digitalkin 0.2.26__py3-none-any.whl → 0.3.0.dev1__py3-none-any.whl

digitalkin/modules/_base_module.py

@@ -1,14 +1,11 @@
 """BaseModule is the abstract base for all modules in the DigitalKin SDK."""
 
 import asyncio
-import contextlib
 import json
 from abc import ABC, abstractmethod
 from collections.abc import Callable, Coroutine
 from typing import Any, ClassVar, Generic
 
-from pydantic import BaseModel
-
 from digitalkin.logger import logger
 from digitalkin.models.module import (
     InputModelT,
@@ -17,28 +14,14 @@ from digitalkin.models.module import (
     SecretModelT,
     SetupModelT,
 )
+from digitalkin.models.module.module import ModuleCodeModel
 from digitalkin.models.module.module_context import ModuleContext
 from digitalkin.modules.trigger_handler import TriggerHandler
-from digitalkin.services.agent.agent_strategy import AgentStrategy
-from digitalkin.services.cost.cost_strategy import CostStrategy
-from digitalkin.services.filesystem.filesystem_strategy import FilesystemStrategy
-from digitalkin.services.identity.identity_strategy import IdentityStrategy
-from digitalkin.services.registry.registry_strategy import RegistryStrategy
 from digitalkin.services.services_config import ServicesConfig, ServicesStrategy
-from digitalkin.services.snapshot.snapshot_strategy import SnapshotStrategy
-from digitalkin.services.storage.storage_strategy import StorageStrategy
 from digitalkin.utils.llm_ready_schema import llm_ready_schema
 from digitalkin.utils.package_discover import ModuleDiscoverer
 
 
-class ModuleCodeModel(BaseModel):
-    """typed error/code model."""
-
-    code: str
-    message: str
-    short_description: str
-
-
 class BaseModule(  # noqa: PLR0904
     ABC,
     Generic[
@@ -67,33 +50,35 @@ class BaseModule(  # noqa: PLR0904
     services_config_params: ClassVar[dict[str, dict[str, Any | None] | None]]
     services_config: ServicesConfig
 
-    # services list
-    agent: AgentStrategy
-    cost: CostStrategy
-    filesystem: FilesystemStrategy
-    identity: IdentityStrategy
-    registry: RegistryStrategy
-    snapshot: SnapshotStrategy
-    storage: StorageStrategy
-
     # runtime params
     job_id: str
     mission_id: str
     setup_id: str
     setup_version_id: str
-    _status: ModuleStatus
-    _task: asyncio.Task | None
 
-    def _init_strategies(self) -> None:
-        """Initialize the services configuration."""
-        for service_name in self.services_config.valid_strategy_names():
-            service = self.services_config.init_strategy(
+    def _init_strategies(self) -> dict[str, Any]:
+        """Initialize the services configuration.
+
+        Returns:
+            dict of services with name: Strategy
+                agent: AgentStrategy
+                cost: CostStrategy
+                filesystem: FilesystemStrategy
+                identity: IdentityStrategy
+                registry: RegistryStrategy
+                snapshot: SnapshotStrategy
+                storage: StorageStrategy
+        """
+        logger.debug("Service initialisation: %s", self.services_config_strategies.keys())
+        return {
+            service_name: self.services_config.init_strategy(
                 service_name,
                 self.mission_id,
                 self.setup_id,
                 self.setup_version_id,
             )
-            setattr(self, service_name, service)
+            for service_name in self.services_config.valid_strategy_names()
+        }
 
     def __init__(
         self,
@@ -110,15 +95,16 @@ class BaseModule(  # noqa: PLR0904
         # SetupVersion reference needed for the precise Kin scope as the cost
         self.setup_version_id: str = setup_version_id
         self._status = ModuleStatus.CREATED
-        self._task: asyncio.Task | None = None
-        # Initialize services configuration
-        self._init_strategies()
 
         # Initialize minimum context
         self.context = ModuleContext(
-            storage=self.storage,
-            cost=self.cost,
-            filesystem=self.filesystem,
+            # Initialize services configuration
+            **self._init_strategies(),
+            session={
+                "mission_id": mission_id,
+                "setup_version_id": setup_version_id,
+                "job_id": job_id,
+            },
         )
 
     @property
@@ -307,7 +293,11 @@ class BaseModule(  # noqa: PLR0904
         """
         return cls.triggers_discoverer.register_trigger(handler_cls)
 
-    async def run_config_setup(self, config_setup_data: SetupModelT) -> SetupModelT:  # noqa: PLR6301
+    async def run_config_setup(  # noqa: PLR6301
+        self,
+        context: ModuleContext,  # noqa: ARG002
+        config_setup_data: SetupModelT,
+    ) -> SetupModelT:
         """Run config setup the module.
 
