Improve jsonrpc client to properly route replies

Signed-off-by: Lee Smet <lee.smet@hotmail.com>

src/clients/supervisor_client.rs

@@ -1,20 +1,20 @@
 use std::sync::Arc;
 use std::sync::atomic::{AtomicU64, Ordering};
+use std::time::Duration;
 
 use base64::Engine;
 use base64::engine::general_purpose::STANDARD as BASE64_STANDARD;
 use serde_json::{Value, json};
 use thiserror::Error;
+use tokio::time::timeout;
 
-use crate::clients::{Destination, MyceliumClient, MyceliumClientError};
+use crate::clients::{Destination, MyceliumClient, MyceliumClientError, SupervisorHub};
 
 #[derive(Clone)]
 pub struct SupervisorClient {
-    mycelium: Arc<MyceliumClient>, // Delegated Mycelium transport
-    destination: Destination,      // ip or pk
-    topic: String,                 // e.g. "supervisor.rpc"
-    secret: Option<String>,        // optional, required by several supervisor methods
-    id_counter: Arc<AtomicU64>,    // JSON-RPC id generator (for inner supervisor requests)
+    hub: Arc<SupervisorHub>,        // Global hub with background pop loop and shared id generator
+    destination: Destination,       // ip or pk
+    secret: Option<String>,         // optional, required by several supervisor methods
 }
 
 #[derive(Debug, Error)]
@@ -46,24 +46,22 @@ impl From<MyceliumClientError> for SupervisorClientError {
 }
 
 impl SupervisorClient {
-    /// Preferred constructor: provide a shared Mycelium client.
-    pub fn new_with_client(
-        mycelium: Arc<MyceliumClient>,
+    /// Preferred constructor using a shared SupervisorHub (single global listener).
+    pub fn new_with_hub(
+        hub: Arc<SupervisorHub>,
         destination: Destination,
-        topic: impl Into<String>,
         secret: Option<String>,
     ) -> Self {
         Self {
-            mycelium,
+            hub,
             destination,
-            topic: topic.into(),
             secret,
-            id_counter: Arc::new(AtomicU64::new(1)),
         }
     }
 
-    /// Backward-compatible constructor that builds a Mycelium client from base_url.
-    /// base_url defaults to Mycelium spec "http://127.0.0.1:8990" if empty.
+    /// Backward-compatible constructor that builds a new Hub from base_url/topic.
+    /// NOTE: This spawns a background popMessage listener for the given topic.
+    /// Prefer `new_with_hub` so the process has a single global hub.
     pub fn new(
         base_url: impl Into<String>,
         destination: Destination,
@@ -78,8 +76,16 @@ impl SupervisorClient {
         Ok(Self::new_with_client(mycelium, destination, topic, secret))
     }
 
-    fn next_id(&self) -> u64 {
-        self.id_counter.fetch_add(1, Ordering::Relaxed)
+    /// Backward-compatible constructor that reuses an existing Mycelium client.
+    /// NOTE: This creates a new hub and its own background listener. Prefer `new_with_hub`.
+    pub fn new_with_client(
+        mycelium: Arc<MyceliumClient>,
+        destination: Destination,
+        topic: impl Into<String>,
+        secret: Option<String>,
+    ) -> Self {
+        let hub = SupervisorHub::new_with_client(mycelium, topic);
+        Self::new_with_hub(hub, destination, secret)
     }
 
     /// Internal helper used by tests to inspect dst JSON shape.
@@ -93,7 +99,7 @@ impl SupervisorClient {
     fn build_supervisor_payload(&self, method: &str, params: Value) -> Value {
         json!({
             "jsonrpc": "2.0",
-            "id": self.next_id(),
+            "id": self.hub.next_id(),
             "method": method,
             "params": params,
         })
@@ -128,50 +134,37 @@ impl SupervisorClient {
             .map(|s| s.to_string())
     }
 
