penguiflow 1.0.2__py3-none-any.whl → 2.0.0__py3-none-any.whl

penguiflow/testkit.py

@@ -0,0 +1,269 @@
+"""Utilities for writing concise PenguiFlow tests.
+
+The helpers in this module provide a minimal harness around ``PenguiFlow`` so
+unit tests can focus on the behaviour of their nodes instead of the runtime
+plumbing.  Each helper intentionally works with the public runtime surface to
+avoid relying on private attributes, keeping the harness forward compatible
+with the v1 API.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import inspect
+from collections import OrderedDict
+from collections.abc import Awaitable, Callable, Iterable, Sequence
+from dataclasses import dataclass, field
+from itertools import groupby
+from typing import Any
+from weakref import WeakKeyDictionary
+
+from .core import PenguiFlow
+from .errors import FlowErrorCode
+from .metrics import FlowEvent
+from .types import Message
+
+__all__ = ["run_one", "assert_node_sequence", "simulate_error"]
+
+
+_MAX_TRACE_HISTORY = 64
+_TRACE_HISTORY: OrderedDict[str, list[FlowEvent]] = OrderedDict()
+_RECORDER_STATE: WeakKeyDictionary[PenguiFlow, _RecorderState] = (
+    WeakKeyDictionary()
+)
+
+
+def _register_trace_history(trace_id: str, events: list[FlowEvent]) -> None:
+    if not trace_id:
+        return
+    if trace_id in _TRACE_HISTORY:
+        _TRACE_HISTORY.move_to_end(trace_id)
+    _TRACE_HISTORY[trace_id] = events
+    while len(_TRACE_HISTORY) > _MAX_TRACE_HISTORY:
+        _TRACE_HISTORY.popitem(last=False)
+
+
+@dataclass(slots=True)
+class _RunLog:
+    events: list[FlowEvent] = field(default_factory=list)
+    traces: dict[str, list[FlowEvent]] = field(default_factory=dict)
+    active_traces: set[str] = field(default_factory=set)
+
+
+class _RecorderState:
+    def __init__(self) -> None:
+        self._lock = asyncio.Lock()
+        self._log = _RunLog()
+        self._middleware = _Recorder(self)
+
+    @property
+    def middleware(self) -> _Recorder:
+        return self._middleware
+
+    def begin(self, traces: Iterable[str] | None = None) -> None:
+        trace_ids = set(traces or [])
+        self._log = _RunLog(active_traces=trace_ids)
+        for trace_id in trace_ids:
+            bucket: list[FlowEvent] = []
+            self._log.traces[trace_id] = bucket
+            _register_trace_history(trace_id, bucket)
+
+    async def record(self, event: FlowEvent) -> None:
+        async with self._lock:
+            self._log.events.append(event)
+            trace_id = event.trace_id
+            if trace_id is None:
+                return
+            bucket = self._log.traces.get(trace_id)
+            if bucket is None:
+                bucket = []
+                self._log.traces[trace_id] = bucket
+                _register_trace_history(trace_id, bucket)
+            bucket.append(event)
+
+    def node_sequence(self, trace_id: str) -> list[str]:
+        bucket = self._log.traces.get(trace_id)
+        if bucket is None:
+            bucket = _TRACE_HISTORY.get(trace_id, [])
+        sequence: list[str] = []
+        for event in bucket:
+            if event.event_type != "node_start":
+                continue
+            name = event.node_name or event.node_id or "<anonymous>"
+            sequence.append(name)
+        return sequence
+
+
+class _Recorder:
+    def __init__(self, state: _RecorderState) -> None:
+        self._state = state
+
+    async def __call__(self, event: FlowEvent) -> None:
+        await self._state.record(event)
+
+
+def _get_state(flow: PenguiFlow) -> _RecorderState:
+    state = _RECORDER_STATE.get(flow)
+    if state is None:
+        state = _RecorderState()
+        _RECORDER_STATE[flow] = state
+    middlewares = getattr(flow, "_middlewares", None)
+    if middlewares is None:
+        raise AttributeError("PenguiFlow instance is missing middleware hooks")
+    middleware = state.middleware
+    if not any(middleware is existing for existing in middlewares):
+        middlewares.append(middleware)
+    return state
+
+
+async def run_one(
+    flow: PenguiFlow,
+    message: Message,
+    *,
+    registry: Any | None = None,
+    timeout_s: float | None = 1.0,
+) -> Any:
+    """Run ``message`` through ``flow`` and return the first Rookery payload.
+
+    The flow is started and stopped for the caller.  The original message's
+    ``trace_id`` is tracked so :func:`assert_node_sequence` can introspect the
+    execution order afterwards.
+    """