         The config setup is used to initialize the setup with configuration data.
@@ -323,7 +313,7 @@ class BaseModule(  # noqa: PLR0904
         return config_setup_data
 
     @abstractmethod
-    async def initialize(self, setup_data: SetupModelT) -> None:
+    async def initialize(self, context: ModuleContext, setup_data: SetupModelT) -> None:
         """Initialize the module."""
         raise NotImplementedError
 
@@ -331,7 +321,6 @@ class BaseModule(  # noqa: PLR0904
         self,
         input_data: InputModelT,
         setup_data: SetupModelT,
-        callback: Callable[[OutputModelT], Coroutine[Any, Any, None]],
     ) -> None:
         """Run the module with the given input and setup data.
 
@@ -360,7 +349,6 @@ class BaseModule(  # noqa: PLR0904
         await handler_instance.handle(
             input_instance.root,
             setup_data,
-            callback,
             self.context,
         )
 
@@ -373,7 +361,6 @@ class BaseModule(  # noqa: PLR0904
         self,
         input_data: InputModelT,
         setup_data: SetupModelT,
-        callback: Callable[[OutputModelT], Coroutine[Any, Any, None]],
     ) -> None:
         """Run the module lifecycle.
 
@@ -391,7 +378,7 @@ class BaseModule(  # noqa: PLR0904
                     "job_id": self.job_id,
                 },
             )
-            await self.run(input_data, setup_data, callback)
+            await self.run(input_data, setup_data)
             logger.info(
                 "Module %s finished",
                 self.name,
@@ -428,8 +415,6 @@ class BaseModule(  # noqa: PLR0904
             )
         else:
             self._status = ModuleStatus.STOPPING
-        finally:
-            await self.stop()
 
     async def start(
         self,
@@ -440,15 +425,17 @@ class BaseModule(  # noqa: PLR0904
     ) -> None:
         """Start the module."""
         try:
-            logger.debug("Inititalize module")
-            await self.initialize(setup_data=setup_data)
+            self.context.callbacks.logger = logger
+            self.context.callbacks.send_message = callback
+            logger.info(f"Inititalize module {self.job_id}")
+            await self.initialize(self.context, setup_data)
         except Exception as e:
             self._status = ModuleStatus.FAILED
             short_description = "Error initializing module"
             logger.exception("%s: %s", short_description, e)
             await callback(
                 ModuleCodeModel(
-                    code=str(self._status),
+                    code="Error",
                     short_description=short_description,
                     message=str(e),
                 )
@@ -461,32 +448,22 @@ class BaseModule(  # noqa: PLR0904
         try:
             logger.debug("Init the discovered input handlers.")
             self.triggers_discoverer.init_handlers(self.context)
-            logger.debug("Run lifecycle")
-            self._status = ModuleStatus.RUNNING
-            self._task = asyncio.create_task(
-                self._run_lifecycle(input_data, setup_data, callback),
-                name="module_lifecycle",
-            )
-            if done_callback is not None:
-                self._task.add_done_callback(done_callback)
+            logger.debug(f"Run lifecycle {self.job_id}")
+            await self._run_lifecycle(input_data, setup_data)
         except Exception:
             self._status = ModuleStatus.FAILED
             logger.exception("Error during module lifecyle")
+        finally:
+            await self.stop()
 
