Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.
Python agent-contextagent-memoryagent-pluginsagentic-ragcontext-databasedsh-pluginknowledge-basememorymemory-managementself-evolvingvector-database
//! Synchronous execution bridge for CacheRuntime.
use super::{CacheError, CacheResult};
use std::future::Future;
use std::sync::mpsc;
type Job = Box<dyn FnOnce(&tokio::runtime::Runtime) + Send + 'static>;
pub(crate) struct RuntimeExecutor {
sender: mpsc::Sender<Job>,
}
impl RuntimeExecutor {
pub(crate) fn new() -> CacheResult<Self> {
let (sender, receiver) = mpsc::channel::<Job>();
let (ready_sender, ready_receiver) = mpsc::sync_channel(1);
std::thread::Builder::new()
.name("ragfs-cache-runtime".into())
.spawn(move || {
let runtime = tokio::runtime::Builder::new_multi_thread()
.worker_threads(1)
.enable_all()
.build();
match runtime {
Ok(runtime) => {
let _ = ready_sender.send(Ok(()));
while let Ok(job) = receiver.recv() {
job(&runtime);
}
}
Err(error) => {
let _ = ready_sender.send(Err(error.to_string()));
}
}
})
.map_err(|error| CacheError::Internal(error.to_string()))?;
ready_receiver
.recv()
.map_err(|_| CacheError::Internal("cache runtime executor failed to start".into()))?
.map_err(CacheError::Internal)?;
Ok(Self { sender })
}
pub(crate) fn run<T, F>(&self, future: F) -> CacheResult<T>
where
T: Send + 'static,
F: Future<Output = CacheResult<T>> + Send + 'static,
{
if tokio::runtime::Handle::try_current().is_ok() {
return Err(CacheError::InvalidExecutionContext);
}
let (result_sender, result_receiver) = mpsc::sync_channel(1);
self.sender
.send(Box::new(move |runtime| {
let _ = result_sender.send(runtime.block_on(future));
}))
.map_err(|_| CacheError::Closed)?;
result_receiver
.recv()
.map_err(|_| CacheError::Internal("cache runtime executor stopped".into()))?
}
}