-    /// Generic call: build supervisor JSON-RPC message, send via Mycelium pushMessage, return outbound message id (hex).
-    pub async fn call(&self, method: &str, params: Value) -> Result<String, SupervisorClientError> {
-        let inner = self.build_supervisor_payload(method, params);
-        let payload_b64 = Self::encode_payload(&inner)?;
-        let result = self
-            .mycelium
-            .push_message(
-                &self.destination,
-                &Self::encode_topic(self.topic.as_bytes()),
-                &payload_b64,
-                None,
-            )
-            .await?;
-
-        if let Some(id) = MyceliumClient::extract_message_id_from_result(&result) {
-            return Ok(id);
-        }
-        // Some servers might return the oneOf wrapped, handle len==1 array defensively (not in spec but resilient)
-        if let Some(arr) = result.as_array()
-            && arr.len() == 1
-            && let Some(id) = MyceliumClient::extract_message_id_from_result(&arr[0])
-        {
-            return Ok(id);
-        }
-        Err(SupervisorClientError::InvalidResponse(format!(
-            "result did not contain message id: {result}"
-        )))
+    fn need_secret(&self) -> Result<&str, SupervisorClientError> {
+        self.secret
+            .as_deref()
+            .ok_or(SupervisorClientError::MissingSecret)
     }
 
-    /// Variant of call that also returns the inner supervisor JSON-RPC id used in the payload.
-    /// This id is required to correlate asynchronous popMessage replies coming from Mycelium.
-    pub async fn call_with_ids(
+    // -----------------------------
+    // Core: request-reply call via Hub with default 10s timeout
+    // -----------------------------
+
+    /// Send a supervisor JSON-RPC request and await its reply via the Hub.
+    /// Returns (outbound_message_id, reply_envelope_json).
+    pub async fn call_with_reply_timeout(
         &self,
         method: &str,
         params: Value,
-    ) -> Result<(String, u64), SupervisorClientError> {
-        let inner_id = self.next_id();
+        timeout_secs: u64,
+    ) -> Result<(String, Value), SupervisorClientError> {
+        let inner_id = self.hub.next_id();
+        // Register waiter before sending to avoid race
+        let rx = self.hub.register_waiter(inner_id).await;
+
         let inner = self.build_supervisor_payload_with_id(method, params, inner_id);
         let payload_b64 = Self::encode_payload(&inner)?;
+
         let result = self
-            .mycelium
+            .hub
+            .mycelium()
             .push_message(
                 &self.destination,
-                &Self::encode_topic(self.topic.as_bytes()),
+                &Self::encode_topic(self.hub.topic().as_bytes()),
                 &payload_b64,
                 None,
             )
@@ -185,29 +178,210 @@ impl SupervisorClient {
         {
             id
         } else {
+            // Clean pending entry to avoid leak
+            let _ = self.hub.remove_waiter(inner_id).await;
             return Err(SupervisorClientError::InvalidResponse(format!(
                 "result did not contain message id: {result}"
             )));
         };
 
-        Ok((out_id, inner_id))
+        let d = Duration::from_secs(timeout_secs);
+        match timeout(d, rx).await {
+            Ok(Ok(reply)) => Ok((out_id, reply)),
+            Ok(Err(_canceled)) => Err(SupervisorClientError::InvalidResponse(
+                "oneshot canceled before receiving reply".into(),
+            )),
+            Err(_elapsed) => {
+                // Cleanup on timeout
+                let _ = self.hub.remove_waiter(inner_id).await;
+                Err(SupervisorClientError::TransportTimeout)
+            }
+        }
     }
 
+    /// Send and await with default 10s timeout.
+    pub async fn call_with_reply(
+        &self,
+        method: &str,
+        params: Value,
+    ) -> Result<(String, Value), SupervisorClientError> {
+        self.call_with_reply_timeout(method, params, 60).await
+    }
 
-    fn need_secret(&self) -> Result<&str, SupervisorClientError> {
-        self.secret
-            .as_deref()
-            .ok_or(SupervisorClientError::MissingSecret)
+    /// Back-compat: Send and await a reply but return only the outbound id (discard reply).
+    /// This keeps existing call sites working while the system migrates to reply-aware paths.
+    pub async fn call(&self, method: &str, params: Value) -> Result<String, SupervisorClientError> {
+        let (out_id, _reply) = self.call_with_reply(method, params).await?;
+        Ok(out_id)
     }
 
     // -----------------------------
-    // Typed wrappers for Supervisor API
-    // Asynchronous-only: returns outbound message id
+    // Typed wrappers for Supervisor API (await replies)
     // -----------------------------
 