+
+    if not isinstance(message, Message):
+        raise TypeError("run_one expects a penguiflow.types.Message instance")
+
+    state = _get_state(flow)
+    state.begin([message.trace_id])
+
+    flow.run(registry=registry)
+    try:
+        await flow.emit(message)
+        result_coro = flow.fetch()
+        if timeout_s is not None:
+            result = await asyncio.wait_for(result_coro, timeout_s)
+        else:
+            result = await result_coro
+    finally:
+        await flow.stop()
+
+    return result
+
+
+def assert_node_sequence(trace_id: str, expected: Sequence[str]) -> None:
+    """Assert that ``expected`` matches the recorded node start order."""
+
+    expected_nodes = list(expected)
+    events = _TRACE_HISTORY.get(trace_id, [])
+    if not events:
+        raise AssertionError(
+            "No recorded events for trace_id="
+            f"{trace_id!r}; run a flow with run_one first."
+        )
+
+    actual_nodes = [
+        event.node_name or event.node_id or "<anonymous>"
+        for event in events
+        if event.event_type == "node_start"
+    ]
+    actual_nodes = [name for name, _ in groupby(actual_nodes)]
+    if actual_nodes != expected_nodes:
+        raise AssertionError(
+            "Node sequence mismatch:\n"
+            f"  expected: {expected_nodes}\n"
+            f"  actual:   {actual_nodes}"
+        )
+
+
+class _ErrorSimulation:
+    def __init__(
+        self,
+        *,
+        node_name: str,
+        code: str,
+        fail_times: int,
+        exception_factory: Callable[[str], Exception],
+        result_factory: Callable[[Any], Awaitable[Any] | Any] | None,
+    ) -> None:
+        self._node_name = node_name
+        self._code = code
+        self._fail_times = fail_times
+        self._exception_factory = exception_factory
+        self._result_factory = result_factory
+        self._attempts = 0
+
+    @property
+    def attempts(self) -> int:
+        return self._attempts
+
+    @property
+    def failures(self) -> int:
+        return min(self._attempts, self._fail_times)
+
+    async def __call__(self, message: Any, _ctx: Any) -> Any:
+        self._attempts += 1
+        if self._attempts <= self._fail_times:
+            text = (
+                f"[{self._code}] simulated failure in {self._node_name}"
+                f" (attempt {self._attempts})"
+            )
+            raise self._exception_factory(text)
+
+        if self._result_factory is None:
+            return message
+
+        result = self._result_factory(message)
+        if inspect.isawaitable(result):
+            return await result
+        return result
+
+
+def simulate_error(
+    node_name: str,
+    code: FlowErrorCode | str,
+    *,
+    fail_times: int = 1,
+    result: Any | None = None,
+    result_factory: Callable[[Any], Awaitable[Any] | Any] | None = None,
+    exception_type: type[Exception] = RuntimeError,
+) -> Callable[[Any, Any], Awaitable[Any]]:
+    """Return an async callable that fails ``fail_times`` before succeeding.
+
+    The returned coroutine is suitable for wrapping in :class:`~penguiflow.node.Node`
+    and is especially useful for retry-centric tests.  By default the callable
+    echoes the incoming ``message`` once the simulated failures are exhausted, but
+    ``result``/``result_factory`` can override the successful return value.
+    """
+
+    if fail_times < 1:
+        raise ValueError("fail_times must be >= 1")
+    if result is not None and result_factory is not None:
+        raise ValueError("Specify only one of result or result_factory")
+
+    resolved_code = code.value if isinstance(code, FlowErrorCode) else str(code)
+
+    def _exception_factory(text: str) -> Exception:
+        return exception_type(text)
+
+    if result_factory is None and result is not None:
+        async def _const_result(_: Any) -> Any:
+            return result
+
+        result_factory = _const_result
+
+    simulation = _ErrorSimulation(
+        node_name=node_name,
+        code=resolved_code,
+        fail_times=fail_times,
+        exception_factory=_exception_factory,
+        result_factory=result_factory,
+    )
+
+    async def _runner(message: Any, ctx: Any) -> Any:
+        return await simulation(message, ctx)
+
+    # Attach useful attributes for introspection in tests without exposing the
+    # internal class.
+    _runner.simulation = simulation  # type: ignore[attr-defined]
+    return _runner
+

