sapiopycommons 2025.8.14a703__py3-none-any.whl → 2026.1.22a847__py3-none-any.whl

sapiopycommons/ai/agent_service_base.py

@@ -0,0 +1,2051 @@
+from __future__ import annotations
+
+import base64
+import json
+import logging
+import os.path
+import re
+import subprocess
+import traceback
+from abc import abstractmethod, ABC
+from enum import Enum
+from logging import Logger
+from os import PathLike
+from subprocess import CompletedProcess
+from typing import Any, Iterable, Mapping, Sequence
+
+from grpc import ServicerContext
+from sapiopylib.rest.User import SapioUser, ensure_logger_initialized
+from sapiopylib.rest.pojo.DataRecord import DataRecord
+from sapiopylib.rest.pojo.DateRange import DateRange
+from sapiopylib.rest.pojo.datatype.FieldDefinition import AbstractVeloxFieldDefinition, SapioDoubleFormat, \
+    SapioStringFormat
+from sapiopylib.rest.utils.recordmodel.PyRecordModel import PyRecordModel, AbstractRecordModel
+from sapiopylib.rest.utils.recordmodel.RecordModelWrapper import WrappedRecordModel
+
+from sapiopycommons.ai.external_credentials import ExternalCredentials
+from sapiopycommons.ai.protoapi.agent.agent_pb2 import AgentDetailsResponsePbo, \
+    AgentDetailsPbo, ProcessStepRequestPbo, ProcessStepResponsePbo, AgentOutputDetailsPbo, AgentIoConfigBasePbo, \
+    AgentInputDetailsPbo, ExampleContainerPbo, ProcessStepResponseStatusPbo, AgentCitationPbo
+from sapiopycommons.ai.protoapi.agent.agent_pb2_grpc import AgentServiceServicer
+from sapiopycommons.ai.protoapi.agent.entry_pb2 import StepOutputBatchPbo, StepItemContainerPbo, \
+    StepBinaryContainerPbo, StepCsvContainerPbo, StepCsvHeaderRowPbo, StepCsvRowPbo, StepJsonContainerPbo, \
+    StepTextContainerPbo
+from sapiopycommons.ai.protoapi.agent.item.item_container_pb2 import ContentTypePbo, StepDataRecordContainerPbo
+from sapiopycommons.ai.protoapi.externalcredentials.external_credentials_pb2 import ExternalCredentialsPbo
+from sapiopycommons.ai.protoapi.fielddefinitions.fields_pb2 import FieldValuePbo, DataRecordPbo
+from sapiopycommons.ai.protoapi.fielddefinitions.velox_field_def_pb2 import VeloxFieldDefPbo, FieldTypePbo, \
+    SelectionPropertiesPbo, IntegerPropertiesPbo, DoublePropertiesPbo, BooleanPropertiesPbo, StringPropertiesPbo, \
+    DatePropertiesPbo, BooleanDependentFieldEntryPbo, SelectionDependentFieldEntryPbo, DateRangePropertiesPbo
+from sapiopycommons.ai.protoapi.session.sapio_conn_info_pb2 import SapioUserSecretTypePbo, SapioConnectionInfoPbo
+from sapiopycommons.ai.protobuf_utils import ProtobufUtils
+from sapiopycommons.files.file_util import FileUtil
+from sapiopycommons.files.temp_files import TempFileHandler
+from sapiopycommons.general.aliases import FieldMap, FieldValue, SapioRecord, AliasUtil
+
+
+# FR-47422: Created classes.
+class ContainerType(Enum):
+    """
+    An enum of the different container contents of a StepItemContainerPbo.
+    """
+    BINARY = "binary"
+    CSV = "csv"
+    DATA_RECORDS = "DataRecord"
+    JSON = "json"
+    TEXT = "text"
+
+
+class AgentResult(ABC):
+    """
+    A class representing a result from a Tool of Tools agent. Instantiate one of the subclasses to create a result
+    object.
+    """
+
+    @abstractmethod
+    def to_pbo(self, content_type: str, file_extensions: list[str] | None = None,
+               data_type_name: str | None = None) -> StepOutputBatchPbo:
+        """
+        Convert this AgentResult object to a StepOutputBatchPbo or list of FieldValueMapPbo proto objects.
+
+        :param content_type: The content type of the result.
+        :param file_extensions: The file extensions of the result. These may be used to download the results as a
+            file within the platform.
+        :param data_type_name: The data type of the result. Only used for DataRecordResults.
+        """
+        pass
+
+
+class BinaryResult(AgentResult):
+    """
+    A class representing binary results from an agent.
+    """
+    binary_data: list[bytes]
+    name: str
+
+    def __init__(self, binary_data: list[bytes], name: str | None = None):
+        """
+        :param binary_data: The binary data as a list of bytes.
+        :param name: An optional identifying name for this result that will be accessible to the next agent.
+        """
+        self.binary_data = binary_data
+        self.name = name
+
+    def to_pbo(self, content_type: str, file_extensions: list[str] | None = None,
+               data_type_name: str | None = None) -> StepOutputBatchPbo:
+        extensions: list[str] = file_extensions if file_extensions else []
+        return StepOutputBatchPbo(
+            item_container=StepItemContainerPbo(
+                content_type=ContentTypePbo(name=content_type, extensions=extensions),
+                container_name=self.name,
+                binary_container=StepBinaryContainerPbo(items=self.binary_data)
+            )
+        )
+
+
+class CsvResult(AgentResult):
+    """
+    A class representing CSV results from an agent.
+    """
+    csv_data: list[dict[str, Any]]
+    name: str
+
+    def __init__(self, csv_data: list[dict[str, Any]], name: str | None = None):
+        """
+        :param csv_data: The list of CSV data results, provided as a list of dictionaries of column name to value.
+        :param name: An optional identifying name for this result that will be accessible to the next agent.
+        """
+        self.csv_data = csv_data
+        self.name = name
+
+    def to_pbo(self, content_type: str, file_extensions: list[str] | None = None,
+               data_type_name: str | None = None) -> StepOutputBatchPbo:
+        return StepOutputBatchPbo(
+            item_container=StepItemContainerPbo(
+                content_type=ContentTypePbo(name=content_type, extensions=[".csv"]),
+                container_name=self.name,
+                csv_container=StepCsvContainerPbo(
+                    header=StepCsvHeaderRowPbo(cells=self.csv_data[0].keys()),
+                    items=[StepCsvRowPbo(cells=[str(x) for x in row.values()]) for row in self.csv_data]
+                )
+            ) if self.csv_data else None
+        )
+
+
+class DataRecordResult(AgentResult):
+    """
+    A class representing data record results from an agent.
+    """
+    new_records: list[FieldMap]
+    existing_records: list[DataRecord]
+    name: str | None
+    data_type_name: str | None
+
+    def __init__(self, new_records: list[FieldMap] | None = None,
+                 existing_records: list[SapioRecord] | None = None, name: str | None = None,
+                 data_type_name: str | None = None):
+        """
+        :param new_records: A list of field maps, where each map is a dictionary of field names to values. Each entry
+            will create a new data record in the system.
+        :param existing_records: A list of records that already exist in the system. Any changes made to the fields of
+            the provided records will be committed to the database after this agent's results are returned.
+        :param name: An optional identifying name for this result that will be accessible to the next agent.
+        :param data_type_name: The data type of the new records created by this result. This is only required if the
+            output configuration related to this result does not have a set output data type.
+        """
+        self.name = name
+        self.data_type_name = data_type_name.strip() if data_type_name else None
+        self.new_records = new_records if new_records is not None else []
+        self.existing_records = []
+        if existing_records is not None:
+            for record in existing_records:
+                record_id: int | None = AliasUtil.to_record_id(record)
+                if record_id is None or record_id <= 0:
+                    raise Exception(f"A record provided as an existing record has an invalid record ID.\n{record}")
+                if isinstance(record, DataRecord):
+                    self.existing_records.append(record)
+                elif isinstance(record, (PyRecordModel, AbstractRecordModel, WrappedRecordModel)):
+                    data_record: DataRecord = record.get_data_record()
+                    data_record.set_fields(record.fields.copy_to_dict())
+                    self.existing_records.append(data_record)
+                else:
+                    raise Exception(f"Unsupported record type: {type(record)}.")
+
+    def to_pbo(self, content_type: str, file_extensions: list[str] | None = None,
+               data_type_name: str | None = None) -> StepOutputBatchPbo:
+        dt: str = data_type_name if data_type_name and data_type_name != "Any" else self.data_type_name
+        data_records: list[DataRecordPbo] = []
+        for field_map in self.new_records:
+            fields: dict[str, FieldValuePbo] = ProtobufUtils.field_map_to_pbo(field_map)
+            data_records.append(DataRecordPbo(data_type_name=dt, record_id=None, fields=fields))
+        for record in self.existing_records:
+            fields: dict[str, FieldValuePbo] = ProtobufUtils.field_map_to_pbo(record.fields)
+            data_records.append(DataRecordPbo(data_type_name=record.data_type_name, record_id=record.record_id,
+                                              fields=fields))
+        return StepOutputBatchPbo(
+            item_container=StepItemContainerPbo(
+                content_type=ContentTypePbo(name=content_type),
+                container_name=self.name,
+                data_record_container=StepDataRecordContainerPbo(items=data_records)
+            )
+        )
+
+
+class JsonResult(AgentResult):
+    """
+    A class representing JSON results from an agent.
+    """
+    json_data: list[dict[str, Any]]
+    name: str
+
+    def __init__(self, json_data: list[dict[str, Any]], name: str | None = None):
+        """
+        :param json_data: The list of JSON data results. Each entry in the list represents a separate JSON object.
+            These entries must be able to be serialized to JSON using json.dumps().
+        :param name: An optional identifying name for this result that will be accessible to the next agent.
+        """
+        # Verify that the given json_data is actually a list of dictionaries.
+        if not isinstance(json_data, list) or not all(isinstance(x, dict) for x in json_data):
+            raise ValueError("json_data must be a list of dictionaries.")
+        self.json_data = json_data
+        self.name = name
+
+    def to_pbo(self, content_type: str, file_extensions: list[str] | None = None,
+               data_type_name: str | None = None) -> StepOutputBatchPbo:
+        return StepOutputBatchPbo(
+            item_container=StepItemContainerPbo(
+                content_type=ContentTypePbo(name=content_type, extensions=[".json"]),
+                container_name=self.name,
+                json_container=StepJsonContainerPbo(items=[json.dumps(x) for x in self.json_data])
+            )
+        )
+
+
+class TextResult(AgentResult):
+    """
+    A class representing text results from an agent.
+    """
+    text_data: list[str]
+    name: str
+
+    def __init__(self, text_data: list[str], name: str | None = None):
+        """
+        :param text_data: The text data as a list of strings.
+        :param name: An optional identifying name for this result that will be accessible to the next agent.
