chore: add loadtest observability setup

server-rs/crates/api-server/src/config.rs

@@ -20,7 +20,10 @@ pub(crate) const DEFAULT_VECTOR_ENGINE_IMAGE_REQUEST_TIMEOUT_MS: u64 = 1_000_000
 pub struct AppConfig {
     pub bind_host: String,
     pub bind_port: u16,
+    pub listen_backlog: i32,
+    pub worker_threads: Option<usize>,
     pub log_filter: String,
+    pub otel_enabled: bool,
     pub admin_username: Option<String>,
     pub admin_password: Option<String>,
     pub admin_token_ttl_seconds: u64,
@@ -147,7 +150,10 @@ impl Default for AppConfig {
         Self {
             bind_host: "127.0.0.1".to_string(),
             bind_port: 3000,
+            listen_backlog: 1024,
+            worker_threads: None,
             log_filter: "info,tower_http=info".to_string(),
+            otel_enabled: false,
             admin_username: None,
             admin_password: None,
             admin_token_ttl_seconds: 4 * 60 * 60,
@@ -301,6 +307,17 @@ impl AppConfig {
         {
             config.log_filter = log_filter;
         }
+        if let Some(listen_backlog) =
+            read_first_positive_i32_env(&["GENARRATIVE_API_LISTEN_BACKLOG"])
+        {
+            config.listen_backlog = listen_backlog;
+        }
+        if let Some(worker_threads) = read_first_usize_env(&["GENARRATIVE_API_WORKER_THREADS"]) {
+            config.worker_threads = Some(worker_threads);
+        }
+        if let Some(otel_enabled) = read_first_bool_env(&["GENARRATIVE_OTEL_ENABLED"]) {
+            config.otel_enabled = otel_enabled;
+        }
 
         config.admin_username = read_first_non_empty_env(&["GENARRATIVE_ADMIN_USERNAME"]);
         config.admin_password = read_first_non_empty_env(&["GENARRATIVE_ADMIN_PASSWORD"]);
@@ -881,6 +898,14 @@ fn read_first_positive_u32_env(keys: &[&str]) -> Option<u32> {
     })
 }
 
+fn read_first_positive_i32_env(keys: &[&str]) -> Option<i32> {
+    keys.iter().find_map(|key| {
+        env::var(key)
+            .ok()
+            .and_then(|value| parse_positive_i32(&value))
+    })
+}
+
 fn read_first_positive_u64_env(keys: &[&str]) -> Option<u64> {
     keys.iter().find_map(|key| {
         env::var(key)
@@ -971,6 +996,15 @@ fn parse_positive_u32(raw: &str) -> Option<u32> {
     Some(value)
 }
 
+fn parse_positive_i32(raw: &str) -> Option<i32> {
+    let value = raw.trim().parse::<i32>().ok()?;
+    if value <= 0 {
+        return None;
+    }
+
+    Some(value)
+}
+
 fn parse_u32(raw: &str) -> Option<u32> {
     raw.trim().parse::<u32>().ok()
 }
@@ -1151,6 +1185,34 @@ mod tests {
         }
     }
 
+    #[test]
+    fn from_env_reads_api_runtime_performance_settings() {
+        let _guard = ENV_LOCK
+            .get_or_init(|| Mutex::new(()))
+            .lock()
+            .expect("env lock should not poison");
+
+        unsafe {
+            std::env::remove_var("GENARRATIVE_API_LISTEN_BACKLOG");
+            std::env::remove_var("GENARRATIVE_API_WORKER_THREADS");
+            std::env::remove_var("GENARRATIVE_OTEL_ENABLED");
+            std::env::set_var("GENARRATIVE_API_LISTEN_BACKLOG", "2048");
+            std::env::set_var("GENARRATIVE_API_WORKER_THREADS", "6");
+            std::env::set_var("GENARRATIVE_OTEL_ENABLED", "true");
+        }
+
+        let config = AppConfig::from_env();
+        assert_eq!(config.listen_backlog, 2048);
+        assert_eq!(config.worker_threads, Some(6));
+        assert!(config.otel_enabled);
+
+        unsafe {
+            std::env::remove_var("GENARRATIVE_API_LISTEN_BACKLOG");
+            std::env::remove_var("GENARRATIVE_API_WORKER_THREADS");
+            std::env::remove_var("GENARRATIVE_OTEL_ENABLED");
+        }
+    }
+
     #[test]
     fn from_env_reads_wechat_pay_settings() {
         let _guard = ENV_LOCK