ci(prerelease): auto-build on every push to main

The build-prerelease workflow was workflow_dispatch-only, which meant
every commit needed a manual run dispatch before any host could
upgrade. That left rolling fixes (e.g. f9f5fa4's StateDirectory fix)
sitting on main with no published RPM behind them, so deploy.sh
silently fell back to an older prerelease.

Add 'push: branches: [main]' alongside the existing workflow_dispatch
trigger; the unstable channel now tracks head automatically. The
concurrency group is keyed on ${{ github.ref }} with
cancel-in-progress so successive rapid-fire pushes coalesce to one
build (latest wins) rather than queueing every intermediate commit.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

.gitea/workflows/build-prerelease.yml

@@ -12,6 +12,11 @@ name: build-prerelease
 # so they sort BELOW the eventual 0.1.16-1 stable release.
 
 on:
+  # Auto-build on every push to main so the unstable channel tracks
+  # head without a manual dispatch step.
+  push:
+    branches: [main]
+  # Manual dispatch still available to build from a non-main ref.
   workflow_dispatch:
     inputs:
       ref:
@@ -20,7 +25,9 @@ on:
         default: ""
 
 concurrency:
-  group: prerelease-build
+  # Coalesce on branch+event so successive pushes don't pile up; the
+  # latest push wins.
+  group: prerelease-build-${{ github.ref }}
   cancel-in-progress: true
 
 env:

crates/neuron/src/api.rs

@@ -69,11 +69,22 @@ async fn load_model(
     let registry = state.registry.read().await;
     match registry.load_model(&spec).await {
         Ok(()) => Json(json!({"status": "loaded"})).into_response(),
-        Err(e) => (
+        Err(e) => {
+            // Log the full anyhow chain server-side so journalctl shows
+            // the underlying failure (hf-hub timeout, permission denied,
+            // disk full, etc.) without needing to inspect the HTTP
+            // response body separately.
+            tracing::warn!(
+                model = %spec.model_id,
+                error = %format!("{e:#}"),
+                "load_model failed"
+            );
+            (
                 StatusCode::BAD_REQUEST,
                 Json(json!({"error": format!("{e:#}")})),
             )
-            .into_response(),
+                .into_response()
+        }
     }
 }