+        """
+        self.text_data = text_data
+        self.name = name
+
+    def to_pbo(self, content_type: str, file_extensions: list[str] | None = None,
+               data_type_name: str | None = None) -> StepOutputBatchPbo:
+        return StepOutputBatchPbo(
+            item_container=StepItemContainerPbo(
+                content_type=ContentTypePbo(name=content_type, extensions=[".txt"]),
+                container_name=self.name,
+                text_container=StepTextContainerPbo(items=self.text_data)
+            )
+        )
+
+
+class AgentServiceBase(AgentServiceServicer, ABC):
+    """
+    A base class for implementing an agent service. Subclasses should implement the register_agents method to register
+    their agents with the service.
+    """
+    debug_mode: bool = False
+
+    def GetAgentDetails(self, request: AgentDetailsPbo, context: ServicerContext) -> AgentDetailsResponsePbo:
+        try:
+            # Get the agent details from the registered agents.
+            details: list[AgentDetailsPbo] = []
+            for agent in self.register_agents():
+                details.append(agent().to_pbo())
+            if not details:
+                raise Exception("No agents registered with this service.")
+            return AgentDetailsResponsePbo(agent_framework_version=self.server_version(), agent_details=details)
+        except Exception as e:
+            # Woe to you if you somehow cause an exception to be raised when just initializing your agents.
+            # There's no way to log this.
+            print(f"CRITICAL ERROR: {e}")
+            print(traceback.format_exc())
+            return AgentDetailsResponsePbo()
+
+    def ProcessData(self, request: ProcessStepRequestPbo, context: ServicerContext) -> ProcessStepResponsePbo:
+        try:
+            # Convert the SapioConnectionInfo proto object to a SapioUser object.
+            user = self._create_user(request.sapio_user)
+            # Locate the agent. If a string is returned instead of an agent object,
+            # that means that the requested agent was not located.
+            agent: AgentBase | str = self.find_agent(user, request, context)
+            if isinstance(agent, str):
+                return ProcessStepResponsePbo(status=ProcessStepResponseStatusPbo.FAILURE, status_message=agent)
+            # Get the agent results from the registered agent matching the request.
+            success, msg, results, logs = self.run(agent)
+            # Convert the results to protobuf objects.
+            output_data: list[StepOutputBatchPbo] = []
+            for result, output_config, file_extensions in zip(results, agent.output_configs,
+                                                              agent.output_file_extensions):
+                base_config: AgentIoConfigBasePbo = output_config.base_config
+                output_data.append(result.to_pbo(base_config.content_type, file_extensions,
+                                                 base_config.data_type_name))
+            # Return a ProcessStepResponse proto object containing the results to the caller.
+            status = ProcessStepResponseStatusPbo.SUCCESS if success else ProcessStepResponseStatusPbo.FAILURE
+            return ProcessStepResponsePbo(status=status, status_message=msg, output=output_data, log=logs)
+        except Exception as e:
+            # This try/except should never be needed, as the agent should handle its own exceptions, but better safe
+            # than sorry.
+            print(f"CRITICAL ERROR: {e}")
+            print(traceback.format_exc())
+            return ProcessStepResponsePbo(status=ProcessStepResponseStatusPbo.FAILURE,
+                                          status_message=f"CRITICAL ERROR: {e}",
+                                          log=[traceback.format_exc()])
+
+    @staticmethod
+    def _create_user(info: SapioConnectionInfoPbo, timeout_seconds: int = 60) -> SapioUser:
+        """
+        Create a SapioUser object from the given SapioConnectionInfo proto object.
+
+        :param info: The SapioConnectionInfo proto object.
+        :param timeout_seconds: The request timeout for calls made from this user object.
+        """
+        guid: str | None = info.app_guid if info.HasField("app_guid") else None
+        user = SapioUser(info.webservice_url.rstrip("/"), True, timeout_seconds, guid=guid)
+        match info.secret_type:
+            case SapioUserSecretTypePbo.SESSION_TOKEN:
+                user.api_token = info.secret
+            case SapioUserSecretTypePbo.PASSWORD:
+                secret: str = info.secret
+                if secret.startswith("Basic "):
+                    secret = secret[6:]
+                credentials: list[str] = base64.b64decode(secret).decode().split(":", 1)
+                user.username = credentials[0]
+                user.password = credentials[1]
+            case _:
+                raise Exception(f"Unexpected secret type: {info.secret_type}")
+        return user
+
+    @staticmethod
+    def server_version() -> int:
+        """
+        :return: The version of this set of agents.
+        """
+        return 1
+
+    @abstractmethod
+    def register_agents(self) -> list[type[AgentBase]]:
+        """
+        Register agent types with this service. Provided agents should implement the AgentBase class.
+
+        :return: A list of agents to register to this service.
+        """
+        pass
+
+    def find_agent(self, user: SapioUser, request: ProcessStepRequestPbo, context: ServicerContext) -> AgentBase | str:
+        """
+        Locate the requested agent from within this agent service.
+
+        :param user: A user object that can be used to initialize manager classes using DataMgmtServer to query the
+            system.
+        :param request: The request object containing the input data.
+        :param context: The gRPC context.
+        :return: The requested agent, or a string reporting that the agent could not be found that lists all agents
+            that this service supports.
+        """
+        # Locate the agent named in the request.
+        agent_name: str = request.agent_name
+        registered_agents: dict[str, type[AgentBase]] = {str(a): a for a in self.register_agents()}
+        # If the agent is not found, list all of the registered agents for this service so that the caller can correct
+        # the agent it is requesting.
+        if agent_name not in registered_agents:
+            all_agent_names: str = "\n".join(registered_agents.keys())
+            msg: str = (f"Agent \"{agent_name}\" not found in the registered agents for this service. The registered "
+                        f"agents for this service are: \n{all_agent_names}")
+            return msg
+
+        # Instantiate the agent class.
+        agent: AgentBase = registered_agents[agent_name]()
+        # Setup the agent with details from the request.
+        agent.setup(user, request, context, self.debug_mode)
+        return agent
+
+    def run(self, agent: AgentBase) -> tuple[bool, str, list[AgentResult], list[str]]:
+        """
+        Execute an agent from this service.
+
+        :param agent: The agent to execute.
+        :return: Whether or not the agent succeeded, the status message, the results of the agent, and any logs
+            generated by the agent.
+        """
+        try:
+            success: bool = True
+            msg: str = ""
+            results: list[AgentResult] = []
+
+            # Validate that the provided inputs match the agent's expected inputs.
+            errors: list[str] | None = self.validate_input(agent)
+            if errors:
+                msg = f"Agent input validation error(s):\n\t" + "\n\t".join(errors)
+                success = False
+
+            # If this is a dry run, then provide the fixed dry run output, even if there were errors with the input.
+            # Otherwise, if the inputs were successfully validated, then the agent is executed normally.
+            if agent.request.dry_run:
+                results = agent.dry_run_output()
+                # If the input was validated without errors, then set the return message to say so.
+                if success:
+                    msg = f"{agent} dry run input successfully validated."
+            elif success:
+                results = agent.run(agent.user)
+                # Verify that the run results match what is expected of this agent.
+                # Otherwise, update the status message to reflect the successful execution of the agent.
+                errors: list[str] | None = self.validate_output(agent, results)
+                if errors:
+                    # Wipe the results list so that malformed results aren't returned
+                    # to the server and mark the run as a failure.
+                    msg = f"Agent output validation error(s):\n\t" + "\n\t".join(errors)
+                    results = []
+                    success = False
+                else:
+                    msg = f"{agent} successfully completed."
+            return success, msg, results, agent.logs
+        except Exception as e:
+            agent.log_exception("Exception occurred during agent execution.", e)
+            return False, str(e), [], agent.logs
+        finally:
+            # Clean up any temporary files created by the agent. If in debug mode, then log the files instead
+            # so that they can be manually inspected.
+            if self.debug_mode:
+                print("Temporary files/directories created during agent execution:")
+                for directory in agent.temp_data.directories:
+                    print(f"\tDirectory: {directory}")
+                for file in agent.temp_data.files:
+                    print(f"\tFile: {file}")
+            else:
+                agent.temp_data.cleanup()
+
+    @staticmethod
+    def validate_input(agent: AgentBase) -> list[str] | None:
+        """
+        Verify that the input from the server matches the expected input according to the agent's input config
+        definitions.
+
+        :param agent: The agent to verify input against.
+        :return: A list of errors with the input, if any.
+        """
+        if len(agent.request.input) != len(agent.input_configs):
+            return [f"Expected {len(agent.input_configs)} inputs for this agent,"
+                    f"but got {len(agent.request.input)} instead."]
+        return agent.validate_input()
+
+    @staticmethod
+    def validate_output(agent: AgentBase, results: list[AgentResult]) -> list[str] | None:
+        """
+        Verify that the output from an agent run matches the expected output according to the agent's output config
+        definitions.
+
+        :param agent: The agent to verify output for.
+        :param results: The results of the agent run.
+        :return: A list of errors with the output, if any.
+        """
+        # Validate the output from the agent. Were we actually given a list of results?
+        if not isinstance(results, list):
+            return [f"Expected a list for results, but got {type(results)} instead."]
+        bad_results: list[Any] = [x for x in results if not isinstance(x, AgentResult)]
+        if bad_results:
+            return [f"The list of results should only contain AgentResult objects. "
+                    f"Got {[type(x) for x in bad_results]} object(s) instead."]
+
+        # Look out for missing or out-of-order outputs.
+        num_expected_outputs: int = len(agent.output_container_types)
+        if len(results) != num_expected_outputs:
+            expected_outputs_str: str = ", ".join(x.value for x in agent.output_container_types)
+            return [f"Expected {num_expected_outputs} outputs for this agent ({expected_outputs_str}), "
+                    f"but got {len(results)} instead."]
+
+        errors: list[str] = []
+        output_index: int = 0
+        for output in results:
+            match agent.output_container_types[output_index]:
+                case ContainerType.BINARY:
+                    if not isinstance(output, BinaryResult):
+                        errors.append(f"Mismatched output type at index {output_index}: expected "
+                                      f"BinaryResult but got {type(output)}.")
+                case ContainerType.CSV:
+                    if not isinstance(output, CsvResult):
+                        errors.append(f"Mismatched output type at index {output_index}: expected "
+                                      f"CsvResult but got {type(output)}.")
+                case ContainerType.DATA_RECORDS:
+                    if not isinstance(output, DataRecordResult):
+                        errors.append(f"Mismatched output type at index {output_index}: expected "
+                                      f"DataRecordResult but got {type(output)}.")
+                        continue
+                    config_dt: str = agent.output_configs[output_index].base_config.data_type_name
+                    result_dt: str = output.data_type_name
+                    if (not config_dt or config_dt == "Any") and not result_dt and output.new_records:
+                        errors.append(f"Output {output_index} was not configured with an output data type, "
+                                      f"is creating new records, and did not specify an output data type name "
+                                      f"on the result-level. Data record results that create new records must "
+                                      f"either be configured with a data type name in the agent's init function, "
+                                      f"or provide a data type name on the result object.")