penguiflow/types.py

@@ -21,6 +21,17 @@ class Message(BaseModel):
     trace_id: str = Field(default_factory=lambda: uuid.uuid4().hex)
     ts: float = Field(default_factory=time.time)
     deadline_s: float | None = None
+    meta: dict[str, Any] = Field(default_factory=dict)
+
+
+class StreamChunk(BaseModel):
+    """Represents a chunk of streamed output."""
+
+    stream_id: str
+    seq: int
+    text: str
+    done: bool = False
+    meta: dict[str, Any] = Field(default_factory=dict)
 
 
 class PlanStep(BaseModel):
@@ -39,7 +50,9 @@ class WM(BaseModel):
     query: str
     facts: list[Any] = Field(default_factory=list)
     hops: int = 0
-    budget_hops: int = 8
+    budget_hops: int | None = 8
+    tokens_used: int = 0
+    budget_tokens: int | None = None
     confidence: float = 0.0
 
 
@@ -51,6 +64,7 @@ class FinalAnswer(BaseModel):
 __all__ = [
     "Headers",
     "Message",
+    "StreamChunk",
     "PlanStep",
     "Thought",
     "WM",

penguiflow/viz.py

@@ -2,4 +2,184 @@
 
 from __future__ import annotations
 
-__all__: list[str] = []
+import re
+from dataclasses import dataclass
+from typing import TYPE_CHECKING
+
+from .core import Endpoint
+from .node import Node
+
+if TYPE_CHECKING:  # pragma: no cover - type checking only
+    from .core import PenguiFlow
+
+__all__ = ["flow_to_mermaid", "flow_to_dot"]
+
+
+@dataclass
+class _VisualNode:
+    identifier: str
+    label: str
+    classes: list[str]
+
+
+@dataclass
+class _VisualEdge:
+    source: str
+    target: str
+    label: str | None
+
+
+def flow_to_mermaid(flow: PenguiFlow, *, direction: str = "TD") -> str:
+    """Render the flow graph as a Mermaid diagram string.
+
+    Parameters
+    ----------
+    flow:
+        The :class:`PenguiFlow` instance to visualize.
+    direction:
+        Mermaid graph direction (``"TD"``, ``"LR"``, etc.). Defaults to top-down.
+    """
+
+    nodes, edges = _collect_graph(flow)
+
+    lines: list[str] = [f"graph {direction}"]
+    class_defs = {
+        "endpoint": "fill:#e0f2fe,stroke:#0369a1,stroke-width:1px",
+        "controller_loop": "fill:#fef3c7,stroke:#b45309,stroke-width:1px",
+    }
+
+    used_definitions: set[str] = set()
+
+    for node in nodes:
+        label = _escape_label(node.label)
+        lines.append(f"    {node.identifier}[\"{label}\"]")
+        for class_name in node.classes:
+            used_definitions.add(class_name)
+
+    for class_name in sorted(used_definitions):
+        style = class_defs.get(class_name)
+        if style:
+            lines.append(f"    classDef {class_name} {style}")
+
+    for node in nodes:
+        if node.classes:
+            classes = " ".join(node.classes)
+            lines.append(f"    class {node.identifier} {classes}")
+
+    for edge in edges:
+        label = f"|{edge.label}|" if edge.label else ""
+        lines.append(f"    {edge.source} -->{label} {edge.target}")
+
+    return "\n".join(lines)
+
+
+def flow_to_dot(flow: PenguiFlow, *, rankdir: str = "TB") -> str:
+    """Render the flow graph as a Graphviz DOT string.
+
+    Parameters
+    ----------
+    flow:
+        The :class:`PenguiFlow` instance to visualize.
+    rankdir:
+        Graph orientation (``"TB"``, ``"LR"``, etc.). Defaults to top-bottom.
+    """
+
+    nodes, edges = _collect_graph(flow)
+
