rappel 0.5.5__py3-none-manylinux_2_39_aarch64.whl

rappel/logger.py

@@ -0,0 +1,39 @@
+"""Simple logging helpers for configurable rappel loggers."""
+
+import logging
+import os
+from typing import Optional
+
+DEFAULT_LEVEL = logging.INFO
+ENV_VAR = "RAPPEL_LOG_LEVEL"
+
+
+def _resolve_level(value: Optional[str]) -> int:
+    if not value:
+        return DEFAULT_LEVEL
+    normalized = value.strip().upper()
+    mapping = {
+        "CRITICAL": logging.CRITICAL,
+        "FATAL": logging.FATAL,
+        "ERROR": logging.ERROR,
+        "WARNING": logging.WARNING,
+        "WARN": logging.WARNING,
+        "INFO": logging.INFO,
+        "DEBUG": logging.DEBUG,
+        "NOTSET": logging.NOTSET,
+    }
+    return mapping.get(normalized, DEFAULT_LEVEL)
+
+
+def configure(name: str) -> logging.Logger:
+    """Return a logger configured from RAPPEL_LOG_LEVEL."""
+
+    logger = logging.getLogger(name)
+    level = _resolve_level(os.environ.get(ENV_VAR))
+    logger.setLevel(level)
+    if not logger.handlers:
+        handler = logging.StreamHandler()
+        handler.setFormatter(logging.Formatter("[%(name)s] %(levelname)s: %(message)s"))
+        handler.setLevel(level)
+        logger.addHandler(handler)
+    return logger

rappel/registry.py

@@ -0,0 +1,106 @@
+from collections.abc import Awaitable, Callable
+from dataclasses import dataclass
+from threading import RLock
+from typing import Any, Optional
+
+AsyncAction = Callable[..., Awaitable[Any]]
+
+
+@dataclass
+class _ActionEntry:
+    module: str
+    name: str
+    func: AsyncAction
+
+
+def _make_key(module: str, name: str) -> str:
+    """Create a registry key from module and action name."""
+    return f"{module}:{name}"
+
+
+class ActionRegistry:
+    """In-memory registry of user-defined actions.
+
+    Actions are keyed by (module, name), allowing the same action name
+    to be used in different modules.
+    """
+
+    def __init__(self) -> None:
+        self._actions: dict[str, _ActionEntry] = {}
+        self._lock = RLock()
+
+    def _source_fingerprint(self, func: AsyncAction) -> tuple[str | None, str | None]:
+        func_any: Any = func
+        try:
+            code = func_any.__code__
+        except AttributeError:
+            return (None, None)
+        try:
+            qualname = func_any.__qualname__
+        except AttributeError:
+            qualname = None
+        filename = code.co_filename
+        if not isinstance(filename, str):
+            filename = None
+        if qualname is not None and not isinstance(qualname, str):
+            qualname = None
+        return (filename, qualname)
+
+    def _is_same_action_definition(self, existing: AsyncAction, new: AsyncAction) -> bool:
+        if existing is new:
+            return True
+        existing_fingerprint = self._source_fingerprint(existing)
+        new_fingerprint = self._source_fingerprint(new)
+        if existing_fingerprint == (None, None) or new_fingerprint == (None, None):
+            return False
+        return existing_fingerprint == new_fingerprint
+
+    def register(self, module: str, name: str, func: AsyncAction) -> None:
+        """Register an action with its module and name.
+
+        Args:
+            module: The Python module containing the action.
+            name: The action name (from @action decorator).
+            func: The async function to execute.
+
+        Raises:
+            ValueError: If an action with the same module:name is already registered
+                with a different implementation.
+        """
+        key = _make_key(module, name)
+        with self._lock:
+            existing = self._actions.get(key)
+            if existing is not None:
+                if self._is_same_action_definition(existing.func, func):
+                    self._actions[key] = _ActionEntry(module=module, name=name, func=func)
+                    return
+                raise ValueError(f"action '{module}:{name}' already registered")
+            self._actions[key] = _ActionEntry(module=module, name=name, func=func)
+
+    def get(self, module: str, name: str) -> Optional[AsyncAction]:
+        """Look up an action by module and name.
+
+        Args:
+            module: The Python module containing the action.