     // Runners
+    pub async fn list_runners_wait(&self) -> Result<(String, Value), SupervisorClientError> {
+        self.call_with_reply("list_runners", json!([])).await
+    }
+
+    pub async fn register_runner_wait(
+        &self,
+        name: impl Into<String>,
+        queue: impl Into<String>,
+    ) -> Result<(String, Value), SupervisorClientError> {
+        let secret = self.need_secret()?;
+        let params = json!([{
+            "secret": secret,
+            "name": name.into(),
+            "queue": queue.into()
+        }]);
+        self.call_with_reply("register_runner", params).await
+    }
+
+    pub async fn remove_runner_wait(
+        &self,
+        actor_id: impl Into<String>,
+    ) -> Result<(String, Value), SupervisorClientError> {
+        self.call_with_reply("remove_runner", json!([actor_id.into()]))
+            .await
+    }
+
+    pub async fn start_runner_wait(
+        &self,
+        actor_id: impl Into<String>,
+    ) -> Result<(String, Value), SupervisorClientError> {
+        self.call_with_reply("start_runner", json!([actor_id.into()]))
+            .await
+    }
+
+    pub async fn stop_runner_wait(
+        &self,
+        actor_id: impl Into<String>,
+        force: bool,
+    ) -> Result<(String, Value), SupervisorClientError> {
+        self.call_with_reply("stop_runner", json!([actor_id.into(), force]))
+            .await
+    }
+
+    pub async fn get_runner_status_wait(
+        &self,
+        actor_id: impl Into<String>,
+    ) -> Result<(String, Value), SupervisorClientError> {
+        self.call_with_reply("get_runner_status", json!([actor_id.into()]))
+            .await
+    }
+
+    pub async fn get_all_runner_status_wait(
+        &self,
+    ) -> Result<(String, Value), SupervisorClientError> {
+        self.call_with_reply("get_all_runner_status", json!([]))
+            .await
+    }
+
+    pub async fn start_all_wait(&self) -> Result<(String, Value), SupervisorClientError> {
+        self.call_with_reply("start_all", json!([])).await
+    }
+
+    pub async fn stop_all_wait(
+        &self,
+        force: bool,
+    ) -> Result<(String, Value), SupervisorClientError> {
+        self.call_with_reply("stop_all", json!([force])).await
+    }
+
+    pub async fn get_all_status_wait(&self) -> Result<(String, Value), SupervisorClientError> {
+        self.call_with_reply("get_all_status", json!([])).await
+    }
+
+    // Jobs (await)
+    pub async fn jobs_create_wait(
+        &self,
+        job: Value,
+    ) -> Result<(String, Value), SupervisorClientError> {
+        let secret = self.need_secret()?;
+        let params = json!([{
+            "secret": secret,
+            "job": job
+        }]);
+        self.call_with_reply("jobs.create", params).await
+    }
+
+    pub async fn jobs_list_wait(&self) -> Result<(String, Value), SupervisorClientError> {
+        self.call_with_reply("jobs.list", json!([])).await
+    }
+
+    pub async fn job_run_wait(&self, job: Value) -> Result<(String, Value), SupervisorClientError> {
+        let secret = self.need_secret()?;
+        let params = json!([{
+            "secret": secret,
+            "job": job
+        }]);
+        self.call_with_reply("job.run", params).await
+    }
+
+    pub async fn job_start_wait(
+        &self,
+        job_id: impl Into<String>,
+    ) -> Result<(String, Value), SupervisorClientError> {
+        let secret = self.need_secret()?;
+        let params = json!([{
+            "secret": secret,
+            "job_id": job_id.into()
+        }]);
+        self.call_with_reply("job.start", params).await
+    }
+
+    pub async fn job_status_wait(
+        &self,
+        job_id: impl Into<String>,
+    ) -> Result<(String, Value), SupervisorClientError> {
+        self.call_with_reply("job.status", json!([job_id.into()]))
+            .await
+    }
+
+    pub async fn job_result_wait(
+        &self,
+        job_id: impl Into<String>,
+    ) -> Result<(String, Value), SupervisorClientError> {
+        self.call_with_reply("job.result", json!([job_id.into()]))
+            .await
+    }
+
+    pub async fn job_stop_wait(
+        &self,
+        job_id: impl Into<String>,
+    ) -> Result<(String, Value), SupervisorClientError> {
+        let secret = self.need_secret()?;
+        let params = json!([{
+            "secret": secret,
+            "job_id": job_id.into()
+        }]);
+        self.call_with_reply("job.stop", params).await
+    }
+
+    pub async fn job_delete_wait(
+        &self,
+        job_id: impl Into<String>,
+    ) -> Result<(String, Value), SupervisorClientError> {
+        let secret = self.need_secret()?;
+        let params = json!([{
+            "secret": secret,
+            "job_id": job_id.into()
+        }]);
+        self.call_with_reply("job.delete", params).await
+    }
+
+    pub async fn rpc_discover_wait(&self) -> Result<(String, Value), SupervisorClientError> {
+        self.call_with_reply("rpc.discover", json!([])).await
+    }
+
+    // -----------------------------
+    // Backward-compatible variants returning only outbound id (discarding reply)
+    // -----------------------------
+
     pub async fn list_runners(&self) -> Result<String, SupervisorClientError> {
-        self.call("list_runners", json!([])).await
+        let (id, _) = self.list_runners_wait().await?;
+        Ok(id)
     }
 