+    lines: list[str] = ["digraph PenguiFlow {", f"    rankdir={rankdir}"]
+    lines.append("    node [shape=box, style=rounded]")
+
+    for node in nodes:
+        attributes: list[str] = [f'label="{node.label}"']
+        if "endpoint" in node.classes:
+            attributes.append('shape=oval')
+            attributes.append('style="filled"')
+            attributes.append('fillcolor="#e0f2fe"')
+        elif "controller_loop" in node.classes:
+            attributes.append('style="rounded,filled"')
+            attributes.append('fillcolor="#fef3c7"')
+        attr_str = ", ".join(attributes)
+        lines.append(f"    {node.identifier} [{attr_str}]")
+
+    for edge in edges:
+        if edge.label:
+            edge_label = _escape_label(edge.label)
+            lines.append(
+                f"    {edge.source} -> {edge.target} [label=\"{edge_label}\"]"
+            )
+        else:
+            lines.append(f"    {edge.source} -> {edge.target}")
+
+    lines.append("}")
+    return "\n".join(lines)
+
+
+def _collect_graph(flow: PenguiFlow) -> tuple[list[_VisualNode], list[_VisualEdge]]:
+    nodes: dict[object, _VisualNode] = {}
+    edges: list[_VisualEdge] = []
+    used_ids: set[str] = set()
+    loop_sources: set[object] = set()
+
+    def ensure_node(entity: object) -> _VisualNode:
+        node = nodes.get(entity)
+        if node is not None:
+            return node
+        label = _display_label(entity)
+        identifier = _unique_id(label, used_ids)
+        used_ids.add(identifier)
+        classes: list[str] = []
+        if isinstance(entity, Endpoint):
+            classes.append("endpoint")
+        if isinstance(entity, Node) and entity.allow_cycle:
+            classes.append("controller_loop")
+        node = _VisualNode(identifier=identifier, label=label, classes=classes)
+        nodes[entity] = node
+        return node
+
+    for floe in flow._floes:  # noqa: SLF001 - visualization inspects internals
+        source = floe.source
+        target = floe.target
+        if source is None or target is None:
+            continue
+        src_node = ensure_node(source)
+        tgt_node = ensure_node(target)
+        if source is target:
+            loop_sources.add(source)
+            label = "loop"
+        elif isinstance(source, Endpoint):
+            label = "ingress"
+        elif isinstance(target, Endpoint):
+            label = "egress"
+        else:
+            label = None
+        edges.append(_VisualEdge(src_node.identifier, tgt_node.identifier, label))
+
+    if loop_sources:
+        for entity, node in nodes.items():
+            if entity in loop_sources and "controller_loop" not in node.classes:
+                node.classes.append("controller_loop")
+
+    return list(nodes.values()), edges
+
+
+def _display_label(entity: object) -> str:
+    if isinstance(entity, Node):
+        return entity.name or entity.node_id
+    if isinstance(entity, Endpoint):
+        return entity.name
+    return str(entity)
+
+
+def _unique_id(label: str, used: set[str]) -> str:
+    base = re.sub(r"[^0-9A-Za-z_]", "_", label) or "node"
+    candidate = base
+    index = 1
+    while candidate in used:
+        index += 1
+        candidate = f"{base}_{index}"
+    return candidate
+
+
+def _escape_label(label: str) -> str:
+    return label.replace("\"", "\\\"")
+