aethergraph 0.1.0a1__py3-none-any.whl → 0.1.0a3__py3-none-any.whl

aethergraph/server/start.py

@@ -1,119 +1,280 @@
-# aethergraph/start.py
 from __future__ import annotations
 
-import asyncio
 import contextlib
-import socket
+from dataclasses import dataclass
+import os
+from pathlib import Path
 import threading
+import time
+from typing import Any
 
+from fastapi import FastAPI
 import uvicorn
 
 from aethergraph.config.context import set_current_settings
 from aethergraph.config.loader import load_settings
+from aethergraph.server.loading import GraphLoader, LoadSpec
+from aethergraph.server.server_state import (
+    get_running_url_if_any,
+    pick_free_port,
+    workspace_lock,
+    write_server_state,
+)
 
-from ..plugins.channel.routes.webui_routes import install_web_channel
 from .app_factory import create_app
 
 _started = False
 _server_thread: threading.Thread | None = None
-_shutdown_flag = threading.Event()
 _url: str | None = None
+_uvicorn_server: uvicorn.Server | None = None
+_loader = GraphLoader()
 
 
-def _pick_free_port(p: int) -> int:
-    if p:
-        return p
-    s = socket.socket()
-    s.bind(("", 0))
-    port = s.getsockname()[1]
-    s.close()
-    return port
+@dataclass
+class ServerHandle:
+    url: str
+    server: uvicorn.Server
+    thread: threading.Thread
 
+    def stop(self, timeout_s: float = 2.0) -> None:
+        self.server.should_exit = True
+        self.thread.join(timeout=timeout_s)
 
-def _run_uvicorn_in_thread(app, host: str, port: int, log_level: str):
-    loop = asyncio.new_event_loop()
-    asyncio.set_event_loop(loop)
-    server = uvicorn.Server(
-        uvicorn.Config(app, host=host, port=port, log_level=log_level, loop="asyncio")
-    )
+    def block(self) -> None:
+        # Loop with a timeout allows Python to process signals (like Ctrl+C)
+        while self.thread.is_alive():
+            self.thread.join(timeout=1.0)
 
-    async def runner():
-        task = asyncio.create_task(server.serve())
-        while not _shutdown_flag.is_set():
-            await asyncio.sleep(0.2)
-        if not server.should_exit:
-            server.should_exit = True
-        await task
 
-    try:
-        loop.run_until_complete(runner())
-    finally:
-        loop.stop()
-        loop.close()
+def _make_uvicorn_server(app: FastAPI, host: str, port: int, log_level: str) -> uvicorn.Server:
+    """
+    Create a uvicorn.Server we can stop via server.should_exit = True.
+    (Safe for background thread.)
+    """
+    config = uvicorn.Config(
+        app=app,
+        host=host,
+        port=port,
+        log_level=log_level,
+        lifespan="on",
+        loop="asyncio",
+    )
+    server = uvicorn.Server(config=config)
+    server.install_signal_handlers = lambda: None  # type: ignore
+    return server
 
 
 def start_server(
     *,
     workspace: str = "./aethergraph_data",
     host: str = "127.0.0.1",
-    port: int = 8000,  # 0 = auto free port
+    port: int = 8745,  # 0 = auto free port
     log_level: str = "warning",
     unvicorn_log_level: str = "warning",
     return_container: bool = False,
-) -> str:
+    return_handle: bool = False,
+    load_modules: list[str] | None = None,
+    load_paths: list[str] | None = None,
+    project_root: str | None = None,
+    strict_load: bool = False,
+) -> str | tuple[str, Any] | tuple[str, ServerHandle] | tuple[str, Any, ServerHandle]:
     """
-    Start the AetherGraph sidecar server in a background thread and install
-    services using the given workspace. Safe to call at top of any script
-    or notebook cell (no main() wrapper needed). Returns base URL.
+    Start (or reuse) the AetherGraph sidecar server in a normalized and flexible way.
+
+    This method manages server lifecycle, workspace locking, and dynamic loading of user code.
+    It supports both in-process and cross-process server reuse, and can return handles for
+    advanced control or integration.
+
+    Examples:
+      Basic usage to start a server and get its URL:
+      ```python
+      url = start_server(workspace="./aethergraph_data", port=0)
+      ```
+
+      Loading user graphs before starting:
+      ```python
+      url = start_server(
+        workspace="./aethergraph_data",
+        port=0,
+        load_paths=["./my_graphs.py"],
+        project_root=".",
+      )
+      ```
+
+      Starting and blocking until server exit (notebook/script mode):
+      ```python
+      url, handle = start_server(workspace="./aethergraph_data", port=0, return_handle=True)
+      print("Server running at", url)
+      try:
+        handle.block()
+      except KeyboardInterrupt:
+        print("Stopping server...")
+        handle.stop()
+      ```
+
+      Returning the dependency injection container for advanced use:
+      ```python
+      url, container = start_server(workspace="./aethergraph_data", return_container=True)
+      ```
+
+      Returning both container and handle:
+      ```python
+      url, container, handle = start_server(
+        workspace="./aethergraph_data",
+        return_container=True,
+        return_handle=True,
+      )
+      ```
+
+    Args:
+      workspace: Persistent storage directory for server state and data.
+      host: Host address to bind the server (default "127.0.0.1").
+      port: Port to bind the server (0 for auto-pick, or specify a fixed port).
+      log_level: Logging level for the application.
+      unvicorn_log_level: Logging level for the Uvicorn server.
+      return_container: If True, also return the app's dependency injection container.
+      return_handle: If True, return a ServerHandle for programmatic control (block/stop).
+      load_modules: List of Python modules to import before server start.
+      load_paths: List of Python file paths to import before server start.
+      project_root: Path to add to sys.path for module resolution during loading.
+      strict_load: If True, raise on import/load errors; otherwise, record errors in loader report.
+
+    Returns:
+      str: The server URL (e.g., "http://127.0.0.1:53421").
+      tuple: Optionally, (url, container), (url, handle), or (url, container, handle)
+        depending on the flags set and whether the server was started in-process.
+
+    Notes:
+      - Workspace is a dedicated directory for server data, including logs, caches, and runtime state; multiple processes using the same
+        workspace will coordinate to reuse a single server instance. Delete the workspace to reset state.
+      - Use handle.block() to wait for server exit when you need to keep the server running in a script or notebook. This is typically not needed
+        when using the server in client mode.
+      - When you are using Aethergraph UI, use handle.block() to keep the server running so that the UI can connect to it and discover agents/apps.
     """
-    global _started, _server_thread, _url
-    if _started:
-        return _url  # type: ignore
-
-    # Build app (installs services inside create_app)
-    cfg = load_settings()
-    set_current_settings(cfg)
-
-    app = create_app(workspace=workspace, cfg=cfg, log_level=log_level)
-
-    picked_port = _pick_free_port(port)
-    t = threading.Thread(
-        target=_run_uvicorn_in_thread,
-        args=(app, host, picked_port, unvicorn_log_level),
-        name="aethergraph-sidecar",
-        daemon=True,
-    )
-    t.start()
+    global _started, _server_thread, _url, _uvicorn_server
+
+    # In-process fast path
+    if _started and _url:
+        if return_container or return_handle:
+            # We can return these because we're in-process
+            # (container is attached to app only when we start it; see below)
+            pass
+        else:
+            print(" - reusing existing in-process server at", _url)
+            return _url
 
