Files
app/backend/fluksio/flow/schemas.py
T
stroblmeandClaude Opus 5 6ff56533f5 Schedule a node across every machine, not just this one
The engine answered "where does this node run" twice, in two ways that could
not see each other: a device sent it to a worker carrying that label, and
resources were counted against the engine's own cores. Declaring both meant the
second answer won and nothing was counted at all — which the data-science
getting-started page and the worked example both do.

One question now, in flow/placement.py: of every machine attached, which could
grant what this node asked for, and which of those has it free. The books move
onto each machine — one accountant per worker, built from the inventory it
reported — and the waiting moves above them, where one condition variable can
be woken by a release anywhere or by a worker attaching. Locks go one way:
placer, then a machine's books, never back.

So a node asking for a card now finds the box that has one, rather than being
clamped down to none and run here. When nothing can grant the ask at all it is
still cut down and run — a flow written on a cluster has to work on a laptop —
but the ceiling is one real machine now, since taking the largest of each
dimension separately can describe a machine nobody has.

Two things fixed on the way. A device on a connector node held every batch run
of its flow forever, waiting for a worker that could never run an entry point.
And `prefer` falling back to the engine skipped the books, so the fallback held
nothing.

The bench flow's node has taken a `params` argument that with_settings has not
forwarded for some time, so the benchmark could not run at all: 62 ms median
submit-to-result with this, against the 61 ms on record.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01A6HeySA27EkGANZN95QySW
2026-08-27 08:49:36 +02:00