+            name: The action name.
+
+        Returns:
+            The action function if found, None otherwise.
+        """
+        key = _make_key(module, name)
+        with self._lock:
+            entry = self._actions.get(key)
+            return entry.func if entry else None
+
+    def names(self) -> list[str]:
+        """Return all registered action keys (module:name format)."""
+        with self._lock:
+            return sorted(self._actions.keys())
+
+    def reset(self) -> None:
+        """Clear all registered actions."""
+        with self._lock:
+            self._actions.clear()
+
+
+registry = ActionRegistry()

rappel/schedule.py

@@ -0,0 +1,357 @@
+"""
+Scheduled workflow execution.
+
+This module provides functions for registering workflows to run on a cron
+schedule or at fixed intervals.
+"""
+
+from dataclasses import dataclass
+from datetime import datetime, timedelta
+from typing import Any, Dict, List, Literal, Optional, Type, Union
+
+from grpc import aio  # type: ignore[attr-defined]
+
+from proto import messages_pb2 as pb2
+
+from .bridge import _workflow_stub, ensure_singleton
+from .serialization import build_arguments_from_kwargs
+from .workflow import Workflow
+
+ScheduleType = Literal["cron", "interval"]
+ScheduleStatus = Literal["active", "paused"]
+
+
+@dataclass
+class ScheduleInfo:
+    """Information about a registered schedule."""
+
+    id: str
+    workflow_name: str
+    schedule_name: str
+    schedule_type: ScheduleType
+    cron_expression: Optional[str]
+    interval_seconds: Optional[int]
+    jitter_seconds: Optional[int]
+    status: ScheduleStatus
+    next_run_at: Optional[datetime]
+    last_run_at: Optional[datetime]
+    last_instance_id: Optional[str]
+    created_at: datetime
+    updated_at: datetime
+
+
+async def schedule_workflow(
+    workflow_cls: Type[Workflow],
+    *,
+    schedule_name: str,
+    schedule: Union[str, timedelta],
+    jitter: Optional[timedelta] = None,
+    inputs: Optional[Dict[str, Any]] = None,
+) -> str:
+    """
+    Register a schedule for a workflow.
+
+    This function registers both the workflow DAG and the schedule in a single
+    call. When the schedule fires, the registered workflow version will be
+    executed.
+
+    Args:
+        workflow_cls: The Workflow class to schedule.
+        schedule_name: Unique name for this schedule. Allows multiple schedules
+                       per workflow with different inputs. Must be unique within
+                       a workflow.
+        schedule: Either a cron expression string (e.g., "0 * * * *" for hourly)
+                  or a timedelta for interval-based scheduling.
+        jitter: Optional jitter window to add to each scheduled run.
+        inputs: Optional keyword arguments to pass to each scheduled run.
+
+    Returns:
+        The schedule ID.
+
+    Examples:
+        # Run every hour at minute 0
+        await schedule_workflow(
+            MyWorkflow,
+            schedule_name="hourly-run",
+            schedule="0 * * * *"
+        )
+
+        # Run every 5 minutes
+        await schedule_workflow(
+            MyWorkflow,
+            schedule_name="frequent-check",
+            schedule=timedelta(minutes=5)
+        )
+
+        # Multiple schedules with different inputs
+        await schedule_workflow(
+            MyWorkflow,
+            schedule_name="small-batch",
+            schedule="0 0 * * *",
+            inputs={"batch_size": 100}
+        )
+        await schedule_workflow(
+            MyWorkflow,
+            schedule_name="large-batch",
+            schedule="0 12 * * *",
+            inputs={"batch_size": 1000}
+        )
+
+    Raises:
+        ValueError: If the cron expression is invalid, interval is non-positive,
+                    or schedule_name is empty.
+        RuntimeError: If the gRPC call fails.