     pub async fn register_runner(
@@ -215,27 +389,24 @@ impl SupervisorClient {
         name: impl Into<String>,
         queue: impl Into<String>,
     ) -> Result<String, SupervisorClientError> {
-        let secret = self.need_secret()?;
-        let params = json!([{
-            "secret": secret,
-            "name": name.into(),
-            "queue": queue.into()
-        }]);
-        self.call("register_runner", params).await
+        let (id, _) = self.register_runner_wait(name, queue).await?;
+        Ok(id)
     }
 
     pub async fn remove_runner(
         &self,
         actor_id: impl Into<String>,
     ) -> Result<String, SupervisorClientError> {
-        self.call("remove_runner", json!([actor_id.into()])).await
+        let (id, _) = self.remove_runner_wait(actor_id).await?;
+        Ok(id)
     }
 
     pub async fn start_runner(
         &self,
         actor_id: impl Into<String>,
     ) -> Result<String, SupervisorClientError> {
-        self.call("start_runner", json!([actor_id.into()])).await
+        let (id, _) = self.start_runner_wait(actor_id).await?;
+        Ok(id)
     }
 
     pub async fn stop_runner(
@@ -243,141 +414,96 @@ impl SupervisorClient {
         actor_id: impl Into<String>,
         force: bool,
     ) -> Result<String, SupervisorClientError> {
-        self.call("stop_runner", json!([actor_id.into(), force]))
-            .await
+        let (id, _) = self.stop_runner_wait(actor_id, force).await?;
+        Ok(id)
     }
 
     pub async fn get_runner_status(
         &self,
         actor_id: impl Into<String>,
     ) -> Result<String, SupervisorClientError> {
-        self.call("get_runner_status", json!([actor_id.into()]))
-            .await
+        let (id, _) = self.get_runner_status_wait(actor_id).await?;
+        Ok(id)
     }
 
     pub async fn get_all_runner_status(&self) -> Result<String, SupervisorClientError> {
-        self.call("get_all_runner_status", json!([])).await
+        let (id, _) = self.get_all_runner_status_wait().await?;
+        Ok(id)
     }
 
     pub async fn start_all(&self) -> Result<String, SupervisorClientError> {
-        self.call("start_all", json!([])).await
+        let (id, _) = self.start_all_wait().await?;
+        Ok(id)
     }
 
     pub async fn stop_all(&self, force: bool) -> Result<String, SupervisorClientError> {
-        self.call("stop_all", json!([force])).await
+        let (id, _) = self.stop_all_wait(force).await?;
+        Ok(id)
     }
 
     pub async fn get_all_status(&self) -> Result<String, SupervisorClientError> {
-        self.call("get_all_status", json!([])).await
+        let (id, _) = self.get_all_status_wait().await?;
+        Ok(id)
     }
 
-    // Jobs
     pub async fn jobs_create(&self, job: Value) -> Result<String, SupervisorClientError> {
-        let secret = self.need_secret()?;
-        let params = json!([{
-            "secret": secret,
-            "job": job
-        }]);
-        self.call("jobs.create", params).await
+        let (id, _) = self.jobs_create_wait(job).await?;
+        Ok(id)
     }
 
     pub async fn jobs_list(&self) -> Result<String, SupervisorClientError> {
-        self.call("jobs.list", json!([])).await
+        let (id, _) = self.jobs_list_wait().await?;
+        Ok(id)
     }
 
     pub async fn job_run(&self, job: Value) -> Result<String, SupervisorClientError> {
-        let secret = self.need_secret()?;
-        let params = json!([{
-            "secret": secret,
-            "job": job
-        }]);
-        self.call("job.run", params).await
-    }
-
-    /// Typed wrapper returning both outbound Mycelium id and inner supervisor JSON-RPC id.
-    pub async fn job_run_with_ids(
-        &self,
-        job: Value,
-    ) -> Result<(String, u64), SupervisorClientError> {
-        let secret = self.need_secret()?;
-        let params = json!([{
-            "secret": secret,
-            "job": job
-        }]);
-        self.call_with_ids("job.run", params).await
+        let (id, _) = self.job_run_wait(job).await?;
+        Ok(id)
     }
 