393 lines
13 KiB
Python

"""The persisted shape of a flow, shared by the store, the API and the editor.
A flow is structure plus code: this module is the structure. Node logic for
``python`` nodes lives beside it as a plain ``.py`` file.
"""
from __future__ import annotations
import re
from typing import Any, Literal
from pydantic import BaseModel, ConfigDict, Field, field_validator
from fluksio.flow.messages import MessageSpec
NAME_PATTERN = re.compile(r"^[a-z][a-z0-9_]*$")
def _validate_name(value: str) -> str:
if not NAME_PATTERN.match(value):
raise ValueError(
"Use lowercase letters, digits and underscores, starting with a letter"
)
return value
class Resources(BaseModel):
"""What one execution of a node needs to have to itself.
Declaring nothing is the default and means what it always did: the node
runs on the shared worker pool and nothing is accounted for it. That is
right for the kind of node most flows are made of — a poll, a threshold, a
message on its way somewhere.
It is wrong for the other kind. A numerical library sizes its thread pool
to every core it can see, so a handful of them at once oversubscribe the
machine badly enough to starve the engine's own event loop, and a GPU
library that preallocates most of the card deadlocks when a second one
arrives. Both are a node saying how much of the machine it takes, which is
what this is.
"""
model_config = ConfigDict(extra="forbid")
cpus: int = Field(
default=1,
ge=1,
description=(
"Cores held for the whole execution. Also what the thread-pool "
"variables are set to, so a library sizing itself to the machine "
"sizes itself to this instead."
),
)
gpus: int = Field(
default=0,
ge=0,
description=(
"Whole devices held for the whole execution, named to the node "
"through CUDA_VISIBLE_DEVICES. Nothing else is given them while it "
"runs, which is what keeps two preallocating processes apart."
),
)
ram: int | None = Field(
default=None,
ge=1,
description=(
"Megabytes held for the whole execution. Counted against machines "
"that said how much they have, and ignored by those that did not — "
"which is a machine with nothing to say about memory, not one with "
"none."
),
)
env: dict[str, str] = Field(
default_factory=dict,
description=(
"Extra environment for the worker this node runs in, applied over "
"what the allocation derives. Where a library's own tuning goes — "
"XLA_FLAGS, XLA_PYTHON_CLIENT_MEM_FRACTION — since those are "
"composed strings the engine must not invent."
),
)
class NodeDef(BaseModel):
"""A node as stored: identity, configuration and ports.
Deliberately no canvas position. The editor lays a flow out itself, so
where a node sits is a fact about the drawing rather than about the flow —
and a graph nobody can arrange is one worth keeping small.
"""
id: str
type: str = "python"
title: str = ""
#: This node's settings: constants of its function, stored with the flow.
#: A function node reads them as keyword arguments beside its ports, so a
#: setting cannot share a name with one.
params: dict[str, Any] = Field(default_factory=dict)
requires: list[MessageSpec] = Field(default_factory=list)
provides: list[MessageSpec] = Field(default_factory=list)
#: Name of a shared source in the library, instead of this node's own file.
#: Editing it edits the copy every flow using it runs.
source_ref: str | None = None
timeout: float | None = Field(
default=None,
ge=0,
description=(
"Seconds this node's code may be silent before it is stopped. A "
"yield or an emit resets the clock, and the first call's imports "
"are not charged to it. 0 disables the limit: the node runs until "
"it finishes, and only a dead worker fails the call. Empty "
"inherits the engine default."
),
)
device: str | None = Field(
default=None,
description=(
"Label of the worker this node's code must run on, such as 'gpu'. "
"Empty means the engine's own workers. A run needing a label no "
"attached worker carries waits rather than failing."
),
)
device_policy: Literal["require", "prefer"] = Field(
default="require",
description=(
"What to do when no worker carries `device`: wait for one, or run "
"locally anyway."
),
)
cache: bool = Field(
default=True,
description=(
"Whether a batch run may reuse an earlier execution of this node "
"with the same source, settings and inputs. Turn it off for a "
"function whose answer can change on its own."
),
)
resources: Resources | None = Field(
default=None,
description=(
"What one execution of this node holds while it runs. Absent — the "
"default — means it is not accounted for and shares the engine's "
"workers, which is right for everything that is not compute-heavy."
),
)
@field_validator("id")
@classmethod
def _check_id(cls, value: str) -> str:
return _validate_name(value)
class FlowOrigin(BaseModel):
"""Where a flow was declared, when that was somewhere other than here.
A flow drawn on the canvas has no origin: the store is where it lives. One
stamped with this was declared with the decorators in somebody's own
repository and put here by ``fluksio sync``, so the node bodies below it
are generated imports and the code they run is versioned twice — once here
and once there. Its presence is what makes a flow code-defined.
Deliberately no timestamp. The store commits every change it is given, so
when a flow was last synced is a fact its own history already holds — and
one that would otherwise change on every sync, making an unchanged upload
look like a new version of the flow.
"""
kind: Literal["python"] = "python"
#: The repository root on the machine that ran ``sync``.
repo: str = ""
#: Its commit, and whether the tree had uncommitted changes at the time —
#: a run stamped with a dirty commit names code that was never stored.