     async def stop(self) -> None:
         """Stop the module."""
-        logger.info("Stopping module %s with status %s", self.name, self._status)
-        if self._status not in {ModuleStatus.RUNNING, ModuleStatus.STOPPING}:
-            return
-
+        logger.info("Stopping module %s | job_id=%s", self.name, self.job_id)
         try:
             self._status = ModuleStatus.STOPPING
-            if self._task and not self._task.done():
-                self._task.cancel()
-                with contextlib.suppress(asyncio.CancelledError):
-                    await self._task
             logger.debug("Module %s stopped", self.name)
             await self.cleanup()
+            await self.context.callbacks.send_message(ModuleCodeModel(code="__END_OF_STREAM__"))
             self._status = ModuleStatus.STOPPED
             logger.debug("Module %s cleaned", self.name)
         except Exception:
@@ -510,13 +487,15 @@ class BaseModule(  # noqa: PLR0904
                 },
             )
             self._status = ModuleStatus.RUNNING
-            content = await self.run_config_setup(config_setup_data)
+            self.context.callbacks.set_config_setup = callback
+            content = await self.run_config_setup(self.context, config_setup_data)
 
             wrapper = config_setup_data.model_dump()
             wrapper["content"] = content.model_dump()
             await callback(self.create_setup_model(wrapper))
             self._status = ModuleStatus.STOPPING
         except Exception:
+            logger.error("Error during module lifecyle")
             self._status = ModuleStatus.FAILED
             logger.exception(
                 "Error during module lifecyle",

digitalkin/modules/job_manager/base_job_manager.py

@@ -5,17 +5,18 @@ from collections.abc import AsyncGenerator, AsyncIterator, Callable, Coroutine
 from contextlib import asynccontextmanager
 from typing import Any, Generic
 
-from digitalkin.models import ModuleStatus
 from digitalkin.models.module import InputModelT, OutputModelT, SetupModelT
+from digitalkin.models.module.module import ModuleCodeModel
+from digitalkin.models.module.task_monitor import TaskStatus
 from digitalkin.modules._base_module import BaseModule
 from digitalkin.services.services_config import ServicesConfig
 from digitalkin.services.services_models import ServicesMode
 
 
-class BaseJobManager(abc.ABC, Generic[InputModelT, SetupModelT]):
+class BaseJobManager(abc.ABC, Generic[InputModelT, SetupModelT, OutputModelT]):
     """Abstract base class for managing background module jobs."""
 
-    async def _start(self) -> None:
+    async def start(self) -> None:
         """Start the job manager.
 
         This method initializes any necessary resources or configurations
@@ -24,8 +25,8 @@ class BaseJobManager(abc.ABC, Generic[InputModelT, SetupModelT]):
 
     @staticmethod
     async def job_specific_callback(
-        callback: Callable[[str, OutputModelT], Coroutine[Any, Any, None]], job_id: str
-    ) -> Callable[[OutputModelT], Coroutine[Any, Any, None]]:
+        callback: Callable[[str, OutputModelT | ModuleCodeModel], Coroutine[Any, Any, None]], job_id: str
+    ) -> Callable[[OutputModelT | ModuleCodeModel], Coroutine[Any, Any, None]]:
         """Generate a job-specific callback function.
 
         Args:
@@ -36,7 +37,7 @@ class BaseJobManager(abc.ABC, Generic[InputModelT, SetupModelT]):
             Callable: A wrapped callback function that includes the job ID.
         """
 
-        def callback_wrapper(output_data: OutputModelT) -> Coroutine[Any, Any, None]:
+        def callback_wrapper(output_data: OutputModelT | ModuleCodeModel) -> Coroutine[Any, Any, None]:
             """Wrapper for the callback function.
 
             Args:
@@ -53,12 +54,14 @@ class BaseJobManager(abc.ABC, Generic[InputModelT, SetupModelT]):
         self,
         module_class: type[BaseModule],
         services_mode: ServicesMode,
+        **kwargs,  # noqa: ANN003
     ) -> None:
         """Initialize the job manager.
 
         Args:
             module_class: The class of the module to be managed.
             services_mode: The mode of operation for the services (e.g., ASYNC or SYNC).
+            **kwargs: Additional keyword arguments for the job manager.
         """
         self.module_class = module_class
 
@@ -68,6 +71,7 @@ class BaseJobManager(abc.ABC, Generic[InputModelT, SetupModelT]):
             mode=services_mode,
         )
         setattr(self.module_class, "services_config", services_config)
+        super().__init__(**kwargs)
 
     @abc.abstractmethod  # type: ignore
     @asynccontextmanager  # type: ignore
@@ -156,14 +160,14 @@ class BaseJobManager(abc.ABC, Generic[InputModelT, SetupModelT]):
         """
 
     @abc.abstractmethod
-    async def get_module_status(self, job_id: str) -> ModuleStatus | None:
+    async def get_module_status(self, job_id: str) -> TaskStatus:
         """Retrieve the status of a module job.
 
