fix: add SSE streaming proxy and robust make try startup

The SSE proxy was buffering the entire response body with a 30s read
timeout, causing 504s on the OpenCode /global/event stream. Add a
streaming path that detects SSE requests (by Accept header or /event
path) and returns a StreamingResponse with no read timeout.

Also fix the make try target to poll the health endpoint for Docker
readiness and wait for the container to reach running status before
opening the browser.

session-manager/http_pool.py

@@ -159,6 +159,24 @@ async def make_http_request(method: str, url: str, **kwargs) -> httpx.Response:
         return await client.request(method, url, **kwargs)
 
 
+@asynccontextmanager
+async def stream_http_request(method: str, url: str, **kwargs):
+    """Stream an HTTP response using a dedicated client with no read timeout.
+
+    Yields an httpx.Response whose body has NOT been read -- caller must
+    iterate over ``response.aiter_bytes()`` / ``aiter_lines()`` etc.
+
+    A separate AsyncClient is used (not the pool) because httpx's
+    ``stream()`` keeps the connection checked-out for the lifetime of the
+    context manager, and SSE streams are effectively infinite.  Using a
+    short-lived client avoids starving the pool.
+    """
+    timeout = httpx.Timeout(connect=10.0, read=None, write=10.0, pool=5.0)
+    async with httpx.AsyncClient(timeout=timeout, follow_redirects=False) as client:
+        async with client.stream(method, url, **kwargs) as response:
+            yield response
+
+
 async def get_connection_pool_stats() -> Dict[str, Any]:
     """Get connection pool statistics."""
     return await _http_pool.get_pool_stats()

session-manager/routes/proxy.py

@@ -4,10 +4,11 @@ import time
 from urllib.parse import urlparse
 
 from fastapi import APIRouter, HTTPException, Request, Response
+from starlette.responses import StreamingResponse
 import httpx
 
 from session_manager import session_manager
-from http_pool import make_http_request
+from http_pool import make_http_request, stream_http_request
 from logging_config import (
     RequestContext,
     log_request,
@@ -270,6 +271,17 @@ async def proxy_internal_session_to_session(request: Request):
     return await proxy_to_session(request, session_id, path)
 
 
+def _is_sse_request(request: Request, path: str) -> bool:
+    """Detect SSE requests by Accept header or path convention."""
+    accept = request.headers.get("accept", "")
+    if "text/event-stream" in accept:
+        return True
+    # OpenCode uses /global/event and /event paths for SSE
+    if path == "event" or path.endswith("/event"):
+        return True
+    return False
+
+
 @router.api_route("/session/{session_id}/{path:path}", methods=ALL_METHODS)
 async def proxy_to_session(request: Request, session_id: str, path: str):
     start_time = time.time()
@@ -319,6 +331,11 @@ async def proxy_to_session(request: Request, session_id: str, path: str):
         headers["X-Session-Token"] = session.auth_token
         headers["X-Session-ID"] = session.session_id
 
+    # --- SSE streaming path ---
+    if _is_sse_request(request, path):
+        return await _proxy_sse(request, session_id, path, url, headers, body, start_time)
+
+    # --- Buffered path (original behaviour) ---
     try:
         log_session_operation(
             session_id, "proxy_request", method=request.method, path=path
@@ -425,3 +442,71 @@ async def proxy_to_session(request: Request, session_id: str, path: str):
             status_code=502,
             detail=f"Failed to connect to session container: {str(e)}",
         )
+
+
+async def _proxy_sse(
+    request: Request,
+    session_id: str,
+    path: str,
+    url: str,
+    headers: dict,
+    body: bytes,
+    start_time: float,
+):
+    """Proxy an SSE event stream without buffering."""
+    log_session_operation(
+        session_id, "proxy_sse_stream", method=request.method, path=path
+    )
+
+    # We need to keep the httpx stream context alive for the lifetime of the
+    # StreamingResponse.  Starlette calls our async generator and only closes
+    # it when the client disconnects, so we enter the context manager inside
+    # the generator and exit on cleanup.
+
+    async def event_generator():
+        try:
+            async with stream_http_request(
+                method=request.method,
+                url=url,
+                headers=headers,
+                content=body,
+            ) as upstream:
+                async for chunk in upstream.aiter_bytes():
+                    yield chunk
+        except httpx.RequestError as e:
+            log_security_event(
+                "proxy_sse_error",
+                "error",
+                session_id=session_id,
+                method=request.method,
+                path=path,
+                error=str(e),
+            )
+        finally:
+            duration_ms = (time.time() - start_time) * 1000
+            log_request(
+                request.method,
+                f"/session/{session_id}/{path}",
+                200,
+                duration_ms,
+                session_id=session_id,
+                operation="proxy_sse_complete",
+            )
+
+    resp = StreamingResponse(
+        event_generator(),
+        media_type="text/event-stream",
+        headers={
+            "Cache-Control": "no-cache",
+            "Connection": "keep-alive",
+            "X-Accel-Buffering": "no",
+        },
+    )
+    resp.set_cookie(
+        key="lovdata_session",
+        value=session_id,
+        httponly=True,
+        samesite="lax",
+        max_age=86400,
+    )
+    return resp