A node's numbers leave through its ports, not a logging call

The first cut had node code call fluksio.log_metric, which was a second,
undeclared way for data to leave a node: invisible to validation, absent from
the canvas, and stored where the graph could not see it. That is precisely the
MLflow discrepancy this framework exists to avoid, so it is gone.

A node that produces values over time is a generator. Every yield is a dict
keyed by output port, published the instant it happens — same port, same type
check, same place on the canvas as any other value — and what it returns is
its result. A port doing this declares stream: true, and a run keeps every
number one takes, so experiment tracking is a consequence of the graph rather
than an API beside it: a chart binds to a training curve the way it binds to a
temperature. fluksio.emit writes the same ports imperatively, for where a
yield cannot reach — inside a training framework's callback.

In a live flow an emission also wakes what is downstream, as a subscriber
publishing does; in a run it does not, because a run's graph is scheduled once
and mid-node cascades would leave 'finished' with nothing to mean. The
enqueued item carries no payload: the value is already in state, and one
carrying it would re-apply an old emission after the node returned.

Verified on the stack: 30 loss values arrived live on the flow socket during a
run, attributed to the node that produced them, and the same node run on the
remote worker streamed its curve back across the socket.

Also caches remote compile results per worker, so attaching a GPU box does not
put a network round trip in every rebuild.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AD8SfVhzXBG2nAfFcVh3iD

backend/tests/flow/test_workers.py

@@ -181,21 +181,19 @@ def test_a_pool_can_stop_while_a_node_is_running(pool):
 
 
 # -----------------------------------------------------------------------------
-# Reporting from inside a node that has not returned yet
+# Producing values before returning: a generator node
 # -----------------------------------------------------------------------------
 
 
-def test_a_node_reports_metrics_while_it_is_still_running(pool):
+def test_a_generator_node_publishes_each_yield_and_returns_the_end(pool):
     seen = []
     result = pool.run(
         "demo",
         "train",
-        "import fluksio\n"
         "def process(params):\n"
         "    for step in range(3):\n"
-        "        fluksio.log_metric('loss', 1.0 / (step + 1), step)\n"
-        "    fluksio.progress(0.5, 'halfway')\n"
-        "    return {'out': 1}\n",
+        "        yield {'loss': 1.0 / (step + 1)}\n"
+        "    return {'weights': 'w', 'final_loss': 0.25}\n",
         {},
         {},
         "demo.train",
@@ -204,32 +202,80 @@ def test_a_node_reports_metrics_while_it_is_still_running(pool):
         on_event=seen.append,
     )
 
-    assert result == {"out": 1}
-    metrics = [event for event in seen if event["event"] == "metric"]
-    assert [(m["name"], m["step"]) for m in metrics] == [
-        ("loss", 0),
-        ("loss", 1),
-        ("loss", 2),
-    ]
-    assert metrics[0]["value"] == 1.0
-    # Every event says which call it belongs to, so a sweep can tell them apart.
-    assert {m["call_id"] for m in metrics} == {"r1:demo.train"}
-    assert [event["event"] for event in seen if event["event"] == "progress"] == [
-        "progress"
+    # What it returned is the node's output; what it yielded went out as it
+    # happened, on the same ports.
+    assert result == {"weights": "w", "final_loss": 0.25}
+    assert [event["outputs"] for event in seen] == [
+        {"loss": 1.0},
+        {"loss": 0.5},
+        {"loss": 1 / 3},
     ]
+    # Every frame says which call it belongs to, so a sweep can tell them apart.
+    assert {event["call_id"] for event in seen} == {"r1:demo.train"}
+
+
+def test_without_a_return_the_last_yield_is_the_result(pool):
+    seen = []
+    result = pool.run(
+        "demo",
+        "count",
+        "def process(params):\n"
+        "    yield {'out': 1}\n"
+        "    yield {'out': 2}\n"
+        "    yield {'out': 3}\n",
+        {},
+        {},
+        "demo.count",
+        timeout=5,
+        on_event=seen.append,
+    )
+
+    assert result == {"out": 3}
+    assert [event["outputs"] for event in seen] == [{"out": 1}, {"out": 2}]
+
+
+def test_emit_reaches_the_same_ports_from_inside_a_callback(pool):
+    # A value produced somewhere a yield cannot reach — inside a framework's
+    # callback — is still an output rather than a log.
+    seen = []
+    result = pool.run(
+        "demo",
+        "fit",
+        "import fluksio\n"
+        "def process(params):\n"
+        "    def on_epoch(n):\n"
+        "        fluksio.emit(loss=1.0 / (n + 1))\n"
+        "    for epoch in range(2):\n"
+        "        on_epoch(epoch)\n"
+        "    return {'done': True}\n",
+        {},
+        {},
+        "demo.fit",
+        timeout=5,
+        on_event=seen.append,
+    )
+
+    assert result == {"done": True}
+    assert [event["outputs"] for event in seen] == [{"loss": 1.0}, {"loss": 0.5}]
+
+
+def test_a_plain_function_still_just_returns(pool):
+    seen = []
+    assert run(pool, "def process(params):\n    return {'out': 7}\n") == {"out": 7}
+    assert seen == []
 
 
 def test_events_hold_off_the_timeout_but_silence_does_not(pool):
-    # The deadline measures silence: a node reporting every 0.05s stays alive
+    # The deadline measures silence: a node yielding every 0.05s stays alive
     # well past a 0.3s timeout, which is what a two-hour training needs.
     result = pool.run(
         "demo",
         "slow",
-        "import time, fluksio\n"
+        "import time\n"
         "def process(params):\n"
         "    for step in range(12):\n"
         "        time.sleep(0.05)\n"
-        "        fluksio.log_metric('beat', step, step)\n"
+        "        yield {'beat': step}\n"
         "    return {'done': True}\n",
         {},
         {},