+                    elif config_dt and config_dt != "Any":
+                        if result_dt:
+                            agent.log_warning(f"Output {output_index} was configured with an output data type "
+                                              f"and also specified a data type on the result-level. The result-level "
+                                              f"data type name will be ignored.")
+                        for record in output.existing_records:
+                            record_dt: str = record.data_type_name
+                            if record_dt != config_dt:
+                                errors.append(f"Output {output_index} expected records of data type "
+                                              f"\"{config_dt}\" but received \"{record_dt}\".\n{record}")
+                    # Leaving record validation up to the platform. For now, at least. May re-implement agent
+                    # server level record validation if we see a need for it.
+                    # else:
+                    #     errors.extend(AgentServiceBase._validate_records_output(agent, config_dt, output))
+                case ContainerType.JSON:
+                    if not isinstance(output, JsonResult):
+                        errors.append(f"Mismatched output type at index {output_index}: expected "
+                                      f"JsonResult but got {type(output)}.")
+                case ContainerType.TEXT:
+                    if not isinstance(output, TextResult):
+                        errors.append(f"Mismatched output type at index {output_index}: expected "
+                                      f"TextResult but got {type(output)}.")
+            output_index += 1
+
+        # Run the agent's output validation, but only if no other errors were encountered.
+        if not errors:
+            errors = agent.validate_output()
+        return errors
+
+    @staticmethod
+    def _validate_records_output(agent: AgentBase, data_type_name: str, result: DataRecordResult) -> list[str]:
+        """
+        UNUSED: Leaving record validation up to the platform.
+        """
+        errors: list[str] = []
+
+        def value_field(fd: VeloxFieldDefPbo) -> bool:
+            valueless_types: list[FieldTypePbo] = [
+                FieldTypePbo.ACTION, FieldTypePbo.PARENTLINK, FieldTypePbo.IDENTIFIER, FieldTypePbo.LINK,
+                FieldTypePbo.MULTIPARENTLINK, FieldTypePbo.CHILDLINK, FieldTypePbo.AUTO_ACCESSION
+            ]
+            return fd.data_field_type not in valueless_types
+
+        data_types: dict[str, AgentDataType] = {x.data_type_name: x for x in agent.define_data_types()}
+        data_type: AgentDataType = data_types.get(data_type_name)
+        if data_type:
+            field_defs: dict[str, VeloxFieldDefPbo] = {x.data_field_name: x for x in data_type.field_defs if value_field(x)}
+
+            for i, field_map in enumerate(result.new_records):
+                errors.extend(AgentServiceBase._validate_field_map(i, field_map, field_defs))
+
+        return errors
+
+    @staticmethod
+    def _validate_field_map(i: int, field_map: FieldMap, field_defs: dict[str, VeloxFieldDefPbo]) -> list[str]:
+        """
+        UNUSED: Leaving record validation up to the platform.
+        """
+        errors: list[str] = []
+
+        missing: set[str] = set(field_defs).difference(field_map.keys())
+        if missing:
+            errors.append(f"Entry {i} is missing field names specified by the AgentDataType: {missing}")
+            return errors
+
+        for field, field_def in field_defs.items():
+            value: Any = field_map[field]
+            if value is None:
+                if field_def.required:
+                    errors.append(f"Entry {i} is missing required value for field {field}.")
+                continue
+            match field_def.data_field_type:
+                case FieldTypePbo.BOOLEAN:
+                    if not isinstance(value, bool):
+                        errors.append(f"Entry {i} field {field} expects a boolean but "
+                                      f"received {type(value)}.")
+                case FieldTypePbo.SHORT | FieldTypePbo.INTEGER | FieldTypePbo.LONG \
+                     | FieldTypePbo.ENUM | FieldTypePbo.DATE | FieldTypePbo.SIDE_LINK:
+                    if not isinstance(value, int):
+                        errors.append(f"Entry {i} field {field} expects an integer but "
+                                      f"received {type(value)}.")
+                case FieldTypePbo.DOUBLE:
+                    if not isinstance(value, (int, float)):
+                        errors.append(f"Entry {i} field {field} expects a float or integer but "
+                                      f"received {type(value)}.")
+                case FieldTypePbo.DATE_RANGE:
+                    if not isinstance(value, str):
+                        errors.append(f"Entry {i} field {field} expects a string but "
+                                      f"received {type(value)}.")
+                    else:
+                        try:
+                            DateRange.from_str(value)
+                        except:
+                            errors.append(f"Entry {i} field {field} expects a date range "
+                                          f"formatted string (<start timestamp>/<end timestamp>) "
+                                          f"but received \"{value}\".")
+                case FieldTypePbo.STRING | FieldTypePbo.SELECTION | FieldTypePbo.PICKLIST \
+                     | FieldTypePbo.ACTION_STRING | FieldTypePbo.FILE_BLOB:
+                    if not isinstance(value, str):
+                        errors.append(f"Entry {i} field {field} expects a string but "
+                                      f"received {type(value)}.")
+
+        return errors
+
+
+class AgentMetaData:
+    """
+    A class that allows agents to self-describe their metadata.
+    """
+    name: str
+    description: str
+    category: str
+    citations: dict[str, str]
+    sub_category: str | None
+    icon: bytes | None
+    license_flag: str | None
+
+    def __init__(self, name: str, description: str, category: str, citations: dict[str, str],
+                 sub_category: str | None = None, icon: bytes | None = None, license_flag: str | None = None):
+        """
+        :param name: The display name of the agent. This should be unique across all agents in the service.
+        :param description: The description of the agent.
+        :param category: The category of the agent. This is used to group similar agents in the pipeline manager.
+        :param citations: Any citations or references for this agent, as a dictionary of citation name to URL.
+        :param sub_category: The sub-category of the agent. Agents with the same category and sub-category will appear
+            as one item in the list of agents in the pipeline manager. Clicking on this item will then expand to allow
+            the selection of which agent in the sub-category should be used. This is used to group multiple agents
+            from the same service into one item as not to clutter the list of agents, as some services may have a large
+            number of individual agents created for them.
+        :param icon: The icon to use for the agent. This will appear in the list of agents and on individual steps
+            in the pipeline manager.
+        :param license_flag: The license flag for this agent. The system must have this license in order to use this
+            agent. If None, the agent is not license locked.
+        """
+        self.name = name
+        self.description = description
+        self.category = category
+        self.sub_category = sub_category
+        self.citations = citations
+        self.icon = icon
+        self.license_flag = license_flag
+
+
+class AgentDataType:
+    """
+    A class that allows agents to self-describe the fields that will be present in the FieldMapResults of the agent.
+    """
+    data_type_name: str
+    display_name: str
+    field_defs: list[VeloxFieldDefPbo]
+    icon: bytes | None
+
+    def __init__(self, data_type_name: str, display_name: str,
+                 field_defs: list[AbstractVeloxFieldDefinition | VeloxFieldDefPbo],
+                 icon: bytes | None = None):
+        """
+        :param data_type_name: The name of the data type to use. The length of the data type name
+            cannot exceed 63 characters.
+        :param display_name: The display name of the data type to use.
+        :param field_defs: The fields to use for this data type. The length of the data field names
+            cannot exceed 63 characters.
+        :param icon: The icon to use for this data type.
+        """
+        self.data_type_name = data_type_name
+        self.display_name = display_name
+        self.icon = icon
+        self.field_defs = []
+        for field_def in field_defs:
+            if isinstance(field_def, AbstractVeloxFieldDefinition):
+                self.field_defs.append(ProtobufUtils.field_def_to_pbo(field_def))
+            else:
+                self.field_defs.append(field_def)
+
+        if len(self.data_type_name) > 63:
+            raise Exception(f"Data type name \"{self.data_type_name}\" exceeds the limit of 63 characters.")
+        for field_def in self.field_defs:
+            if len(field_def.data_field_name) > 63:
+                raise Exception(f"Data field name \"{field_def.data_field_name}\" exceeds the limit of 63 characters.")
+
+
+class AgentBase(ABC):
+    """
+    A base class for implementing an agent.
+    """
+    input_configs: list[AgentInputDetailsPbo]
+    input_container_types: list[ContainerType]
+    output_configs: list[AgentOutputDetailsPbo]
+    output_container_types: list[ContainerType]
+    output_file_extensions: list[list[str]]
+    config_fields: list[VeloxFieldDefPbo]
+
+    logs: list[str]
+    logger: Logger
+    _verbose_logging: bool | None = None
+
+    _temp_data: TempFileHandler | None = None
+
+    _user: SapioUser | None = None
+    _request: ProcessStepRequestPbo | None = None
+    _context: ServicerContext | None = None
+    _debug_mode: bool | None = None
+    
+    __is_setup: bool
+
+    @property
+    def verbose_logging(self) -> bool:
+        if not self.__is_setup:
+            raise Exception("Agent must be set up to respond to a request before accessing this property.")
+        return self._verbose_logging
+
+    @property
+    def temp_data(self) -> TempFileHandler:
+        if not self.__is_setup:
+            raise Exception("Agent must be set up to respond to a request before accessing this property.")
+        return self._temp_data
+
+    @property
+    def user(self) -> SapioUser:
+        if not self.__is_setup:
+            raise Exception("Agent must be set up to respond to a request before accessing this property.")
+        return self._user
+
+    @property
+    def request(self) -> ProcessStepRequestPbo:
+        if not self.__is_setup:
+            raise Exception("Agent must be set up to respond to a request before accessing this property.")
+        return self._request
+
+    @property
+    def context(self) -> ServicerContext:
+        if not self.__is_setup:
+            raise Exception("Agent must be set up to respond to a request before accessing this property.")
+        return self._context
+
+    @property
+    def debug_mode(self) -> bool:
+        if not self.__is_setup:
+            raise Exception("Agent must be set up to respond to a request before accessing this property.")
+        return self._debug_mode
+
+    @classmethod
+    @abstractmethod
+    def identifier(cls):
+        """
+        :return: The unique identifier of the agent. This is used by the system to determine which agent should be
+            updated if an agent is re-imported. This should not be changed after the first time that an agent is
+            imported, otherwise a duplicate agent will be created.
+        """
+        pass
+
+    @classmethod
+    @abstractmethod
+    def agent_meta_data(cls) -> AgentMetaData:
+        """
+        :return: The metadata information of this agent.
+        """
+        pass
+
+    @classmethod
+    @abstractmethod
+    def define_data_types(cls) -> list[AgentDataType] | None:
+        """
+        :return: The information of the data types of this agent, if applicable. When this agent is first
+            imported, these details will be used to create new data types in the system. Note that further
+            customization may be made to the data type after it has been imported that can cause it to differ
+            from this definition.
+        """
+        pass
+
+    def __init__(self):
+        self.__is_setup = False
+        self.input_configs = []
+        self.input_container_types = []
+        self.output_configs = []
+        self.output_container_types = []
+        self.output_file_extensions = []
+        self.config_fields = []
+        self.logs = []
+        self.logger = logging.getLogger(f"AgentBase.{self.__class__.__name__}")
+        ensure_logger_initialized(self.logger)
+
+    def __str__(self):
+        return f"{self.agent_meta_data().name} ({self.identifier()})"
+
+    def __hash__(self):
+        return hash(self.identifier())
+
+    def to_pbo(self) -> AgentDetailsPbo:
+        """
+        :return: The AgentDetailsPbo proto object representing this agent.
