Non-autoregressive System 1 decision engine. Typed choice, score and yes/no decisions over any text in a single forward pass, in 100+ languages, with a router that picks the right checkpoint per request.
Python calibrationclassificationdecision-modelhuggingfacejevmodernbertmultilingualnlppythonpytorchroutingtyped-decisionszero-shot
docs(docker): distinguish HTTP liveness from inference
Maxfield Allisonf4135c235ce6
diff --git a/compose.http.yaml b/compose.http.yamlindex 9893608d..43f8b32e 100644--- a/compose.http.yaml+++ b/compose.http.yaml@@ -97,8 +97,9 @@ services:# - ${LAYA_API_KEY_PATH:?Set LAYA_API_KEY_PATH}:/run/secrets/laya_api_key:roinit: truerestart: unless-stopped- # The server preloads before it binds, so /health answers once the preloaded- # checkpoints are ready. `up --wait` therefore waits for a usable server.+ # /health checks HTTP liveness. With the default LAYA_PRELOAD=0, it can pass+ # with no checkpoint loaded. `up --wait` does not verify a model forward;+ # even LAYA_PRELOAD=1 builds checkpoints without running an inference.healthcheck:test: ["CMD", "python", "-c", "import os, urllib.request; urllib.request.urlopen('http://127.0.0.1:' + os.environ['LAYA_PORT'] + '/health', timeout=4)"]interval: 30sdiff --git a/docs/docker.md b/docs/docker.mdindex 36df8567..30c27181 100644--- a/docs/docker.md+++ b/docs/docker.md@@ -314,9 +314,9 @@ The port is published on `127.0.0.1` only. The API has no authentication untilworking; with a key set it answers an unauthenticated caller `{"status": "ok"}` and withholds thecheckpoint, revision and device fields, which need the bearer.-The service has a healthcheck on `/health`. The server preloads before it starts-listening, so with `LAYA_PRELOAD=1` a healthy container has its checkpoints loaded.-`docker compose ... up -d --wait laya-serve` returns once it is healthy.+The service has a healthcheck on `/health`. `docker compose ... up -d --wait laya-serve` waits for HTTP liveness. With the Compose default `LAYA_PRELOAD=0`, it can return with no checkpoint loaded; the first valid inference request loads the checkpoint it needs. With `LAYA_PRELOAD=1`, the server builds the selected checkpoints before listening, but preloading does not run a model forward.++These are three separate checks: `/health` confirms the HTTP process responds, an authenticated `/health` response lists resident checkpoints, and a successful authenticated `POST /v1/systemone` confirms that the requested inference ran. A healthy container or a nonempty `loaded` list alone does not establish successful inference. The Docker workflow's HTTP job checks startup, loopback publication and authentication with preload disabled and invalid request bodies; it does not run a model forward.`/health` reports `device` as the device a resident checkpoint actually computeson, which is not always what `LAYA_DEVICE` asked for: a checkpoint that wants a