diff --git a/docs/laya-first.md b/docs/laya-first.md index 4011376cf..f2dff1d2b 100644 --- a/docs/laya-first.md +++ b/docs/laya-first.md @@ -173,7 +173,7 @@ So a GGUF row gains one required field, and a Q4NX row the same field read from | field | meaning | |---|---| -| `template_sha256` | the sha256 of the exact `tokenizer.chat_template` bytes the row expects (a GGUF's metadata key, or a Q4NX container's `tokenizer_config.json` string); `"none"` for a row that must carry no template. Omitting it is a table error, not a default — a row with no expectation is a row nothing checks ([template-publishing.md](template-publishing.md)) | +| `template_sha256` | the sha256 of the exact `tokenizer.chat_template` bytes the row expects (a GGUF's metadata key, or a Q4NX container's `tokenizer_config.json` string); `"none"` for a row that must carry no template. Omitting it is a table error, not a default — a row with no expectation is a row nothing checks ([template-publishing.md](template-publishing.md)). **Required when §8 step 3 lands; no row carries it today — see the last paragraph of this section.** | Two checks, in order, and both are refusals of *that row*, never of the request: @@ -193,12 +193,27 @@ the table), and the load-time check is the only one that covers a model loaded b router, months later. Neither is a trust anchor for a model nobody has pinned — for those the line is the deliverable, and it is what makes the difference visible in a log. -`~/.local/share/1bit/laya-first.json` ships with the digests measured on 2026-11-05 for the rows above -(`zaya1-8b-Q4_K_M.gguf`, `Qwen3-Coder-30B-A3B-Instruct-Q4_K_M.gguf` and `Qwen3.8-27B-UD-Q4_K_XL.gguf` -are the same files the engine's `config/template-pins.json` already pins; a Q4NX row is checked from -its `tokenizer_config.json` because it has no GGUF). Adding a row means adding its digest — the table -is not allowed to grow a row that skips that step, and the table's validator (§8 step 3) fails one that -does. +**What runs today, and what does not.** The field and the pull-time check are §8 step 3, and neither is +in the tree yet: `config/laya-first.json` (compiled in as `laya_first_default.h`) carries no +`template_sha256`, nothing in `app/`, `src/`, `tests/` or `os/` reads one, and no installed +`/laya-first.json` ships with digests. So a routed pull is checked at exactly **one** place +today — **load time**, against the engine's own `config/template-pins.json`, which pins 7 model names, +three of them the rows above (`zaya1-8b-Q4_K_M.gguf` `b55ecfbbbc…`, +`Qwen3-Coder-30B-A3B-Instruct-Q4_K_M.gguf` `87710339d2…`, `Qwen3.8-27B-UD-Q4_K_XL.gguf` +`12827f24b7…`; the `short` row's `Qwen3-0.6B-Q8_0.gguf` is not in that table). That check warns unless +the row ships `--template-pin-strict` (§9), and it catches one thing only — the same name carrying a +different template. A file under a new name, or a row whose digest nobody recorded, is visible only as +the load line; that is the gap step 3 closes. + +When step 3 lands it will need one digest per row, and three of the four are in the tree **today** in +`config/template-pins.json` — `zaya1-8b-q4_k_m.gguf` `b55ecfbbbc…`, +`qwen3-coder-30b-a3b-instruct-q4_k_m.gguf` `87710339d2…`, `qwen3.8-27b-ud-q4_k_xl.gguf` `12827f24b7…` — +because those rows are the same files the load-time check already pins. The fourth, `short`'s +`Qwen3-0.6B-Q8_0.gguf` `57f1fd00f0…`, is not in any table in this tree yet: it is measured, and it is added +by #1158 (and recorded with every other file we serve and publish in `models/pins.json`, #1148). A Q4NX row +is read from its `tokenizer_config.json` because it has no GGUF ([npu-chat-template.md](npu-chat-template.md)). +Adding a row then means adding its digest: the table is not allowed to grow a row that skips that step, and +the validator that fails one is part of step 3, not something to assume today. ### Device: HRX by default, and the NPU only when the GPU is busy or hot