+        """
+        metadata: AgentMetaData = self.agent_meta_data()
+        # TODO: Support multiple data types.
+        data_types: list[AgentDataType] | None = self.define_data_types()
+        data_type: AgentDataType | None = data_types[0] if data_types else None
+        return AgentDetailsPbo(
+            import_id=self.identifier(),
+            name=metadata.name,
+            description=metadata.description,
+            category=metadata.category,
+            sub_category=metadata.sub_category,
+            icon=metadata.icon,
+            citation=[AgentCitationPbo(title=x, url=y) for x, y in metadata.citations.items()],
+            license_info=metadata.license_flag,
+            config_fields=self.config_fields,
+            input_configs=self.input_configs,
+            output_configs=self.output_configs,
+            output_data_type_name=data_type.data_type_name if data_type else None,
+            output_data_type_display_name=data_type.display_name if data_type else None,
+            output_type_fields=data_type.field_defs if data_type else None,
+            output_data_type_icon=data_type.icon if data_type else None,
+        )
+
+    def setup(self, user: SapioUser, request: ProcessStepRequestPbo, context: ServicerContext, debug_mode: bool) -> None:
+        """
+        Setup the agent with the user, request, and context. This method can be overridden by subclasses to perform
+        additional setup.
+
+        :param user: A user object that can be used to initialize manager classes using DataMgmtServer to query the
+            system.
+        :param request: The request object containing the input data.
+        :param context: The gRPC context.
+        :param debug_mode: If true, the agent should run in debug mode, providing additional logging and not cleaning
+            up temporary files.
+        """
+        self.__is_setup = True
+        self._user = user
+        self._request = request
+        self._context = context
+        self._verbose_logging = request.verbose_logging
+        self._debug_mode = debug_mode
+        self._temp_data = TempFileHandler()
+
+    @staticmethod
+    def _parse_jsonl(example: str) -> list[dict[str, Any]]:
+        """
+        Given a testing example for a JSON output, parse the JSONL into plain JSON.
+        """
+        # Use a JSON decoder to find all valid JSON strings within the testing example.
+        decoder = json.JSONDecoder()
+        idx: int = 0
+        json_strings: list[str] = []
+        while idx < len(example):
+            try:
+                # Find the next valid JSON object in the example.
+                obj, end = decoder.raw_decode(example, idx)
+                # Extract the exact substring corresponding to this JSON object
+                json_strings.append(example[idx:end].strip())
+                idx = end
+            except json.JSONDecodeError:
+                # Skip invalid character and keep searching
+                idx += 1
+        # If no valid JSON was encountered, then a bad example was given.
+        if not json_strings:
+            raise Exception("No valid JSON encountered in testing example.")
+
+        # At this point, each line is its own top-level JSON object. Verify that every line is a dictionary.
+        # For lines that are dictionaries, parse them as JSON.
+        data: list[dict[str, Any]] = []
+        for line in json_strings:
+            line = line.strip()
+            if not line.startswith("{") and not line.endswith("}"):
+                raise Exception("Testing examples must be JSON dictionaries in the JSONL format. "
+                                "(The top-level object may not be a list.)")
+            data.append(json.loads(line))
+        return data
+
+    @staticmethod
+    def get_file_string(path: str) -> str | None:
+        """
+        :param path: The path of the file to read.
+        :return: The file contents as a string, or None if the file does not exist.
+        """
+        if not os.path.exists(path):
+            return None
+        with open(path, "r") as f:
+            return f.read()
+
+    @staticmethod
+    def get_file_bytes(path: str) -> bytes | None:
+        """
+        :param path: The path of the file to read.
+        :return: The file contents as bytes, or None if the file does not exist.
+        """
+        if not os.path.exists(path):
+            return None
+        with open(path, "rb") as f:
+            return f.read()
+
+    @staticmethod
+    def _build_example_container(example: str | bytes | None) -> ExampleContainerPbo | None:
+        """
+        Create an example container from the given example data.
+        """
+        if example is None:
+            return None
+        if isinstance(example, str):
+            return ExampleContainerPbo(text_example=example)
+        if isinstance(example, bytes):
+            return ExampleContainerPbo(binary_example=example)
+        raise ValueError(f"Unexpected example type: {type(example)}")
+
+    @staticmethod
+    def _data_record_content_type(data_type_name: str | None) -> str:
+        data_type: str = data_type_name.strip() if data_type_name else "Any"
+        return f"{data_type} DataRecord"
+
+    def add_binary_input(self, content_type: str, display_name: str, description: str,
+                         validation: str | None = None, input_count: tuple[int, int] | None = None,
+                         is_paged: bool = False, page_size: tuple[int, int] | None = None,
+                         max_request_bytes: int | None = None) -> None:
+        """
+        Add a binary input configuration to the agent. This determines how many inputs this agent will accept in the
+        plan manager, as well as what those inputs are. The IO number of the input will be set to the current number of
+        inputs. That is, the first time an add_X_input function is called, the IO number will be 0, the second time it
+        is called, the IO number will be 1, and so on.
+
+        :param content_type: The content type of the input.
+        :param display_name: The display name of the input.
+        :param description: The description of the input.
+        :param validation: An optional validation string for the input.
+        :param input_count: A tuple of the minimum and maximum number of inputs allowed for this agent.
+        :param is_paged: If true, this input will be paged. If false, this input will not be paged.
+        :param page_size: A tuple of the minimum and maximum page size for this agent. The input must be paged in order
+            for this to have an effect.
+        :param max_request_bytes: The maximum request size in bytes for this agent.
+        """
+        base_config = AgentIoConfigBasePbo(
+            io_number=len(self.input_configs),
+            content_type=content_type,
+            display_name=display_name,
+            description=description,
+            structure_example=None,
+            # The testing example on the input is never used, hence why it can't be set by this function.
+            # The testing example is only used during dry runs, in which the testing_example of the output
+            # of the previous step is what gets passed to the next step's input validation.
+            testing_example=None
+        )
+        self._add_input_config(ContainerType.BINARY, base_config, validation, input_count, is_paged, page_size,
+                               max_request_bytes)
+
+    def add_csv_input(self, content_type: str, display_name: str, description: str,
+                      structure_example: str | bytes, validation: str | None = None,
+                      input_count: tuple[int, int] | None = None, is_paged: bool = False,
+                      page_size: tuple[int, int] | None = None, max_request_bytes: int | None = None) -> None:
+        """
+        Add a CSV input configuration to the agent. This determines how many inputs this agent will accept in the
+        plan manager, as well as what those inputs are. The IO number of the input will be set to the current number of
+        inputs. That is, the first time an add_X_input function is called, the IO number will be 0, the second time it
+        is called, the IO number will be 1, and so on.
+
+        :param content_type: The content type of the input.
+        :param display_name: The display name of the input.
+        :param description: The description of the input.
+        :param structure_example: An example of the structure of the input. This does not need to be an entirely valid
+            example, and should often be truncated for brevity.
+        :param validation: An optional validation string for the input.
+        :param input_count: A tuple of the minimum and maximum number of inputs allowed for this agent.
+        :param is_paged: If true, this input will be paged. If false, this input will not be paged.
+        :param page_size: A tuple of the minimum and maximum page size for this agent. The input must be paged in order
+            for this to have an effect.
+        :param max_request_bytes: The maximum request size in bytes for this agent.
+        """
+        base_config = AgentIoConfigBasePbo(
+            io_number=len(self.input_configs),
+            content_type=content_type,
+            display_name=display_name,
+            description=description,
+            structure_example=self._build_example_container(structure_example),
+            testing_example=None
+        )
+        self._add_input_config(ContainerType.CSV, base_config, validation, input_count, is_paged, page_size,
+                               max_request_bytes)
+
+    def add_data_record_input(self, data_type_name: str | None, display_name: str, description: str,
+                              structure_example: str | bytes, validation: str | None = None,
+                              input_count: tuple[int, int] | None = None, is_paged: bool = False,
+                              page_size: tuple[int, int] | None = None, max_request_bytes: int | None = None) -> None:
+        """
+        Add a data record input configuration to the agent. This determines how many inputs this agent will accept in
+        the plan manager, as well as what those inputs are. The IO number of the input will be set to the current number
+        of inputs. That is, the first time an add_X_input function is called, the IO number will be 0, the second time
+        it is called, the IO number will be 1, and so on.
+
+        :param data_type_name: The data type name for the data records input. If None, then any data record can be
+            provided as input.
+        :param display_name: The display name of the input.
+        :param description: The description of the input.
+        :param structure_example: An example of the structure of the input. This does not need to be an entirely valid
+            example, and should often be truncated for brevity.
+        :param validation: An optional validation string for the input.
+        :param input_count: A tuple of the minimum and maximum number of inputs allowed for this agent.
+        :param is_paged: If true, this input will be paged. If false, this input will not be paged.
+        :param page_size: A tuple of the minimum and maximum page size for this agent. The input must be paged in order
+            for this to have an effect.
+        :param max_request_bytes: The maximum request size in bytes for this agent.
+        """
+        if data_type_name:
+            data_type_name = data_type_name.strip()
+        base_config = AgentIoConfigBasePbo(
+            io_number=len(self.input_configs),
+            content_type=self._data_record_content_type(data_type_name),
+            data_type_name=data_type_name,
+            display_name=display_name,
+            description=description,
+            structure_example=self._build_example_container(structure_example),
+            testing_example=None
+        )
+        self._add_input_config(ContainerType.DATA_RECORDS, base_config, validation, input_count, is_paged, page_size,
+                               max_request_bytes)
+
+    def add_json_input(self, content_type: str, display_name: str, description: str,
+                       structure_example: str | bytes, validation: str | None = None,
+                       input_count: tuple[int, int] | None = None, is_paged: bool = False,
+                       page_size: tuple[int, int] | None = None, max_request_bytes: int | None = None) -> None:
+        """
+        Add a JSON input configuration to the agent. This determines how many inputs this agent will accept in the
+        plan manager, as well as what those inputs are. The IO number of the input will be set to the current number of
+        inputs. That is, the first time an add_X_input function is called, the IO number will be 0, the second time it
+        is called, the IO number will be 1, and so on.
+
+        :param content_type: The content type of the input.
+        :param display_name: The display name of the input.
+        :param description: The description of the input.
+        :param structure_example: An example of the structure of the input. This does not need to be an entirely valid
+            example, and should often be truncated for brevity.
+        :param validation: An optional validation string for the input.
+        :param input_count: A tuple of the minimum and maximum number of inputs allowed for this agent.
+        :param is_paged: If true, this input will be paged. If false, this input will not be paged.