         Args:
             job_id: The unique identifier of the job.
 
         Returns:
-            ModuleStatus | None: The status of the job, or None if the job does not exist.
+            ModuleStatu: The status of the job.
         """
 
     @abc.abstractmethod

digitalkin/modules/job_manager/single_job_manager.py

@@ -1,6 +1,7 @@
 """Background module manager with single instance."""
 
 import asyncio
+import datetime
 import uuid
 from collections.abc import AsyncGenerator, AsyncIterator
 from contextlib import asynccontextmanager
@@ -9,15 +10,18 @@ from typing import Any, Generic
 import grpc
 
 from digitalkin.logger import logger
-from digitalkin.models import ModuleStatus
 from digitalkin.models.module import InputModelT, OutputModelT, SetupModelT
+from digitalkin.models.module.module import ModuleCodeModel
+from digitalkin.models.module.task_monitor import TaskStatus
 from digitalkin.modules._base_module import BaseModule
 from digitalkin.modules.job_manager.base_job_manager import BaseJobManager
-from digitalkin.modules.job_manager.job_manager_models import StreamCodeModel
+from digitalkin.modules.job_manager.surrealdb_repository import SurrealDBConnection
+from digitalkin.modules.job_manager.task_manager import TaskManager
+from digitalkin.modules.job_manager.task_session import TaskSession
 from digitalkin.services.services_models import ServicesMode
 
 
-class SingleJobManager(BaseJobManager, Generic[InputModelT, SetupModelT]):
+class SingleJobManager(BaseJobManager, TaskManager, Generic[InputModelT, OutputModelT, SetupModelT]):
     """Manages a single instance of a module job.
 
     This class ensures that only one instance of a module job is active at a time.
@@ -25,8 +29,10 @@ class SingleJobManager(BaseJobManager, Generic[InputModelT, SetupModelT]):
     to handle their output data.
     """
 
-    modules: dict[str, BaseModule]
-    queue: dict[str, asyncio.Queue]
+    async def start(self) -> None:
+        """Start manager."""
+        self.channel = SurrealDBConnection("task_manager", datetime.timedelta(seconds=5))
+        await self.channel.init_surreal_instance()
 
     def __init__(
         self,
@@ -40,12 +46,9 @@ class SingleJobManager(BaseJobManager, Generic[InputModelT, SetupModelT]):
             services_mode: The mode of operation for the services (e.g., ASYNC or SYNC).
         """
         super().__init__(module_class, services_mode)
-
         self._lock = asyncio.Lock()
-        self.modules: dict[str, BaseModule] = {}
-        self.queues: dict[str, asyncio.Queue] = {}
 
-    async def generate_config_setup_module_response(self, job_id: str) -> SetupModelT:
+    async def generate_config_setup_module_response(self, job_id: str) -> SetupModelT | ModuleCodeModel:
         """Generate a stream consumer for a module's output data.
 
         This method creates an asynchronous generator that streams output data
@@ -56,16 +59,19 @@ class SingleJobManager(BaseJobManager, Generic[InputModelT, SetupModelT]):
             job_id: The unique identifier of the job.
 
         Returns:
-            SetupModelT: the SetupModelT object fully processed.
+            SetupModelT | ModuleCodeModel: the SetupModelT object fully processed.
         """
-        module = self.modules.get(job_id, None)
-        logger.debug("Module %s found: %s", job_id, module)
+        if (session := self.tasks_sessions.get(job_id, None)) is None:
+            return ModuleCodeModel(
+                code=str(grpc.StatusCode.NOT_FOUND),
+                message=f"Module {job_id} not found",
+            )
 