commit: str = ""
dirty: bool = False
class FlowInput(BaseModel):
"""A message the flow starts with rather than computes."""
spec: MessageSpec
initial: Any | None = None
class FlowDef(BaseModel):
"""One atomic flow."""
name: str
title: str = ""
nodes: list[NodeDef] = Field(default_factory=list)
inputs: list[FlowInput] = Field(default_factory=list)
version: int = 1
mode: Literal["live", "batch"] = Field(
default="live",
description=(
"A live flow reacts to what arrives: its subscriptions, schedules "
"and webhooks run until it is stopped. A batch flow only runs when "
"a run asks it to, from its inputs to its outputs, and is never "
"activated."
),
)
outputs: list[str] = Field(
default_factory=list,
description=(
"Messages a batch run reports as its result, unqualified. Empty "
"means every message the flow ends up holding."
),
)
origin: FlowOrigin | None = Field(
default=None,
description=(
"Set when the flow was declared in code elsewhere and uploaded by "
"`fluksio sync`. Absent for a flow drawn on the canvas."
),
)
@field_validator("name")
@classmethod
def _check_name(cls, value: str) -> str:
return _validate_name(value)
class NodeSource(BaseModel):
"""The Python source of a node."""
code: str
#: True when nothing is stored and `code` is the new-node template. An
#: editor opens on it either way; a client deciding whether somebody wrote
#: that code needs to know it was nobody. Read-only — set on the way out.
missing: bool = False
Health = Literal["ok", "degraded", "down"]
class NodeStatusPublic(BaseModel):
"""Whether a node loaded, and how its connection is doing."""
id: str
status: str = "active"
error: str | None = None
health: Health = "ok"
health_detail: str | None = None
#: The node's last runtime failure, kept after it runs again: a failure
#: that fired an alert should leave a trace of what it was.
last_error: str = ""
last_error_ts: float | None = None
class MessageValue(BaseModel):
"""The last payload seen on a message."""
value: Any = None
ts: float | None = None
class HistoryPoint(BaseModel):
"""One numeric value a message carried, and when."""
ts: float
value: float
class MessageHistory(BaseModel):
"""A message's recent numeric values, oldest first.
Only numbers are recorded, so ``numeric`` tells the panel whether an empty
series means "nothing plottable here" or "nothing has arrived yet".
"""
message: str
numeric: bool = False
points: list[HistoryPoint] = Field(default_factory=list)
class FlowSummary(BaseModel):
name: str
title: str = ""
node_count: int = 0
error_count: int = 0
has_draft: bool = False
enabled: bool = True
paused: bool = False
# Its background tasks kept crashing, so the engine stopped restarting them.
quarantined: bool = False
#: Of the working copy, so publishing from a list needs no second read.
version: int = 1
class FlowsPublic(BaseModel):
data: list[FlowSummary]
count: int
class LibraryNode(BaseModel):
"""A node source shared across flows, and who is using it."""
name: str
used_by: list[str] = Field(default_factory=list)
class FlowStatePublic(BaseModel):
values: dict[str, MessageValue] = Field(default_factory=dict)
nodes: list[NodeStatusPublic] = Field(default_factory=list)
class ModulePackage(BaseModel):
"""One package installed in the venv node code runs on."""
name: str
version: str
class ModulesInfo(BaseModel):
"""The venv node code imports from, and the manifest that describes it."""
python_version: str = ""
venv_path: str = ""
requirements: str = ""
packages: list[ModulePackage] = Field(default_factory=list)
#: Whether what is installed matches the manifest.
applied: bool = False
#: True when node code runs on the venv Fluksio was installed into rather
#: than one the engine built. That venv belongs to whoever made it, so the
#: manifest does not describe it and nothing here installs into it.
adopted: bool = False
class ApplyRequest(BaseModel):
"""A pip manifest, one requirement per line."""
requirements: str = ""
class ApplyResult(BaseModel):
ok: bool
output: str = ""
class BrainNode(BaseModel):
"""One neuron: a thing the engine talks to, or a node that only computes.
Nodes of the same type pointing at the same outside thing — one broker
topic, one URL, one bucket — are a single entry here, whichever flows they
sit in. ``members`` are the ``flow.node_id`` names behind it, which is also
what the live events are keyed by.
"""
id: str
label: str
kind: str
members: list[str] = Field(default_factory=list)
flows: list[str] = Field(default_factory=list)
issue: str | None = Field(
default=None,
description=(
"Why this neuron cannot run, if validation found something. A "
"failure the engine hits while running arrives over the socket "
"instead; this is the part that is already true before anything "
"fires, and so has to travel with the graph."
),
)
class BrainEdge(BaseModel):
"""Messages carrying values from one neuron to another."""
source: str
target: str
messages: list[str] = Field(default_factory=list)
class BrainGraph(BaseModel):
"""Every published flow at once, merged on what its nodes talk to."""
nodes: list[BrainNode] = Field(default_factory=list)
edges: list[BrainEdge] = Field(default_factory=list)
class NodeTypeInfo(BaseModel):
"""A node type the editor can offer, with its parameter schema."""
type: str
title: str
description: str
params_schema: dict[str, Any] = Field(default_factory=dict)
has_source: bool = False
#: Whether this type takes settings beyond the ones its schema declares.
#: A function node's settings are its author's to name, and reach `process`
#: as keyword arguments beside its ports.
free_params: bool = False
#: The package a connector came from; empty for the built-in types.
plugin: str | None = None