+        :param page_size: A tuple of the minimum and maximum page size for this agent. The input must be paged in order
+            for this to have an effect.
+        :param max_request_bytes: The maximum request size in bytes for this agent.
+        """
+        base_config = AgentIoConfigBasePbo(
+            io_number=len(self.input_configs),
+            content_type=content_type,
+            display_name=display_name,
+            description=description,
+            structure_example=self._build_example_container(structure_example),
+            testing_example=None
+        )
+        self._add_input_config(ContainerType.JSON, base_config, validation, input_count, is_paged, page_size,
+                               max_request_bytes)
+
+    def add_text_input(self, content_type: str, display_name: str, description: str,
+                       structure_example: str | bytes, validation: str | None = None,
+                       input_count: tuple[int, int] | None = None, is_paged: bool = False,
+                       page_size: tuple[int, int] | None = None, max_request_bytes: int | None = None) -> None:
+        """
+        Add a text input configuration to the agent. This determines how many inputs this agent will accept in the
+        plan manager, as well as what those inputs are. The IO number of the input will be set to the current number of
+        inputs. That is, the first time an add_X_input function is called, the IO number will be 0, the second time it
+        is called, the IO number will be 1, and so on.
+
+        :param content_type: The content type of the input.
+        :param display_name: The display name of the input.
+        :param description: The description of the input.
+        :param structure_example: An example of the structure of the input. This does not need to be an entirely valid
+            example, and should often be truncated for brevity.
+        :param validation: An optional validation string for the input.
+        :param input_count: A tuple of the minimum and maximum number of inputs allowed for this agent.
+        :param is_paged: If true, this input will be paged. If false, this input will not be paged.
+        :param page_size: A tuple of the minimum and maximum page size for this agent. The input must be paged in order
+            for this to have an effect.
+        :param max_request_bytes: The maximum request size in bytes for this agent.
+        """
+        base_config = AgentIoConfigBasePbo(
+            io_number=len(self.input_configs),
+            content_type=content_type,
+            display_name=display_name,
+            description=description,
+            structure_example=self._build_example_container(structure_example),
+            testing_example=None
+        )
+        self._add_input_config(ContainerType.TEXT, base_config, validation, input_count, is_paged, page_size,
+                               max_request_bytes)
+
+    def _add_input_config(self, container_type: ContainerType, base_config: AgentIoConfigBasePbo,
+                          validation: str | None = None, input_count: tuple[int, int] | None = None,
+                          is_paged: bool = False, page_size: tuple[int, int] | None = None,
+                          max_request_bytes: int | None = None) -> None:
+        """
+        Add an input configuration to the agent.
+        """
+        self.input_configs.append(AgentInputDetailsPbo(
+            base_config=base_config,
+            validation=validation,
+            min_input_count=input_count[0] if input_count else None,
+            max_input_count=input_count[1] if input_count else None,
+            paged=is_paged,
+            min_page_size=page_size[0] if page_size else None,
+            max_page_size=page_size[1] if page_size else None,
+            max_request_bytes=max_request_bytes,
+        ))
+        self.input_container_types.append(container_type)
+
+    def add_binary_output(self, content_type: str, display_name: str, description: str,
+                          testing_example: str | bytes, file_extensions: list[str] | None = None) -> None:
+        """
+        Add a binary output configuration to the agent. This determines how many outputs this agent will produce in the
+        plan manager, as well as what those outputs are. The IO number of the output will be set to the current number
+        of outputs. That is, the first time an add_X_input function is called, the IO number will be 0, the second time
+        it is called, the IO number will be 1, and so on.
+
+        :param content_type: The content type of the output.
+        :param display_name: The display name of the output.
+        :param description: The description of the output.
+        :param testing_example: An example of the input to be used when testing this agent in the system. This must be
+            an entirely valid example of what an output of this agent could look like so that it can be properly used
+            to run tests with. The provided example may be a string, such as for representing JSON or CSV outputs,
+            or bytes, such as for representing binary outputs like images or files.
+        :param file_extensions: A list of file extensions that this binary output can be saved as.
+        """
+        base_config = AgentIoConfigBasePbo(
+            io_number=len(self.output_configs),
+            content_type=content_type,
+            display_name=display_name,
+            description=description,
+            structure_example=None,
+            testing_example=self._build_example_container(testing_example)
+        )
+        self._add_output_config(ContainerType.BINARY, base_config, file_extensions=file_extensions)
+
+    def add_csv_output(self, content_type: str, display_name: str, description: str,
+                       structure_example: str | bytes, testing_example: str | bytes) -> None:
+        """
+        Add a CSV output configuration to the agent. This determines how many outputs this agent will produce in the
+        plan manager, as well as what those outputs are. The IO number of the output will be set to the current number
+        of outputs. That is, the first time an add_X_input function is called, the IO number will be 0, the second time
+        it is called, the IO number will be 1, and so on.
+
+        :param content_type: The content type of the output.
+        :param display_name: The display name of the output.
+        :param description: The description of the output.
+        :param structure_example: An optional example of the structure of the input. This does not need to be an
+            entirely valid example, and should often be truncated for brevity.
+        :param testing_example: An example of the input to be used when testing this agent in the system. This must be
+            an entirely valid example of what an output of this agent could look like so that it can be properly used
+            to run tests with. The provided example may be a string, such as for representing JSON or CSV outputs,
+            or bytes, such as for representing binary outputs like images or files.
+        """
+        base_config = AgentIoConfigBasePbo(
+            io_number=len(self.output_configs),
+            content_type=content_type,
+            display_name=display_name,
+            description=description,
+            structure_example=self._build_example_container(structure_example),
+            testing_example=self._build_example_container(testing_example)
+        )
+        self._add_output_config(ContainerType.CSV, base_config, file_extensions=[".csv"])
+
+    def add_data_record_output(self, data_type_name: str | None, display_name: str, description: str,
+                               structure_example: str | bytes, testing_example: str | bytes) -> None:
+        """
+        Add a data record output configuration to the agent. This determines how many outputs this agent will produce in
+        the plan manager, as well as what those outputs are. The IO number of the output will be set to the current
+        number of outputs. That is, the first time an add_X_input function is called, the IO number will be 0, the
+        second time it is called, the IO number will be 1, and so on.
+
+        :param data_type_name: The data type name for the data record output. If None, then any data record may be
+            provided as input.
+        :param display_name: The display name of the output.
+        :param description: The description of the output.
+        :param structure_example: An optional example of the structure of the input. This does not need to be an
+            entirely valid example, and should often be truncated for brevity.
+        :param testing_example: An example of the input to be used when testing this agent in the system. This must be
+            an entirely valid example of what an output of this agent could look like so that it can be properly used
+            to run tests with. The provided example may be a string, such as for representing JSON or CSV outputs,
+            or bytes, such as for representing binary outputs like images or files.
+        """
+        if data_type_name:
+            data_type_name = data_type_name.strip()
+        # See add_json_output for details on parsing testing examples.
+        self._parse_jsonl(testing_example)
+        base_config = AgentIoConfigBasePbo(
+            io_number=len(self.output_configs),
+            content_type=self._data_record_content_type(data_type_name),
+            data_type_name=data_type_name,
+            display_name=display_name,
+            description=description,
+            structure_example=self._build_example_container(structure_example),
+            testing_example=self._build_example_container(testing_example)
+        )
+        self._add_output_config(ContainerType.DATA_RECORDS, base_config, file_extensions=[])
+
+    def add_json_output(self, content_type: str, display_name: str, description: str,
+                        structure_example: str | bytes, testing_example: str | bytes) -> None:
+        """
+        Add a JSON output configuration to the agent. This determines how many outputs this agent will produce in the
+        plan manager, as well as what those outputs are. The IO number of the output will be set to the current number
+        of outputs. That is, the first time an add_X_input function is called, the IO number will be 0, the second time
+        it is called, the IO number will be 1, and so on.
+
+        :param content_type: The content type of the output.
+        :param display_name: The display name of the output.
+        :param description: The description of the output.
+        :param structure_example: An optional example of the structure of the input. This does not need to be an
+            entirely valid example, and should often be truncated for brevity.
+        :param testing_example: An example of the input to be used when testing this agent in the system. This must be
+            an entirely valid example of what an output of this agent could look like so that it can be properly used
+            to run tests with. The provided example may be a string, such as for representing JSON or CSV outputs,
+            or bytes, such as for representing binary outputs like images or files.
+        """
+        # The JSON testing example MUST be in the JSONL format with top-level objects being dictionaries.
+        # Attempt to parse it as JSONL. An exception will be thrown if this fails.
+        self._parse_jsonl(testing_example)
+        base_config = AgentIoConfigBasePbo(
+            io_number=len(self.output_configs),
+            content_type=content_type,
+            display_name=display_name,
+            description=description,
+            structure_example=self._build_example_container(structure_example),
+            testing_example=self._build_example_container(testing_example)
+        )
+        self._add_output_config(ContainerType.JSON, base_config, file_extensions=[".json"])
+
+    def add_text_output(self, content_type: str, display_name: str, description: str,
+                        structure_example: str | bytes, testing_example: str | bytes) -> None:
+        """
+        Add a text output configuration to the agent. This determines how many outputs this agent will produce in the
+        plan manager, as well as what those outputs are. The IO number of the output will be set to the current number
+        of outputs. That is, the first time an add_X_input function is called, the IO number will be 0, the second time
+        it is called, the IO number will be 1, and so on.
+
+        :param content_type: The content type of the output.
+        :param display_name: The display name of the output.
+        :param description: The description of the output.
+        :param structure_example: An optional example of the structure of the input. This does not need to be an
+            entirely valid example, and should often be truncated for brevity.
+        :param testing_example: An example of the input to be used when testing this agent in the system. This must be
+            an entirely valid example of what an output of this agent could look like so that it can be properly used
+            to run tests with. The provided example may be a string, such as for representing JSON or CSV outputs,
+            or bytes, such as for representing binary outputs like images or files.
+        """
+        base_config = AgentIoConfigBasePbo(
+            io_number=len(self.output_configs),
+            content_type=content_type,
+            display_name=display_name,
+            description=description,
+            structure_example=self._build_example_container(structure_example),
+            testing_example=self._build_example_container(testing_example)
+        )
+        self._add_output_config(ContainerType.TEXT, base_config, file_extensions=[".txt"])
+
+    def _add_output_config(self, container_type: ContainerType, base_config: AgentIoConfigBasePbo,
+                           file_extensions: list[str] | None = None) -> None:
+        """
+        Add an output configuration to the agent for the given container type.
+        """
+        self.output_configs.append(AgentOutputDetailsPbo(base_config=base_config))
+        self.output_container_types.append(container_type)
+        self.output_file_extensions.append(file_extensions if file_extensions else [])
+
+    def add_config_field(self, field: VeloxFieldDefPbo) -> None:
+        """
+        Add a configuration field to the agent. This field will be used to configure the agent in the plan manager.
+
+        :param field: The configuration field details.
