Run python nodes out of process, with modules of their own

User code no longer execs in the engine. A pool of persistent worker
subprocesses speaks one JSON object per line; the controller installs a
proxy as the node's function, so every execution path funnels through it
and the pipeline is untouched. A crash costs one subprocess, a per-node
timeout is a kill, and cancelling from the canvas is that same kill.

The workers run a venv of the user's own on the data volume, filled from
a pip manifest versioned beside the flows. Applying it retires the
workers and rebuilds, so a package lands without restarting the engine.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017MeiWk3Yq12n2pTvnQWYvt

backend/app/mcp/server.py

@@ -238,3 +238,30 @@ async def pause_flow(name: str) -> Any:
 async def resume_flow(name: str) -> Any:
     """Let a paused flow carry on, running whatever was held back."""
     return await _call("POST", f"/flows/{name}/resume")
+
+
+@mcp.tool()
+async def cancel_node(name: str, node_id: str) -> Any:
+    """Stop a node's code while it is running. Nothing to stop is not an error."""
+    return await _call("POST", f"/flows/{name}/nodes/{node_id}/cancel")
+
+
+# -----------------------------------------------------------------------------
+# Modules
+# -----------------------------------------------------------------------------
+
+
+@mcp.tool()
+async def get_modules() -> Any:
+    """The python packages node code can import, and the manifest asking for them."""
+    return await _call("GET", "/modules/")
+
+
+@mcp.tool()
+async def apply_modules(requirements: str) -> Any:
+    """Install exactly these requirements, one pip line each.
+
+    This replaces the whole manifest: a package left out is uninstalled. A
+    manifest that does not resolve changes nothing.
+    """
+    return await _call("POST", "/modules/apply", json={"requirements": requirements})