The fastest, cheapest, most efficient open-source agent harness. Run more than 500 agents on a $10 VPS.
Rust agent-orchestrationai-agentsai-assistantdesktopllmlocal-firstmcppersonal-aiprivacyrustsecond-braintauri
fix(storage): avoid self-deadlock in async backend init
Steven Enamakel0d9f5be89854
diff --git a/crates/openhuman-core/src/storage/local.rs b/crates/openhuman-core/src/storage/local.rsindex 8a09f63472..fb8cb24c01 100644--- a/crates/openhuman-core/src/storage/local.rs+++ b/crates/openhuman-core/src/storage/local.rs@@ -215,14 +215,20 @@ pub fn backend_for(workspace_dir: &Path) -> Result<Arc<dyn StorageBackend>> {/// Async counterpart to [`backend_for`] for boot and module-proxy paths that/// already run inside a future. Cold initialization is delegated to Tokio's/// blocking pool so the shared OnceCell initializer is driven by the storage-/// bridge rather than by the caller's runtime. A synchronous store may wait for-/// that same initializer, including from a current-thread runtime, without+/// bridge rather than by the caller's runtime. When already on the bridge's+/// runtime, initialize directly: queueing a synchronous bridge call from a+/// bridge worker would wait on itself. A synchronous store may wait for the+/// same initializer, including from a current-thread runtime, without/// preventing it from making progress.pub async fn backend_for_async(workspace_dir: &Path) -> Result<Arc<dyn StorageBackend>> {if let Some(backend) = installed() {return Ok(backend);}let workspace_dir = workspace_dir.to_path_buf();+ if super::ON_STORAGE_RUNTIME.with(std::cell::Cell::get) {+ let cell = workspace_backend(&workspace_dir);+ return initialized_backend(cell, workspace_dir).await;+ }tokio::task::spawn_blocking(move || backend_for(&workspace_dir)).await.context("join the workspace storage opener")?