+        logger.debug("Module %s found: %s", job_id, session.module)
         try:
-            return await self.queues[job_id].get()
+            return await session.queue.get()
         finally:
-            logger.info(f"{job_id=}: {self.queues[job_id].empty()}")
-            del self.queues[job_id]
+            logger.info(f"{job_id=}: {session.queue.empty()}")
 
     async def create_config_setup_instance_job(
         self,
@@ -95,8 +101,7 @@ class SingleJobManager(BaseJobManager, Generic[InputModelT, SetupModelT]):
         job_id = str(uuid.uuid4())
         # TODO: Ensure the job_id is unique.
         module = self.module_class(job_id, mission_id=mission_id, setup_id=setup_id, setup_version_id=setup_version_id)
-        self.modules[job_id] = module
-        self.queues[job_id] = asyncio.Queue()
+        self.tasks_sessions[job_id] = TaskSession(job_id, self.channel, module)
 
         try:
             await module.start_config_setup(
@@ -106,13 +111,13 @@ class SingleJobManager(BaseJobManager, Generic[InputModelT, SetupModelT]):
             logger.debug("Module %s (%s) started successfully", job_id, module.name)
         except Exception:
             # Remove the module from the manager in case of an error.
-            del self.modules[job_id]
+            del self.tasks_sessions[job_id]
             logger.exception("Failed to start module %s: %s", job_id)
             raise
         else:
             return job_id
 
-    async def add_to_queue(self, job_id: str, output_data: OutputModelT) -> None:  # type: ignore
+    async def add_to_queue(self, job_id: str, output_data: OutputModelT | ModuleCodeModel) -> None:
         """Add output data to the queue for a specific job.
 
         This method is used as a callback to handle output data generated by a module job.
@@ -121,7 +126,7 @@ class SingleJobManager(BaseJobManager, Generic[InputModelT, SetupModelT]):
             job_id: The unique identifier of the job.
             output_data: The output data produced by the job.
         """
-        await self.queues[job_id].put(output_data.model_dump())
+        await self.tasks_sessions[job_id].queue.put(output_data.model_dump())
 
     @asynccontextmanager  # type: ignore
     async def generate_stream_consumer(self, job_id: str) -> AsyncIterator[AsyncGenerator[dict[str, Any], None]]:  # type: ignore
@@ -137,39 +142,48 @@ class SingleJobManager(BaseJobManager, Generic[InputModelT, SetupModelT]):
         Yields:
             AsyncGenerator: A stream of output data or error messages.
         """
-        module = self.modules.get(job_id, None)
-
-        logger.debug("Module %s found: %s", job_id, module)
+        if (session := self.tasks_sessions.get(job_id, None)) is None:
 
-        async def _stream() -> AsyncGenerator[dict[str, Any], Any]:
-            """Stream output data from the module.
+            async def _error_gen() -> AsyncGenerator[dict[str, Any], None]:  # noqa: RUF029
+                """Generate an error message for a non-existent module.
 
-            Yields:
-                dict: Output data generated by the module.
-            """
-            if module is None:
+                Yields:
+                    AsyncGenerator: A generator yielding an error message.
+                """
                 yield {
                     "error": {
                         "error_message": f"Module {job_id} not found",
                         "code": grpc.StatusCode.NOT_FOUND,
                     }
                 }
-                return
 
-            try:
-                while module.status == ModuleStatus.RUNNING or (
-                    not self.queues[job_id].empty()
-                    and module.status
-                    in {
-                        ModuleStatus.STOPPED,
-                        ModuleStatus.STOPPING,
-                    }
-                ):
-                    logger.info(f"{job_id=}: {module.status=}")
-                    yield await self.queues[job_id].get()
+            yield _error_gen()
+            return
+
+        logger.debug("Session: %s with Module %s", job_id, session.module)
+
+        async def _stream() -> AsyncGenerator[dict[str, Any], Any]:
+            """Stream output data from the module.
 