     pub async fn job_start(
         &self,
         job_id: impl Into<String>,
     ) -> Result<String, SupervisorClientError> {
-        let secret = self.need_secret()?;
-        let params = json!([{
-            "secret": secret,
-            "job_id": job_id.into()
-        }]);
-        self.call("job.start", params).await
+        let (id, _) = self.job_start_wait(job_id).await?;
+        Ok(id)
     }
 
     pub async fn job_status(
         &self,
         job_id: impl Into<String>,
     ) -> Result<String, SupervisorClientError> {
-        self.call("job.status", json!([job_id.into()])).await
+        let (id, _) = self.job_status_wait(job_id).await?;
+        Ok(id)
     }
 
-    /// Asynchronous job.status returning outbound and inner IDs for correlation
-    pub async fn job_status_with_ids(
-        &self,
-        job_id: impl Into<String>,
-    ) -> Result<(String, u64), SupervisorClientError> {
-        self.call_with_ids("job.status", json!([job_id.into()])).await
-    }
-
-
     pub async fn job_result(
         &self,
         job_id: impl Into<String>,
     ) -> Result<String, SupervisorClientError> {
-        self.call("job.result", json!([job_id.into()])).await
+        let (id, _) = self.job_result_wait(job_id).await?;
+        Ok(id)
     }
 
-    /// Asynchronous job.result returning outbound and inner IDs for correlation
-    pub async fn job_result_with_ids(
-        &self,
-        job_id: impl Into<String>,
-    ) -> Result<(String, u64), SupervisorClientError> {
-        self.call_with_ids("job.result", json!([job_id.into()])).await
-    }
-
-
     pub async fn job_stop(
         &self,
         job_id: impl Into<String>,
     ) -> Result<String, SupervisorClientError> {
-        let secret = self.need_secret()?;
-        let params = json!([{
-            "secret": secret,
-            "job_id": job_id.into()
-        }]);
-        self.call("job.stop", params).await
+        let (id, _) = self.job_stop_wait(job_id).await?;
+        Ok(id)
     }
 
     pub async fn job_delete(
         &self,
         job_id: impl Into<String>,
     ) -> Result<String, SupervisorClientError> {
-        let secret = self.need_secret()?;
-        let params = json!([{
-            "secret": secret,
-            "job_id": job_id.into()
-        }]);
-        self.call("job.delete", params).await
+        let (id, _) = self.job_delete_wait(job_id).await?;
+        Ok(id)
     }
 
-    // Discovery
     pub async fn rpc_discover(&self) -> Result<String, SupervisorClientError> {
-        self.call("rpc.discover", json!([])).await
+        let (id, _) = self.rpc_discover_wait().await?;
+        Ok(id)
     }
 }
 
@@ -390,27 +516,27 @@ mod tests {
     use std::net::IpAddr;
 
     fn mk_client() -> SupervisorClient {
-        // Uses the legacy constructor but will not issue real network calls in these tests.
-        SupervisorClient::new(
-            "http://127.0.0.1:8990",
+        // Build a hub but it won't issue real network calls in these serializer-only tests.
+        let mycelium = Arc::new(MyceliumClient::new("http://127.0.0.1:8990").unwrap());
+        let hub = SupervisorHub::new_with_client(mycelium, "supervisor.rpc");
+        SupervisorClient::new_with_hub(
+            hub,
             Destination::Pk(
                 "bb39b4a3a4efd70f3e05e37887677e02efbda14681d0acd3882bc0f754792c32".to_string(),
             ),
-            "supervisor.rpc",
             Some("secret".to_string()),
         )
-        .unwrap()
     }
 
     #[test]
     fn builds_dst_ip_and_pk() {
-        let c_ip = SupervisorClient::new(
-            "http://127.0.0.1:8990",
+        let mycelium = Arc::new(MyceliumClient::new("http://127.0.0.1:8990").unwrap());
+        let hub_ip = SupervisorHub::new_with_client(mycelium.clone(), "supervisor.rpc");
+        let c_ip = SupervisorClient::new_with_hub(
+            hub_ip,
             Destination::Ip("2001:db8::1".parse().unwrap()),
-            "supervisor.rpc",
             None,
-        )
-        .unwrap();
+        );
         let v_ip = c_ip.build_dst();
         assert_eq!(v_ip.get("ip").unwrap().as_str().unwrap(), "2001:db8::1");