ework-web 0.10.126 → 0.10.128

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ework-web",
-  "version": "0.10.126",
+  "version": "0.10.128",
   "type": "module",
   "description": "ework-web — standalone multi-project issue tracker. Local SQLite-backed, no external API dependency. Bun + TypeScript + SSR HTML.",
   "license": "MIT",

package/src/index.ts

@@ -420,6 +420,9 @@ const SESSION_BATCH_RE = /^\/api\/sessions\/([A-Za-z0-9_-]+)\/batch$/;
 const server = Bun.serve({
   port: cfg.port,
   hostname: cfg.host,
+  // Streaming routes (translate, live tails) legitimately sit silent past the
+  // 10s default while upstream queues; 60s matches their in-route heartbeats.
+  idleTimeout: 60,
   async fetch(req, server) {
     const url = new URL(req.url);
     const ip = remoteAddr(req, server);
@@ -1009,6 +1012,16 @@ async function handle(req: Request, url: URL, ip: string, ctx: { authed: boolean
       async start(controller) {
         const enc = new TextEncoder();
         const send = (obj: unknown) => controller.enqueue(enc.encode(JSON.stringify(obj) + "\n"));
+        // Bun closes connections idle >10s by default; under model-server load
+        // the first translation chunk can queue longer than that. Heartbeat
+        // lines ({"s":1}) keep the connection alive and are ignored by clients.
+        const heartbeat = setInterval(() => {
+          try {
+            send({ s: 1 });
+          } catch {
+            /* client disconnected */
+          }
+        }, 5000);
         try {
           let full = "";
           for await (const chunk of translateTextStream(cfg, text)) {
@@ -1019,6 +1032,7 @@ async function handle(req: Request, url: URL, ip: string, ctx: { authed: boolean
         } catch (e) {
           send({ e: e instanceof Error ? e.message : String(e) });
         } finally {
+          clearInterval(heartbeat);
           controller.close();
         }
       },

package/src/translate.ts

@@ -45,6 +45,9 @@ export async function translateText(cfg: Config, text: string): Promise<string>
       body: JSON.stringify({
         model: cfg.translateModel,
         stream: false,
+        // Reasoning-off: translation needs none, and reasoning tokens delay
+        // time-to-first-token past interactive patience on the shared server.
+        chat_template_kwargs: { enable_thinking: false },
         messages: [
           { role: "system", content: SYSTEM_PROMPT },
           { role: "user", content: body },
@@ -121,6 +124,7 @@ async function* streamOneChunk(cfg: Config, chunk: string): AsyncGenerator<strin
       body: JSON.stringify({
         model: cfg.translateModel,
         stream: true,
+        chat_template_kwargs: { enable_thinking: false },
         messages: [
           { role: "system", content: SYSTEM_PROMPT },
           { role: "user", content: chunk },