+        """
+        self.config_fields.append(field)
+
+    def add_config_field_def(self, field: AbstractVeloxFieldDefinition) -> None:
+        """
+        Add a configuration field to the agent. This field will be used to configure the agent in the plan manager.
+
+        :param field: The configuration field details.
+        """
+        self.config_fields.append(ProtobufUtils.field_def_to_pbo(field))
+
+    def add_boolean_config_field(self, field_name: str, display_name: str, description: str,
+                                 default_value: bool | None = None, optional: bool = False,
+                                 dependencies: dict[bool, list[str]] | None = None,
+                                 is_hide_disabled_fields: bool = False) -> None:
+        """
+        Add a boolean configuration field to the agent. This field will be used to configure the agent in the plan
+        manager.
+
+        :param field_name: The name of the field.
+        :param display_name: The display name of the field.
+        :param description: The description of the field.
+        :param default_value: The default value of the field.
+        :param optional: If true, this field is optional. If false, this field is required.
+        :param dependencies: A dictionary of field dependencies. The value of the dictionary is a possible value of
+            this field, and the key is a list of field names for other config fields of this agent that will be
+            disabled if the config field matches the corresponding value.
+        :param is_hide_disabled_fields: If true, fields disabled by a field dependency will be hidden. If false, the
+            dependent fields will be visible, but uneditable.
+        """
+        dependent_fields: list[BooleanDependentFieldEntryPbo] | None = None
+        if dependencies:
+            dependent_fields = []
+            for key, values in dependencies.items():
+                dependent_fields.append(BooleanDependentFieldEntryPbo(key=key, dependent_field_names=values))
+        self.config_fields.append(VeloxFieldDefPbo(
+            data_field_type=FieldTypePbo.BOOLEAN,
+            data_field_name=field_name,
+            display_name=display_name,
+            description=description,
+            required=not optional,
+            editable=True,
+            boolean_properties=BooleanPropertiesPbo(
+                default_value=default_value,
+                dependent_fields=dependent_fields,
+                is_hide_disabled_fields=is_hide_disabled_fields
+            )
+        ))
+
+    def add_double_config_field(self, field_name: str, display_name: str, description: str,
+                                default_value: float | None = None, min_value: float = -10.**120,
+                                max_value: float = 10.**120, precision: int = 2, optional: bool = False,
+                                double_format: SapioDoubleFormat | None = None) -> None:
+        """
+        Add a double configuration field to the agent. This field will be used to configure the agent in the plan
+        manager.
+
+        :param field_name: The name of the field.
+        :param display_name: The display name of the field.
+        :param description: The description of the field.
+        :param default_value: The default value of the field.
+        :param min_value: The minimum value of the field.
+        :param max_value: The maximum value of the field.
+        :param precision: The precision of the field.
+        :param optional: If true, this field is optional. If false, this field is required.
+        :param double_format: The format of the field.
+        """
+        self.config_fields.append(VeloxFieldDefPbo(
+            data_field_type=FieldTypePbo.DOUBLE,
+            data_field_name=field_name,
+            display_name=display_name,
+            description=description,
+            required=not optional,
+            editable=True,
+            double_properties=DoublePropertiesPbo(
+                default_value=default_value,
+                min_value=min_value,
+                max_value=max_value,
+                precision=precision,
+                double_format=ProtobufUtils.double_format_to_pbo(double_format)
+            )
+        ))
+
+    def add_integer_config_field(self, field_name: str, display_name: str, description: str,
+                                 default_value: int | None = None, min_value: int = -2**31, max_value: int = 2**31-1,
+                                 optional: bool = False) -> None:
+        """
+        Add an integer configuration field to the agent. This field will be used to configure the agent in the plan
+        manager.
+
+        :param field_name: The name of the field.
+        :param display_name: The display name of the field.
+        :param description: The description of the field.
+        :param default_value: The default value of the field.
+        :param min_value: The minimum value of the field.
+        :param max_value: The maximum value of the field.
+        :param optional: If true, this field is optional. If false, this field is required.
+        """
+        self.config_fields.append(VeloxFieldDefPbo(
+            data_field_type=FieldTypePbo.INTEGER,
+            data_field_name=field_name,
+            display_name=display_name,
+            description=description,
+            required=not optional,
+            editable=True,
+            integer_properties=IntegerPropertiesPbo(
+                default_value=default_value,
+                min_value=min_value,
+                max_value=max_value
+            )
+        ))
+
+    def add_string_config_field(self, field_name: str, display_name: str, description: str,
+                                default_value: str | None = None, max_length: int = 1000, optional: bool = False,
+                                validation_regex: str | None = None, error_msg: str | None = None,
+                                string_format: SapioStringFormat | None = None) -> None:
+        """
+        Add a string configuration field to the agent. This field will be used to configure the agent in the plan
+        manager.
+
+        :param field_name: The name of the field.
+        :param display_name: The display name of the field.
+        :param description: The description of the field.
+        :param default_value: The default value of the field.
+        :param max_length: The maximum length of the field.
+        :param optional: If true, this field is optional. If false, this field is required.
+        :param validation_regex: An optional regex that the field value must match.
+        :param error_msg: An optional error message to display if the field value does not match the regex.
+        :param string_format: The format of the field.
+        """
+        self.config_fields.append(VeloxFieldDefPbo(
+            data_field_type=FieldTypePbo.STRING,
+            data_field_name=field_name,
+            display_name=display_name,
+            description=description,
+            required=not optional,
+            editable=True,
+            string_properties=StringPropertiesPbo(
+                default_value=default_value,
+                max_length=max_length,
+                field_validator=ProtobufUtils.field_validator_pbo(validation_regex, error_msg),
+                string_format=ProtobufUtils.string_format_to_pbo(string_format)
+            )
+        ))
+
+    def add_list_config_field(self, field_name: str, display_name: str, description: str,
+                              default_value: str | None = None, allowed_values: list[str] | None = None,
+                              direct_edit: bool = False, optional: bool = False,
+                              validation_regex: str | None = None, error_msg: str | None = None,
+                              dependencies: dict[str, list[str]] | None = None,
+                              is_hide_disabled_fields: bool = False) -> None:
+        """
+        Add a list configuration field to the agent. This field will be used to configure the agent in the plan
+        manager.
+
+        :param field_name: The name of the field.
+        :param display_name: The display name of the field.
+        :param description: The description of the field.
+        :param default_value: The default value of the field.
+        :param allowed_values: The list of allowed values for the field.
+        :param direct_edit: If true, the user can enter a value that is not in the list of allowed values. If false,
+            the user can only select from the list of allowed values.
+        :param optional: If true, this field is optional. If false, this field is required.
+        :param validation_regex: An optional regex that the field value must match.
+        :param error_msg: An optional error message to display if the field value does not match the regex.
+        :param dependencies: A dictionary of field values to the fields that should be disabled while this field
+            is set to the key value.
+        :param dependencies: A dictionary of field dependencies. The value of the dictionary is a possible value of
+            this field, and the key is a list of field names for other config fields of this agent that will be
+            disabled if the config field matches the corresponding value.
+        :param is_hide_disabled_fields: If true, fields disabled by a field dependency will be hidden. If false, the
+            dependent fields will be visible, but uneditable.
+        """
+        dependent_fields: list[SelectionDependentFieldEntryPbo] | None = None
+        if dependencies:
+            dependent_fields = []
+            for key, values in dependencies.items():
+                dependent_fields.append(SelectionDependentFieldEntryPbo(key=key, dependent_field_names=values))
+        self.config_fields.append(VeloxFieldDefPbo(
+            data_field_type=FieldTypePbo.SELECTION,
+            data_field_name=field_name,
+            display_name=display_name,
+            description=description,
+            required=not optional,
+            editable=True,
+            selection_properties=SelectionPropertiesPbo(
+                default_value=default_value,
+                static_list_values=allowed_values,
+                direct_edit=direct_edit,
+                field_validator=ProtobufUtils.field_validator_pbo(validation_regex, error_msg),
+                dependent_fields=dependent_fields,
+                is_hide_disabled_fields=is_hide_disabled_fields
+            )
+        ))
+
+    def add_multi_list_config_field(self, field_name: str, display_name: str, description: str,
+                                    default_value: list[str] | None = None, allowed_values: list[str] | None = None,
+                                    direct_edit: bool = False, optional: bool = False,
+                                    validation_regex: str | None = None, error_msg: str | None = None,
+                                    dependencies: dict[str, list[str]] | None = None,
+                                    is_hide_disabled_fields: bool = False) -> None:
+        """
+        Add a multi-select list configuration field to the agent. This field will be used to configure the agent in the
+        plan manager.
+
+        :param field_name: The name of the field.
+        :param display_name: The display name of the field.
+        :param description: The description of the field.
+        :param default_value: The default value of the field.
+        :param allowed_values: The list of allowed values for the field.
+        :param direct_edit: If true, the user can enter a value that is not in the list of allowed values. If false,
+            the user can only select from the list of allowed values.
+        :param optional: If true, this field is optional. If false, this field is required.
+        :param validation_regex: An optional regex that the field value must match.
+        :param error_msg: An optional error message to display if the field value does not match the regex.
+        :param dependencies: A dictionary of field dependencies. The value of the dictionary is a possible value of
+            this field, and the key is a list of field names for other config fields of this agent that will be
+            disabled if the config field matches the corresponding value.
+        :param is_hide_disabled_fields: If true, fields disabled by a field dependency will be hidden. If false, the
+            dependent fields will be visible, but uneditable.
+        """
+        dependent_fields: list[SelectionDependentFieldEntryPbo] | None = None
+        if dependencies:
+            dependent_fields = []
+            for key, values in dependencies.items():
+                dependent_fields.append(SelectionDependentFieldEntryPbo(key=key, dependent_field_names=values))
+        self.config_fields.append(VeloxFieldDefPbo(
+            data_field_type=FieldTypePbo.SELECTION,
+            data_field_name=field_name,
+            display_name=display_name,
+            description=description,
+            required=not optional,
+            editable=True,
+            selection_properties=SelectionPropertiesPbo(
+                default_value=",".join(default_value) if default_value else None,
+                static_list_values=allowed_values,
+                multi_select=True,
+                direct_edit=direct_edit,
+                field_validator=ProtobufUtils.field_validator_pbo(validation_regex, error_msg),
+                dependent_fields=dependent_fields,
+                is_hide_disabled_fields=is_hide_disabled_fields
+            )
+        ))
+
+    def add_date_config_field(self, field_name: str, display_name: str, description: str, optional: bool = False,
+                              date_time_format: str = "MMM dd, yyyy", default_to_today: bool = False,
+                              is_static_date: bool = False) -> None:
+        """
+        Add a date configuration field to the agent. This field will be used to configure the agent in the plan
+        manager.
+
+        :param field_name: The name of the field.
+        :param display_name: The display name of the field.
+        :param description: The description of the field.
+        :param optional: If true, this field is optional. If false, this field is required.
+        :param date_time_format: The format that this date field should appear in. The date format is Java-style.
