experiment: reduce keepalive to 5s and add progress bar

EXPERIMENT (NOT FOR PRODUCTION YET)
Changes:
- KEEPALIVE_INTERVAL_SECONDS reduced from 20s to 5s
- Keepalive messages now show progress bar with dots: ●●●●○○○○○○
- Show elapsed time, total wait, and remaining seconds
- Example: ⏳ Waiting for instructions... ●●●●○○○○○○ 20s / 50s (agent=copilot, 30s remaining)
Goal: Test if more frequent progress updates provide better UX and prevent
      perceived freezing during the 50s wait. No functional change - the
      60s client timeout limit remains the binding constraint.

app/mcp_server.py

@@ -33,6 +33,14 @@ mcp_asgi_app = mcp.streamable_http_app()
 # asyncio.Event wakeup, so there is no practical danger in long waits.
 _MAX_WAIT_SECONDS = 86400  # 24 hours
 
+# Default wait time when no instruction is available (seconds)
+# Set to 50s to stay safely under the 60s client timeout while allowing
+# multiple keepalive progress updates.
+DEFAULT_WAIT_SECONDS = 50
+
+# Default response when queue is empty after waiting
+DEFAULT_EMPTY_RESPONSE = "call this tool `get_user_request` again to fetch latest user input..."
+
 # Per-agent generation counter — incremented on every new call.
 # The wait loop only consumes an instruction when it holds the latest generation,
 # preventing abandoned (timed-out) coroutines from silently consuming queue items.
@@ -50,7 +58,8 @@ _agent_generations: dict[str, int] = {}
 # start and are unaffected by intermediate SSE events.
 #
 # Set to 0 to disable keepalives entirely.
-KEEPALIVE_INTERVAL_SECONDS: float = 20.0
+# EXPERIMENT: Reduced from 20.0 to 5.0 for more frequent progress updates
+KEEPALIVE_INTERVAL_SECONDS: float = 5.0
 
 
 @mcp.tool()
@@ -74,10 +83,8 @@ async def get_user_request(
         A dict with keys: status, result_type, instruction, response,
         remaining_pending, waited_seconds.
     """
-    cfg = config_service.get_config()
+    # Wait time is hardcoded to stay safely under the 60s client timeout
-
+    actual_wait = min(DEFAULT_WAIT_SECONDS, _MAX_WAIT_SECONDS)
-    # Wait time is entirely server-controlled — the user sets it via the web UI.
-    actual_wait = min(cfg.default_wait_seconds, _MAX_WAIT_SECONDS)
 
     # Register this call as the newest for this agent.
     my_gen = _agent_generations.get(agent_id, 0) + 1
@@ -171,14 +178,19 @@ async def get_user_request(
         if KEEPALIVE_INTERVAL_SECONDS > 0 and ctx is not None:
             if now - last_keepalive >= KEEPALIVE_INTERVAL_SECONDS:
                 waited_so_far = int(now - start)
+                remaining_sec = max(0, actual_wait - waited_so_far)
+                # Progress bar: filled dots proportional to elapsed time
+                progress_pct = min(100, int((waited_so_far / actual_wait) * 100))
+                filled = int(progress_pct / 10)
+                bar = "●" * filled + "○" * (10 - filled)
                 try:
                     await ctx.info(
-                        f"keepalive: waiting for instructions "
+                        f"⏳ Waiting for instructions... {bar} "
-                        f"(agent={agent_id}, waited={waited_so_far}s)"
+                        f"{waited_so_far}s / {actual_wait}s (agent={agent_id}, {remaining_sec}s remaining)"
                     )
                     logger.debug(
-                        "get_user_request: keepalive sent agent=%s waited=%ds",
+                        "get_user_request: keepalive sent agent=%s waited=%ds progress=%d%%",
-                        agent_id, waited_so_far,
+                        agent_id, waited_so_far, progress_pct,
                     )
                 except Exception as exc:
                     # Client disconnected — no point continuing
@@ -213,7 +225,7 @@ async def get_user_request(
     empty_response = (
         default_response_override
         if default_response_override is not None
-        else cfg.default_empty_response
+        else DEFAULT_EMPTY_RESPONSE
     )
     result_type = "default_response" if empty_response else "empty"