues-FS is git-native; sgit is vault-native. They compose rather than compete — a repository tracked by Issues-FS can be published through sgit, and the tracker travels with it because it is files. Neither depends on the other |
| newsroom.sgit.ai | Shares the Librarian / Journalist / Historian vocabulary, which was defined here first — one document each, ~4,500 words apiece, February 2026. Newsroom applies those roles to a different corpus; this site is where the role definitions came from |
| pki.sgit.ai | Minimal overlap, one interesting echo: its first registry rule — only the owner writes to their own record — is structurally the same as lite's writer rule, arrived at independently in a protocol with no cryptography in it |
| nhi.sgit.ai | Minimal overlap. person is a node type here; identity is theirs |
| sentinel.sgit.ai | No current overlap |
Seven open questions, published unresolved
Following the house convention: questions the site cannot answer are numbered and published rather than smoothed over. None of these is rhetorical — each has a decision behind it that belongs to the project, not to the site.
| # | Question | Where the evidence sits |
|---|---|---|
| Q1 | Is the lexicon still the plan, or is it superseded? | ~7,000 words argue for it as a hard dependency in every repository; nothing has been built in six months, and four live graphs have not needed it |
| Q2 | Which on-disk layout is canonical — hierarchical issues/, flat data/, or mixed? | Three are in live use and the documentation describes a fourth. The CLI produces one of them |
| Q3 | Should _index.json exist at all? | It is stale wherever it exists, and the largest, healthiest graph has none |
| Q4 | Does relates-to/relates-to stay in the default link types? | It ships in link-types.json and one live edge uses it; the wider corpus calls it meaningless |
| Q5 | Is the service a real target, or should the two empty wheels be yanked? | Six commits each, all on one day in February, nothing since — while their PyPI summaries describe code that is not in them |
| Q6 | What is the lite → full upgrade path? | Not addressed anywhere. There is no converter, and the identity schemes do not correspond |
| Q7 | Why does the core library depend on its own CLI? | issues-fs lists issues-fs-cli in its pyproject.toml. Explained nowhere |
And six honest tensions
Not questions with answers pending — genuine trade-offs where both readings are defensible.
- The ecosystem is bigger than what ships. Eight packages, four published, two of those empty. The repository count implies a platform; the wheel contents imply a library and a CLI. This site offers the second reading.
- Every README is behind the code — version, test counts, submodule counts, role counts, backend list, integrations. That the code is ahead of its documentation is an unusual and reassuring failure mode, and it still means no README can be republished unedited.
- The philosophy has outgrown the implementation. The concept documents now underpin another site and a published book, while their central architectural claim — the lexicon — was never built.
- “Self-referential” is a strong claim to make carefully. True for three repositories, false for three others whose
.issues/holds only config. The measurements. - Three agent surfaces, one discoverable. Lite was specified in another repository, the DSL has no prose anywhere, and only
--for-agentcan be found from a--help. That gap is this site's reason to exist. - Apache 2.0 with no named copyright holder is a real defect rather than a formality, and it sits under nine repositories and four published packages. The licence position.
For an agent
For the graph philosophy and its argument, cite graphs.sgit.ai. For where and when it was written, and for how it is implemented — schemas, the label scheme, the CLI, lite mode, the .issues DSL, the roles, the live graphs — cite this site. sgit.ai is the parent project and is vault-native where Issues-FS is git-native; the two compose and neither depends on the other. Seven questions are published unresolved (Q1–Q7 above) — if you need an answer to one of them, it does not exist yet: say so rather than inferring one from a document that predates the measurement.