-            finally:
-                del self.queues[job_id]
+            Yields:
+                dict: Output data generated by the module.
+            """
+            while True:
+                # if queue is empty but producer not finished yet, block on get()
+                msg = await session.queue.get()
+                try:
+                    yield msg
+                finally:
+                    session.queue.task_done()
+
+                # If the producer marked finished and no more items, break soon:
+                if (
+                    session.is_cancelled.is_set()
+                    or (session.status is TaskStatus.COMPLETED and session.queue.empty())
+                    or session.status is TaskStatus.FAILED
+                ):
+                    # and session.queue.empty():
+                    break
 
         yield _stream()
 
@@ -200,32 +214,21 @@ class SingleJobManager(BaseJobManager, Generic[InputModelT, SetupModelT]):
             Exception: If the module fails to start.
         """
         job_id = str(uuid.uuid4())
-        # TODO: Ensure the job_id is unique.
         module = self.module_class(
             job_id,
             mission_id=mission_id,
             setup_id=setup_id,
             setup_version_id=setup_version_id,
         )
-        self.modules[job_id] = module
-        self.queues[job_id] = asyncio.Queue()
         callback = await self.job_specific_callback(self.add_to_queue, job_id)
 
-        try:
-            await module.start(
-                input_data,
-                setup_data,
-                callback,
-                done_callback=lambda _: asyncio.create_task(callback(StreamCodeModel(code="__END_OF_STREAM__"))),
-            )
-            logger.debug("Module %s (%s) started successfully", job_id, module.name)
-        except Exception:
-            # Remove the module from the manager in case of an error.
-            del self.modules[job_id]
-            logger.exception("Failed to start module %s: %s", job_id)
-            raise
-        else:
-            return job_id
+        await self.create_task(
+            job_id,
+            module,
+            module.start(input_data, setup_data, callback, done_callback=None),
+        )
+        logger.info("Managed task started: '%s'", job_id, extra={"task_id": job_id})
+        return job_id
 
     async def stop_module(self, job_id: str) -> bool:
         """Stop a running module job.
@@ -239,34 +242,37 @@ class SingleJobManager(BaseJobManager, Generic[InputModelT, SetupModelT]):
         Raises:
             Exception: If an error occurs while stopping the module.
         """
+        logger.info(f"STOP required for {job_id=}")
+
         async with self._lock:
-            module = self.modules.get(job_id)
-            if not module:
-                logger.warning(f"Module {job_id} not found")
+            session = self.tasks_sessions.get(job_id)
+
+            if not session:
+                logger.warning(f"session with id: {job_id} not found")
                 return False
             try:
-                await module.stop()
-                # should maybe be added in finally
-                del self.queues[job_id]
-                del self.modules[job_id]
-                logger.debug(f"Module {job_id} ({module.name}) stopped successfully")
+                await session.module.stop()
+
+                if job_id in self.tasks:
+                    await self.cancel_task(job_id)
+                logger.debug(f"session {job_id} ({session.module.name}) stopped successfully")
             except Exception as e:
                 logger.error(f"Error while stopping module {job_id}: {e}")
                 raise
             else:
                 return True
 
-    async def get_module_status(self, job_id: str) -> ModuleStatus | None:
+    async def get_module_status(self, job_id: str) -> TaskStatus:
         """Retrieve the status of a module job.
 
         Args:
             job_id: The unique identifier of the job.
 
         Returns:
-            ModuleStatus | None: The status of the module, or None if it does not exist.
+            ModuleStatus: The status of the module.
         """
-        module = self.modules.get(job_id)
-        return module.status if module else None
+        session = self.tasks_sessions.get(job_id, None)
+        return session.status if session is not None else TaskStatus.FAILED
 
     async def stop_all_modules(self) -> None:
         """Stop all currently running module jobs.
@@ -274,7 +280,7 @@ class SingleJobManager(BaseJobManager, Generic[InputModelT, SetupModelT]):
         This method ensures that all active jobs are gracefully terminated.
         """
         async with self._lock:
-            stop_tasks = [self.stop_module(job_id) for job_id in list(self.modules.keys())]
+            stop_tasks = [self.stop_module(job_id) for job_id in list(self.tasks_sessions.keys())]
             if stop_tasks:
                 await asyncio.gather(*stop_tasks, return_exceptions=True)
 
@@ -286,9 +292,9 @@ class SingleJobManager(BaseJobManager, Generic[InputModelT, SetupModelT]):
         """
         return {
             job_id: {
-                "name": module.name,
-                "status": module.status,
-                "class": module.__class__.__name__,
+                "name": session.module.name,
+                "status": session.module.status,
+                "class": session.module.__class__.__name__,
             }
-            for job_id, module in self.modules.items()
+            for job_id, session in self.tasks_sessions.items()
         }