+    """
+    if not schedule_name:
+        raise ValueError("schedule_name is required")
+
+    workflow_name = workflow_cls.short_name()
+
+    # Build schedule definition
+    schedule_def = pb2.ScheduleDefinition()
+    if isinstance(schedule, str):
+        schedule_def.type = pb2.SCHEDULE_TYPE_CRON
+        schedule_def.cron_expression = schedule
+    elif isinstance(schedule, timedelta):
+        interval_seconds = int(schedule.total_seconds())
+        if interval_seconds <= 0:
+            raise ValueError("Interval must be positive")
+        schedule_def.type = pb2.SCHEDULE_TYPE_INTERVAL
+        schedule_def.interval_seconds = interval_seconds
+    else:
+        raise TypeError(f"schedule must be str or timedelta, got {type(schedule)}")
+
+    if jitter is not None:
+        jitter_seconds = int(jitter.total_seconds())
+        if jitter_seconds < 0:
+            raise ValueError("jitter must be non-negative")
+        schedule_def.jitter_seconds = jitter_seconds
+
+    # Build the workflow registration payload to ensure the DAG is registered
+    # This is required for the schedule to execute - the scheduler needs a
+    # registered workflow version to create instances from.
+    registration = workflow_cls._build_registration_payload()
+
+    # Build request with both registration and schedule
+    request = pb2.RegisterScheduleRequest(
+        workflow_name=workflow_name,
+        schedule_name=schedule_name,
+        schedule=schedule_def,
+        registration=registration,
+    )
+
+    # Add inputs if provided
+    if inputs:
+        request.inputs.CopyFrom(build_arguments_from_kwargs(inputs))
+
+    # Send to server
+    async with ensure_singleton():
+        stub = await _workflow_stub()
+
+    try:
+        response = await stub.RegisterSchedule(request, timeout=30.0)
+    except aio.AioRpcError as exc:
+        raise RuntimeError(f"Failed to register schedule: {exc}") from exc
+
+    return response.schedule_id
+
+
+async def pause_schedule(workflow_cls: Type[Workflow], *, schedule_name: str) -> bool:
+    """
+    Pause a workflow's schedule.
+
+    The schedule will not fire until resumed. Existing running instances
+    are not affected.
+
+    Args:
+        workflow_cls: The Workflow class whose schedule to pause.
+        schedule_name: The name of the schedule to pause.
+
+    Returns:
+        True if a schedule was found and paused, False otherwise.
+
+    Raises:
+        ValueError: If schedule_name is empty.
+        RuntimeError: If the gRPC call fails.
+    """
+    if not schedule_name:
+        raise ValueError("schedule_name is required")
+
+    request = pb2.UpdateScheduleStatusRequest(
+        workflow_name=workflow_cls.short_name(),
+        schedule_name=schedule_name,
+        status=pb2.SCHEDULE_STATUS_PAUSED,
+    )
+    async with ensure_singleton():
+        stub = await _workflow_stub()
+
+    try:
+        response = await stub.UpdateScheduleStatus(request, timeout=30.0)
+    except aio.AioRpcError as exc:
+        raise RuntimeError(f"Failed to pause schedule: {exc}") from exc
+
+    return response.success
+
+
+async def resume_schedule(workflow_cls: Type[Workflow], *, schedule_name: str) -> bool:
+    """
+    Resume a paused workflow schedule.
+
+    Args:
+        workflow_cls: The Workflow class whose schedule to resume.
+        schedule_name: The name of the schedule to resume.
+
+    Returns:
+        True if a schedule was found and resumed, False otherwise.
+
+    Raises:
+        ValueError: If schedule_name is empty.
+        RuntimeError: If the gRPC call fails.
+    """
+    if not schedule_name:
+        raise ValueError("schedule_name is required")
+
+    request = pb2.UpdateScheduleStatusRequest(
+        workflow_name=workflow_cls.short_name(),
+        schedule_name=schedule_name,
+        status=pb2.SCHEDULE_STATUS_ACTIVE,
+    )
+    async with ensure_singleton():
+        stub = await _workflow_stub()
+
+    try:
+        response = await stub.UpdateScheduleStatus(request, timeout=30.0)
+    except aio.AioRpcError as exc:
+        raise RuntimeError(f"Failed to resume schedule: {exc}") from exc
+
+    return response.success
+
+
+async def delete_schedule(workflow_cls: Type[Workflow], *, schedule_name: str) -> bool:
+    """
+    Delete a workflow's schedule.
+
+    The schedule is soft-deleted and can be recreated by calling
+    schedule_workflow again.
