Skip to content

Commit 326387b

Browse files
Guard changeset release status
1 parent 87ebb8f commit 326387b

40 files changed

Lines changed: 18 additions & 38 deletions
Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,4 @@
11
---
2-
"@razroo/ray": patch
32
---
43

54
Repo-only: bound the VPS deploy Bun bootstrap download with curl retry, connect, and overall timeouts, and guard curl-to-shell workflow installers in runtime coverage.

.changeset/deploy-apt-timeouts.md

Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,4 @@
11
---
2-
"@razroo/ray": patch
32
---
43

54
Repo-only: bound VPS deploy apt-get bootstrap calls with retry, lock, and overall timeouts, and guard those workflow package installs in runtime coverage.
Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,4 @@
11
---
2-
"@razroo/ray": patch
32
---
43

54
Repo-only: refuse GitHub VPS deploys that would install Caddy for a resolved config with auth disabled, and guard that workflow safety check in runtime coverage.
Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,4 @@
11
---
2-
"ray": patch
32
---
43

54
Add a strict deploy doctor preflight that verifies Caddy is installed before relying on the generated reverse proxy config.
Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,4 @@
11
---
2-
"@razroo/ray": patch
32
---
43

54
Repo-only: validate the generated Caddyfile with the installed Caddy runtime during strict deploy doctor runs so reverse-proxy syntax or runtime incompatibilities fail before installing or reloading Caddy on a VPS.
Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,4 @@
11
---
2-
"@razroo/ray": patch
32
---
43

54
Repo-only: check the configured llama.cpp model file for a GGUF header during strict deploy doctor runs so empty or corrupt model artifacts fail before restarting inference services.
Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,4 @@
11
---
2-
"ray": patch
32
---
43

54
Add a strict deploy doctor probe that runs `llama-server --help` to catch wrong-architecture binaries and missing shared libraries before restarting the generated backend service.
Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,4 @@
11
---
2-
"ray": patch
32
---
43

54
Add a strict deploy doctor check that catches ARM CAX11 versus x64 CX23 sub-1B launch profile architecture mismatches before staging llama.cpp binaries.
Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,4 @@
11
---
2-
"@razroo/ray": patch
32
---
43

54
Repo-only: fail deploy doctor when async queue storage is configured under systemd ProtectSystem read-only roots so durable job writes are kept in service-writable state such as /var/lib/ray/async-queue.
Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,4 @@
11
---
2-
"@razroo/ray": patch
32
---
43

54
Repo-only: treat /var/lib/ray async queue storage as systemd StateDirectory-managed during strict deploy doctor checks so a fresh VPS is not rejected before systemd creates the service state root.

0 commit comments

Comments
 (0)