-    _server_thread = t
-    _started = True
-    _url = f"http://{host}:{picked_port}"
+    print(" - acquiring workspace lock...")
+    # Cross-process coordination: one workspace => one server
+    with workspace_lock(workspace):
+        running_url = get_running_url_if_any(workspace)
+        if running_url:
+            # Reuse the already-running sidecar for this workspace
+            print(" - reusing existing sidecar server at", running_url)
+            _started = True
+            _url = running_url
+            # Cross-process: we cannot return container/handle
+            return running_url
 
-    install_web_channel(app)
+        # Load graphs BEFORE server start so /apps, /agents are populated immediately
+        spec = LoadSpec(
+            modules=load_modules or [],
+            paths=load_paths or [],
+            project_root=project_root,
+            strict=strict_load,
+        )
 
-    if return_container:
-        return _url, app.state.container
-    return _url
+        print(" Loading user graphs with spec:", spec)
+        if spec.modules or spec.paths:
+            report = _loader.load(spec)
+            # Optional: stash report for debugging. We'll attach it to app below.
+            _loader.last_report = report
+
+        # Build app (installs services inside create_app)
+        cfg = load_settings()
+        set_current_settings(cfg)
+        app = create_app(workspace=workspace, cfg=cfg, log_level=log_level)
+        # Optional debug info
+        app.state.last_load_report = getattr(_loader, "last_report", None)
+
+        picked_port = pick_free_port(port)
+        url = f"http://{host}:{picked_port}"
+
+        # Create stoppable server object
+        server = _make_uvicorn_server(app, host, picked_port, unvicorn_log_level)
+
+        def _target():
+            server.run()
+
+        t = threading.Thread(
+            target=_target,
+            name="aethergraph-sidecar",
+            daemon=True,
+        )
+        t.start()
+
+        # Update globals
+        _server_thread = t
+        _uvicorn_server = server
+        _started = True
+        _url = url
+
+        # Write server.json for discovery
+        write_server_state(
+            workspace,
+            {
+                "pid": os.getpid(),
+                "host": host,
+                "port": picked_port,
+                "url": url,
+                "workspace": str(Path(workspace).resolve()),
+                "started_at": time.time(),
+            },
+        )
+
+        print("\n" + "=" * 50)
+        # We align the labels to 18 characters (the length of the longest label)
+        print(f"[AetherGraph] 🚀 {'Server started at:':<18} {url}")
+        print(
+            f"[AetherGraph] 🖥️  {'UI:':<18} {url}/ui   (if built)"
+        )  # strangly, this needs two spaces unlike the rest
+        print(f"[AetherGraph] 📡 {'API:':<18} {url}/api/v1/")
+        print(f"[AetherGraph] 📂 {'Workspace:':<18} {workspace}")
+        print("=" * 50 + "\n")
+
+        handle = ServerHandle(url=url, server=server, thread=t)
+
+        if return_container and return_handle:
+            return url, app.state.container, handle
+        if return_container:
+            return url, app.state.container
+        if return_handle:
+            return url, handle
+
+        return url
 
 
 async def start_server_async(**kw) -> str:
     # Async-friendly wrapper; still uses a thread to avoid clashing with caller loop
-    return start_server(**kw)
+    return start_server(**kw)  # type: ignore[return-value]
 
 
 def stop_server():
-    """Optional: stop the background server (useful in tests)."""
-    global _started, _server_thread, _url
+    """Stop the in-process background server (useful in tests/notebooks)."""
+    global _started, _server_thread, _url, _uvicorn_server
     if not _started:
         return
-    _shutdown_flag.set()
+
+    if _uvicorn_server is not None:
+        _uvicorn_server.should_exit = True
+
     if _server_thread and _server_thread.is_alive():
         with contextlib.suppress(Exception):
             _server_thread.join(timeout=5)
+
     _started = False
     _server_thread = None
+    _uvicorn_server = None
     _url = None
-    _shutdown_flag.clear()
 
 
 # backward compatibility