GANJAEDI
Ganjaedi script

For AI · agents · tools

One machine contract for Ganjaedi.

Fetch a compact language context, a full corpus, or structured JSON. These outputs are secondary serializations of active repository data and accepted Owner Canon; they do not create new Canon.

Start with one fetch.

For an ordinary LLM task, load quick.md first. Use grammar.json when the model needs exact structured rules, and full.md when it also needs the complete dictionary in one context window.

curl https://ganjaedi.pages.dev/ai/quick.md

01

Quick context

Compact bootstrap: authority, alphabet, structured grammar map, guardrails, and generation pseudocode.

02

Full context

One-document bundle with the structured core grammar, adversarial checks, and complete generated dictionary.

03

Structured spec

Top-level JSON contract with policy, orthography, structured grammar, dictionary metadata, and evals.

04

Grammar JSON

Core grammar snapshot serialized from active Owner Canon, including relations, valency, clause linking, comparison, aspect, and modality.

05

Dictionary JSON

Compact projection generated directly from dictionaryEntries. No separately maintained dictionary copy.

Fail closed, not creative.

Machine consumers should prefer explicit dictionary surfaces, preserve Unicode spelling, keep Latin and Cyrillic separate inside a token, follow the structured grammar snapshot, and report undefined forms instead of inventing them.

Discovery: /llms.txt. The contract also marks known traps such as LAT j → Ж/ДЖ ambiguity, non-canonical ASCII substitutions, vowel-final V y-stems, and forbidden passive inference.