+            See https://docs.oracle.com/en/java/javase/18/docs/api/java.base/java/text/SimpleDateFormat.html for more
+            details.
+        :param default_to_today: If true, the default value of the field will be set to today's date. If false, the
+            default value will be None.
+        :param is_static_date: If true, the user will input the date as UTC. If false, the user will input the date
+            as local time.
+        """
+        self.config_fields.append(VeloxFieldDefPbo(
+            data_field_type=FieldTypePbo.DATE,
+            data_field_name=field_name,
+            display_name=display_name,
+            description=description,
+            required=not optional,
+            editable=True,
+            date_properties=DatePropertiesPbo(
+                default_value="@Today" if default_to_today else None,
+                static_date=is_static_date,
+                date_time_format=date_time_format
+            )
+        ))
+
+    def add_date_range_config_field(self, field_name: str, display_name: str, description: str, optional: bool = False,
+                                    date_time_format: str = "MMM dd, yyyy", is_static_date: bool = False) -> None:
+        """
+        Add a date range configuration field to the agent. This field will be used to configure the agent in the plan
+        manager. The returned value of a date range field is a string that should be parsed using the DateRange class.
+
+        :param field_name: The name of the field.
+        :param display_name: The display name of the field.
+        :param description: The description of the field.
+        :param optional: If true, this field is optional. If false, this field is required.
+        :param date_time_format: The format that this date field should appear in. The date format is Java-style.
+            See https://docs.oracle.com/en/java/javase/18/docs/api/java.base/java/text/SimpleDateFormat.html for more
+            details.
+        :param is_static_date: If true, the user will input the date as UTC. If false, the user will input the date
+            as local time.
+        """
+        self.config_fields.append(VeloxFieldDefPbo(
+            data_field_type=FieldTypePbo.DATE_RANGE,
+            data_field_name=field_name,
+            display_name=display_name,
+            description=description,
+            required=not optional,
+            editable=True,
+            date_range_properties=DateRangePropertiesPbo(
+                is_static=is_static_date,
+                date_time_format=date_time_format
+            )
+        ))
+
+    def add_credentials_config_field(self, field_name: str, display_name: str, description: str, optional: bool = False,
+                                     category: str | None = None, dependencies: dict[str, list[str]] | None = None,
+                                     is_hide_disabled_fields: bool = False) -> None:
+        """
+        Add a list field that asks the user to choose which credentials to use. This field will be used to
+        configure the agent in the plan manager.
+
+        :param field_name: The name of the field.
+        :param display_name: The display name of the field.
+        :param description: The description of the field.
+        :param optional: If true, this field is optional. If false, this field is required.
+        :param category: If provided, only credentials in this category will be shown to the user.
+        :param dependencies: A dictionary of field dependencies. The value of the dictionary is a possible value of
+            this field, and the key is a list of field names for other config fields of this agent that will be
+            disabled if the config field matches the corresponding value.
+        :param is_hide_disabled_fields: If true, fields disabled by a field dependency will be hidden. If false, the
+            dependent fields will be visible, but uneditable.
+        """
+        dependent_fields: list[SelectionDependentFieldEntryPbo] | None = None
+        if dependencies:
+            dependent_fields = []
+            for key, values in dependencies.items():
+                dependent_fields.append(SelectionDependentFieldEntryPbo(key=key, dependent_field_names=values))
+        self.config_fields.append(VeloxFieldDefPbo(
+            data_field_type=FieldTypePbo.SELECTION,
+            data_field_name=field_name,
+            display_name=display_name,
+            description=description,
+            required=not optional,
+            editable=True,
+            selection_properties=SelectionPropertiesPbo(
+                # A credentials field is just a selection field with its list mode set to [ExternalCredentials].
+                list_mode=f"[ExternalCredentials]{category.strip() if category else ''}",
+                multi_select=False,
+                default_value=None,
+                direct_edit=False,
+                dependent_fields=dependent_fields,
+                is_hide_disabled_fields=is_hide_disabled_fields
+            )
+        ))
+
+    @abstractmethod
+    def validate_input(self) -> list[str] | None:
+        """
+        Validate the request given to this agent. If the request is validly formatted, this method should return None.
+        If the request is not valid, this method should return an error message indicating what is wrong with the
+        request.
+
+        This method should not perform any actual processing of the request. It should only validate the inputs and
+        configurations provided in the request.
+
+        The request inputs can be accessed using the self.get_input_*() methods.
+        The request settings can be accessed using the self.get_config_fields() method.
+        The request itself can be accessed using self.request.
+
+        :return: A list of the error messages if the request is not valid. If the request is valid, return an empty
+            list or None.
+        """
+        pass
+
+    def validate_output(self: list[AgentResult]) -> list[str] | None:
+        """
+        Validate the output returned by this agent. This is an optional check that agents are able to run before
+        returning their results to the server. This is called after the agent service verifies that the results
+        match the output configurations defined by the agent.
+
+        :return: A list of the error messages if the results are not valid. If the results are not valid, return an
+            empty list or None.
+        """
+        pass
+
+    def dry_run_output(self) -> list[AgentResult]:
+        """
+        Provide fixed results for a dry run of this agent. This method should not perform any actual processing of the
+        request. It should only return example outputs that can be used to test the next agent in the plan.
+
+        The default implementation of this method looks at the testing_example field of each output configuration
+        and returns a AgentResult object based on the content type of the output.
+
+        :return: A list of AgentResult objects containing example outputs for this agent. Each result in the list
+            corresponds to a separate output from the agent.
+        """
+        results: list[AgentResult] = []
+        for output, container_type in zip(self.output_configs, self.output_container_types):
+            config: AgentIoConfigBasePbo = output.base_config
+            example: ExampleContainerPbo = config.testing_example
+            match container_type:
+                case ContainerType.BINARY:
+                    example: bytes = example.binary_example
+                    results.append(BinaryResult(binary_data=[example]))
+                case ContainerType.CSV:
+                    example: str = example.text_example
+                    results.append(CsvResult(FileUtil.tokenize_csv(example.encode())[0]))
+                case ContainerType.DATA_RECORDS:
+                    example: str = example.text_example
+                    records: list[DataRecord] = [DataRecord.from_json(x)  for x in self._parse_jsonl(example)]
+                    results.append(DataRecordResult(existing_records=records))
+                case ContainerType.JSON:
+                    example: str = example.text_example
+                    results.append(JsonResult(json_data=self._parse_jsonl(example)))
+                case ContainerType.TEXT:
+                    example: str = example.text_example
+                    results.append(TextResult(text_data=[example]))
+        return results
+
+    @abstractmethod
+    def run(self, user: SapioUser) -> list[AgentResult]:
+        """
+        Execute this agent.
+
+        The request inputs can be accessed using the self.get_input_*() methods.
+        The request settings can be accessed using the self.get_config_fields() method.
+        The request itself can be accessed using self.request.
+
+        :param user: A user object that can be used to initialize manager classes using DataMgmtServer to query the
+            system.
+        :return: A list of AgentResult objects containing the response data. Each result in the list corresponds to
+            a separate output from the agent. Field map results do not appear as agent output in the plan manager,
+            instead appearing as records related to the plan step during the run.
+        """
+        pass
+
+    def get_credentials(self, identifier: str | None = None, display_name: str | None = None,
+                        category: str | None = None) -> ExternalCredentials:
+        """
+        Get credentials for the given criteria.
+
+        :param identifier: The unique identifier for the credentials.
+        :param display_name: The display name of the credentials.
+        :param category: The category that the credentials are in.
+        :return: An ExternalCredentials object containing the credentials for the given category and host.
+        """
+        if not self.__is_setup:
+            raise Exception("Cannot call this function before the agent has been set up to respond to a request.")
+        # Remove leading/trailing whitespace
+        identifier = identifier.strip() if identifier else None
+        display_name = display_name.strip() if display_name else None
+        category = category.strip() if category else None
+
+        matching_creds: list[ExternalCredentialsPbo] = []
+        for cred in self.request.external_credential:
+            # Do case insensitive comparison
+            if identifier and cred.id.lower != identifier.lower():
+                continue
+            if display_name and cred.display_name.lower != display_name.lower():
+                continue
+            if category and cred.category.lower() != category.lower():
+                continue
+            matching_creds.append(cred)
+        if len(matching_creds) == 0:
+            raise ValueError(f"No credentials found for the criteria. "
+                             f"(identifier={identifier}, display_name={display_name}, category={category})")
+        if len(matching_creds) > 1:
+            raise ValueError(f"Multiple credentials found for the given criteria. "
+                             f"(identifier={identifier}, display_name={display_name}, category={category})")
+
+        return ExternalCredentials.from_pbo(matching_creds[0])
+
+    def _get_credentials_from_config(self, value: str) -> ExternalCredentials:
+        """
+        Get credentials given the value of a credentials config field.
+
+        :param value: The value of the credentials config field.
+        :return: An ExternalCredentials object containing the credentials.
+        """
+        if not self.__is_setup:
+            raise Exception("Cannot call this function before the agent has been set up to respond to a request.")
+        # Values should be of the format "Name (Identifier)"
+        match = re.match(r"^(.*) \((.*)\)$", value)
+        if not match:
+            raise ValueError(f"Invalid credentials value '{value}'. Expected format 'Name (Identifier)'.")
+        identifier: str = match.group(2)
+        for cred in self.request.external_credential:
+            if cred.id == identifier:
+                return ExternalCredentials.from_pbo(cred)
+        raise ValueError(f"No credentials found with identifier '{identifier}'.")
+
+    def call_subprocess(self,
+                        args: str | bytes | PathLike[str] | PathLike[bytes] | Sequence[str | bytes | PathLike[str] | PathLike[bytes]],
+                        cwd: str | bytes | PathLike[str] | PathLike[bytes] | None = None,
+                        **kwargs) -> CompletedProcess[str]:
+        """
+        Call a subprocess with the given arguments, logging the command and any errors that occur.
+        This function will raise an exception if the return code of the subprocess is non-zero. The output of the
+        subprocess will be captured and returned as part of the CompletedProcess object.
+
+        :param args: The list of arguments to pass to the subprocess.
+        :param cwd: The working directory to run the subprocess in. If None, the current working directory is used.
+        :param kwargs: Additional keyword arguments to pass to subprocess.run().
+        :return: The CompletedProcess object returned by subprocess.run().
+        """
+        try:
+            self.log_info(f"Running subprocess with command: {' '.join(args)}")
+            p: CompletedProcess[str] = subprocess.run(args, check=True, capture_output=True, text=True, cwd=cwd,
+                                                      **kwargs)
+            if p.stdout:
+                self.log_info(f"STDOUT: {p.stdout}")
+            if p.stderr:
+                self.log_info(f"STDERR: {p.stderr}")
+            return p
+        except subprocess.CalledProcessError as e:
+            self.log_error(f"Error running subprocess. Return code: {e.returncode}")
+            if e.stdout:
+                self.log_error(f"STDOUT: {e.stdout}")
+            if e.stderr:
+                self.log_error(f"STDERR: {e.stderr}")
+            raise
+
+    def log_info(self, message: str) -> None:
+        """
+        Log an info message for this agent. If verbose logging is enabled, this message will be included in the logs
+        returned to the caller. Empty/None inputs will not be logged.
