fix: yaml stdio per-connection, iac catalog path, makefile port

- yaml-language-server: rewrite to stdio per WebSocket (fixes crash loop)
  vscode-jsonrpc v9 createServerSocketTransport is a TCP client, not server
  now spawns yaml-language-server --stdio per connection via asyncio subprocess

- bicep/modules.py: add /iac_source_catalog.json as first path in _IAC_SOURCE_PATHS
  Dockerfile copies to /iac_source_catalog.json but path wasn't listed

- server.py: remove YAML_LSP_PORT daemon (no longer needed with stdio mode)

- Makefile: add -e HTTP_PORT=$(HEALTH_PORT) to all docker run commands
  server defaulted to :8000 but Makefile exposed :2089 with no override

ilsp/yaml_lsp/proxy.py

@@ -273,31 +273,48 @@ class _YamlSession:
 # ── Main WS handler ───────────────────────────────────────────────────────────
 
 
-async def yaml_ws_handler(request: web.Request, yaml_lsp_port: int) -> web.WebSocketResponse:
+async def yaml_ws_handler(request: web.Request, yaml_lsp_port: int = 0) -> web.WebSocketResponse:
     """
     WebSocket handler for the /yaml endpoint.
 
-    Bridges the editor WS connection to yaml-language-server TCP, intercepting
-    completion messages to inject pipeline template completions.
+    Spawns yaml-language-server --stdio per editor connection (one process per
+    session). Bridges WS ↔ process stdin/stdout, intercepting completion messages
+    to inject pipeline template completions.
+
+    Note: yaml_lsp_port is unused — kept for API compatibility.
+    yaml-language-server uses --stdio so no TCP port is needed.
     """
+    import shutil
+
     ws = web.WebSocketResponse()
     await ws.prepare(request)
 
-    try:
-        tcp_reader, tcp_writer = await asyncio.wait_for(
-            asyncio.open_connection("127.0.0.1", yaml_lsp_port), timeout=3.0
-        )
-    except (OSError, asyncio.TimeoutError) as exc:
-        logger.error("Cannot connect to yaml-language-server on port %d: %s", yaml_lsp_port, exc)
-        await ws.close(code=1011, message=b"YAML LSP backend unavailable", timeout=2.0)
+    yaml_ls = shutil.which("yaml-language-server")
+    if not yaml_ls:
+        logger.error("yaml-language-server not found in PATH")
+        await ws.close(code=1011, message=b"yaml-language-server not installed", timeout=2.0)
         return ws
 
+    try:
+        proc = await asyncio.create_subprocess_exec(
+            yaml_ls, "--stdio",
+            stdin=asyncio.subprocess.PIPE,
+            stdout=asyncio.subprocess.PIPE,
+            stderr=asyncio.subprocess.DEVNULL,
+        )
+    except Exception as exc:
+        logger.error("Failed to start yaml-language-server: %s", exc)
+        await ws.close(code=1011, message=b"YAML LSP failed to start", timeout=2.0)
+        return ws
+
+    logger.info("yaml-language-server started (--stdio) PID=%d", proc.pid)
+
     session = _YamlSession()
     ws_buf = _LspFrameBuffer()
-    tcp_buf = _LspFrameBuffer()
+    proc_buf = _LspFrameBuffer()
 
     async def client_to_server() -> None:
-        """WS → TCP: track document content and completion requests."""
+        """WS → stdin: track document content and completion requests."""
         try:
             async for msg in ws:
                 if msg.type not in (WSMsgType.BINARY, WSMsgType.TEXT):
@@ -313,7 +330,6 @@ async def yaml_ws_handler(request: web.Request, yaml_lsp_port: int) -> web.WebSo
                             uri = params.get("textDocument", {}).get("uri", "")
                             text = params.get("textDocument", {}).get("text") or ""
                             if not text:
-                                # didChange has contentChanges
                                 changes = params.get("contentChanges", [])
                                 if changes:
                                     text = changes[-1].get("text", "")
@@ -323,21 +339,25 @@ async def yaml_ws_handler(request: web.Request, yaml_lsp_port: int) -> web.WebSo
                             session.record_request(parsed)
                     except Exception:
                         pass
-                    tcp_writer.write(_lsp_frame(frame))
-                await tcp_writer.drain()
+                    proc.stdin.write(_lsp_frame(frame))
+                await proc.stdin.drain()
         except Exception:
             pass
         finally:
-            tcp_writer.close()
+            try:
+                proc.stdin.close()
+            except Exception:
+                pass
+            proc.kill()
 
     async def server_to_client() -> None:
-        """TCP → WS: inject completions into completion responses."""
+        """stdout → WS: inject completions into completion responses."""
         try:
             while True:
-                data = await tcp_reader.read(_CHUNK)
+                data = await proc.stdout.read(_CHUNK)
                 if not data:
                     break
-                frames = tcp_buf.feed(data)
+                frames = proc_buf.feed(data)
                 for frame in frames:
                     try:
                         parsed = json.loads(frame)
@@ -356,4 +376,9 @@ async def yaml_ws_handler(request: web.Request, yaml_lsp_port: int) -> web.WebSo
             await ws.close()
 
     await asyncio.gather(client_to_server(), server_to_client(), return_exceptions=True)
+    try:
+        proc.kill()
+    except Exception:
+        pass
+    logger.info("yaml-language-server session ended PID=%d", proc.pid)
     return ws