+
+    Args:
+        workflow_cls: The Workflow class whose schedule to delete.
+        schedule_name: The name of the schedule to delete.
+
+    Returns:
+        True if a schedule was found and deleted, False otherwise.
+
+    Raises:
+        ValueError: If schedule_name is empty.
+        RuntimeError: If the gRPC call fails.
+    """
+    if not schedule_name:
+        raise ValueError("schedule_name is required")
+
+    request = pb2.DeleteScheduleRequest(
+        workflow_name=workflow_cls.short_name(),
+        schedule_name=schedule_name,
+    )
+    async with ensure_singleton():
+        stub = await _workflow_stub()
+
+    try:
+        response = await stub.DeleteSchedule(request, timeout=30.0)
+    except aio.AioRpcError as exc:
+        raise RuntimeError(f"Failed to delete schedule: {exc}") from exc
+
+    return response.success
+
+
+def _parse_iso_datetime(value: str) -> Optional[datetime]:
+    """Parse an ISO 8601 datetime string, returning None if empty."""
+    if not value:
+        return None
+    return datetime.fromisoformat(value.replace("Z", "+00:00"))
+
+
+def _proto_schedule_type_to_str(
+    schedule_type: "pb2.ScheduleType.V",
+) -> ScheduleType:
+    """Convert protobuf ScheduleType to string literal."""
+    if schedule_type == pb2.SCHEDULE_TYPE_CRON:
+        return "cron"
+    elif schedule_type == pb2.SCHEDULE_TYPE_INTERVAL:
+        return "interval"
+    else:
+        return "cron"  # Default fallback
+
+
+def _proto_schedule_status_to_str(
+    status: "pb2.ScheduleStatus.V",
+) -> ScheduleStatus:
+    """Convert protobuf ScheduleStatus to string literal."""
+    if status == pb2.SCHEDULE_STATUS_ACTIVE:
+        return "active"
+    elif status == pb2.SCHEDULE_STATUS_PAUSED:
+        return "paused"
+    else:
+        return "active"  # Default fallback
+
+
+async def list_schedules(
+    status_filter: Optional[ScheduleStatus] = None,
+) -> List[ScheduleInfo]:
+    """
+    List all registered workflow schedules.
+
+    Args:
+        status_filter: Optional filter by status ("active" or "paused").
+                       If None, returns all non-deleted schedules.
+
+    Returns:
+        A list of ScheduleInfo objects containing schedule details.
+
+    Examples:
+        # List all schedules
+        schedules = await list_schedules()
+        for s in schedules:
+            print(f"{s.workflow_name}: {s.status}")
+
+        # List only active schedules
+        active = await list_schedules(status_filter="active")
+
+        # List only paused schedules
+        paused = await list_schedules(status_filter="paused")
+
+    Raises:
+        RuntimeError: If the gRPC call fails.
+    """
+    request = pb2.ListSchedulesRequest()
+    if status_filter is not None:
+        request.status_filter = status_filter
+
+    async with ensure_singleton():
+        stub = await _workflow_stub()
+
+    try:
+        response = await stub.ListSchedules(request, timeout=30.0)
+    except aio.AioRpcError as exc:
+        raise RuntimeError(f"Failed to list schedules: {exc}") from exc
+
+    schedules = []
+    for s in response.schedules:
+        schedules.append(
+            ScheduleInfo(
+                id=s.id,
+                workflow_name=s.workflow_name,
+                schedule_name=s.schedule_name,
+                schedule_type=_proto_schedule_type_to_str(s.schedule_type),
+                cron_expression=s.cron_expression if s.cron_expression else None,
+                interval_seconds=s.interval_seconds if s.interval_seconds else None,
+                jitter_seconds=s.jitter_seconds if s.jitter_seconds else None,
+                status=_proto_schedule_status_to_str(s.status),
+                next_run_at=_parse_iso_datetime(s.next_run_at),
+                last_run_at=_parse_iso_datetime(s.last_run_at),
+                last_instance_id=s.last_instance_id if s.last_instance_id else None,
+                created_at=_parse_iso_datetime(s.created_at),  # type: ignore
+                updated_at=_parse_iso_datetime(s.updated_at),  # type: ignore
+            )
+        )
+
+    return schedules