+        
+        Logging info can be done during initialization, but those logs will not be returned to the caller. Other
+        log calls will be returned to the caller, even if done during initialization.
+
+        :param message: The message to log.
+        """
+        if not message:
+            return
+        if self.__is_setup and self.verbose_logging:
+            self.logs.append(f"INFO: {self}: {message}")
+        self.logger.info(message)
+
+    def log_warning(self, message: str) -> None:
+        """
+        Log a warning message for this agent. This message will be included in the logs returned to the caller.
+        Empty/None inputs will not be logged.
+
+        :param message: The message to log.
+        """
+        if not message:
+            return
+        self.logs.append(f"WARNING: {self}: {message}")
+        self.logger.warning(message)
+
+    def log_error(self, message: str) -> None:
+        """
+        Log an error message for this agent. This message will be included in the logs returned to the caller.
+        Empty/None inputs will not be logged.
+
+        :param message: The message to log.
+        """
+        if not message:
+            return
+        self.logs.append(f"ERROR: {self}: {message}")
+        self.logger.error(message)
+
+    def log_exception(self, message: str, e: Exception) -> None:
+        """
+        Log an exception for this agent. This message will be included in the logs returned to the caller.
+        Empty/None inputs will not be logged.
+
+        :param message: The message to log.
+        :param e: The exception to log.
+        """
+        if not message and not e:
+            return
+        self.logs.append(f"EXCEPTION: {self}: {message} - {e}")
+        self.logger.error(f"{message}\n{traceback.format_exc()}")
+
+    def is_input_partial(self, index: int = 0) -> bool:
+        """
+        Check if the input at the given index is marked as partial.
+
+        :param index: The index of the input to check. Defaults to 0. Used for agents that accept multiple inputs.
+        :return: True if the input is marked as partial, False otherwise.
+        """
+        if not self.__is_setup:
+            raise Exception("Cannot call this function before the agent has been set up to respond to a request.")
+        return self.request.input[index].is_partial
+
+    def get_input_name(self, index: int = 0) -> str | None:
+        """
+        Get the name of the input from the request object.
+
+        :param index: The index of the input to parse. Defaults to 0. Used for agents that accept multiple inputs.
+        :return: The name of the input from the request object, or None if no name is set.
+        """
+        if not self.__is_setup:
+            raise Exception("Cannot call this function before the agent has been set up to respond to a request.")
+        return self.request.input[index].item_container.container_name
+
+    def get_input_content_type(self, index: int = 0) -> ContentTypePbo:
+        """
+        Get the content type of the input from the request object.
+
+        :param index: The index of the input to parse. Defaults to 0. Used for agents that accept multiple inputs.
+        :return: The content type of the input from the request object.
+        """
+        if not self.__is_setup:
+            raise Exception("Cannot call this function before the agent has been set up to respond to a request.")
+        return self.request.input[index].item_container.content_type
+
+    def _validate_get_input(self, index: int, get_type: ContainerType) -> None:
+        """
+        Given an index and the container type being requested from it, validate that the agent is setup to respond
+        to a request, that the index is not out of range, and that the input type from the config matches the
+        input type being requested. If any errors are encountered, raise an exception.
+
+        :param index: The index of the input.
+        :param get_type: The type of the input being requested.
+        """
+        if not self.__is_setup:
+            raise Exception("Cannot call this function before the agent has been set up to respond to a request.")
+        total_inputs: int = len(self.request.input)
+        if index >= total_inputs:
+            raise Exception(f"Index out of range. This agent only has {total_inputs} inputs. Attempted to retrieve "
+                            f"input with index {index}.")
+        config: ContainerType = self.input_container_types[index]
+        if config != get_type:
+            raise Exception(f"Input {index} is not a \"{get_type.value}\" input. The container type for this input "
+                            f"is \"{config.value}\".")
+
+    def get_input_binary(self, index: int = 0) -> list[bytes]:
+        """
+        Get the binary data from the request object.
+
+        :param index: The index of the input to parse. Defaults to 0. Used for agents that accept multiple inputs.
+        :return: The binary data from the request object.
+        """
+        self._validate_get_input(index, ContainerType.BINARY)
+        container: StepItemContainerPbo = self.request.input[index].item_container
+        if not container.HasField("binary_container"):
+            return []
+        return list(container.binary_container.items)
+
+    def get_input_csv(self, index: int = 0) -> tuple[list[str], list[dict[str, str]]]:
+        """
+        Parse the CSV data from the request object.
+
+        :param index: The index of the input to parse. Defaults to 0. Used for agents that accept multiple inputs.
+        :return: A tuple containing the header row and the data rows. The header row is a list of strings representing
+            the column names, and the data rows are a list of dictionaries where each dictionary represents a row in the
+            CSV with the column names as keys and the corresponding values as strings.
+        """
+        self._validate_get_input(index, ContainerType.CSV)
+        container: StepItemContainerPbo = self.request.input[index].item_container
+        if not container.HasField("csv_container"):
+            return [], []
+        ret_val: list[dict[str, str]] = []
+        headers: Iterable[str] = container.csv_container.header.cells
+        for row in container.csv_container.items:
+            row_dict: dict[str, str] = {}
+            for header, value in zip(headers, row.cells):
+                row_dict[header] = value
+            ret_val.append(row_dict)
+        return list(headers), ret_val
+
+    def get_input_records(self, index: int = 0) -> list[DataRecord]:
+        """
+        Parse the DATA_RECORD data from the request object.
+
+        :param index: The index of the input to parse. Defaults to 0. Used for agents that accept multiple inputs.
+        :return: A list containing the data records from the request object.
+        """
+        self._validate_get_input(index, ContainerType.DATA_RECORDS)
+        container: StepItemContainerPbo = self.request.input[index].item_container
+        if not container.HasField("data_record_container"):
+            return []
+        ret_val: list[DataRecord] = []
+        for record in container.data_record_container.items:
+            ret_val.append(ProtobufUtils.pbo_to_data_record(record))
+        return ret_val
+
+    def get_input_json(self, index: int = 0) -> list[dict[str, Any]]:
+        """
+        Parse the JSON data from the request object.
+
+        :param index: The index of the input to parse. Defaults to 0. Used for agents that accept multiple inputs.
+        :return: A list of parsed JSON objects, which are represented as dictionaries.
+        """
+        self._validate_get_input(index, ContainerType.JSON)
+        container: StepItemContainerPbo = self.request.input[index].item_container
+        if not container.HasField("json_container"):
+            return []
+        input_json: list[Any] = [json.loads(x) for x in container.json_container.items]
+        # Verify that the given JSON actually is a list of dictionaries. If they aren't then the previous step provided
+        # bad input. Agents are enforced to result in a list of dictionaries when returning JSON data, so this is likely
+        # an error caused by a script or static input step.
+        for i, entry in enumerate(input_json):
+            if not isinstance(entry, dict):
+                raise Exception(f"Element {i} of input {index} is not a dictionary object. All top-level JSON inputs "
+                                f"are expected to be dictionaries.")
+        return input_json
+
+    def get_input_text(self, index: int = 0) -> list[str]:
+        """
+        Parse the text data from the request object.
+
+        :param index: The index of the input to parse. Defaults to 0. Used for agents that accept multiple inputs.
+        :return: A list of text data as strings.
+        """
+        self._validate_get_input(index, ContainerType.TEXT)
+        container: StepItemContainerPbo = self.request.input[index].item_container
+        if not container.HasField("text_container"):
+            return []
+        return list(container.text_container.items)
+
+    def get_config_defs(self) -> dict[str, VeloxFieldDefPbo]:
+        """
+        Get the config field definitions for this agent.
+
+        :return: A dictionary of field definitions, where the keys are the field names and the values are the
+            VeloxFieldDefPbo objects representing the field definitions.
+        """
+        field_defs: dict[str, VeloxFieldDefPbo] = {}
+        for field_def in self.to_pbo().config_fields:
+            field_defs[field_def.data_field_name] = field_def
+        return field_defs
+
+    def get_config_fields(self) -> dict[str, FieldValue | list[str] | DateRange | ExternalCredentials]:
+        """
+        Get the configuration field values from the request object. If a field is not present in the request,
+        the default value from the config definition will be returned. The returned value for each field will match the
+        field type (e.g. numeric fields return int or float, boolean fields return bool, etc.),
+        along with a few special cases:
+        - Multi-select selection list fields will return a list of strings.
+        - Credentials selection list fields will return an ExternalCredentials object.
+        - Date range fields will return a DateRange object.
+
+        :return: A dictionary of configuration field names and their values.
+        """
+        if not self.__is_setup:
+            raise Exception("Cannot call this function before the agent has been set up to respond to a request.")
+        config_fields: dict[str, Any] = {}
+        raw_configs: Mapping[str, FieldValuePbo] = self.request.config_field_values
+        for field_name, field_def in self.get_config_defs().items():
+            # If the field is present in the request, convert the protobuf value to a Python value.
+            if field_name in raw_configs:
+                field_value: FieldValue = ProtobufUtils.field_pbo_to_value(raw_configs[field_name])
+            # If the field isn't present, use the default value from the field definition.
+            else:
+                field_value: FieldValue = ProtobufUtils.field_def_pbo_to_default_value(field_def)
+            # If the field value is None, continue to the next field.
+            if field_value is None:
+                config_fields[field_name] = field_value
+                continue
+            # If the field is a multi-select selection list, split the value by commas and strip whitespace.
+            # If the field is a credentials selection list, convert the string value(s) to an ExternalCredentials.
+            if field_def.data_field_type == FieldTypePbo.SELECTION:
+                if field_def.selection_properties.multi_select:
+                    field_value: list[str] = [x.strip() for x in re.split(r',(?!\s)', field_value) if x.strip()]
+                if field_def.selection_properties.list_mode.startswith("[ExternalCredentials]"):
+                    if isinstance(field_value, list):
+                        field_value: list[ExternalCredentials] = [self._get_credentials_from_config(x) for x in field_value]
+                    else:
+                        field_value: ExternalCredentials = self._get_credentials_from_config(field_value)
+            # If the field type is a date range, convert the string value to a DateRange.
+            elif field_def.data_field_type == FieldTypePbo.DATE_RANGE:
+                field_value: DateRange = DateRange.from_str(field_value)
+            config_fields[field_name] = field_value
+        return config_fields
+
+    def get_current_data_type_name(self) -> str:
+        """
+        :return: The data type name that is currently configured for this agent in the system. If the data type name
+            of the output data type is needed during the agent's run function, this is how it should be accessed, as
+            opposed to getting the name from the output_data_type function, as the data type name that the system is
+            currently using may differ from the initial self-described name.
+        """
+        if not self.__is_setup:
+            raise Exception("Cannot call this function before the agent has been set up to respond to a request.")
+        return self.request.output_data_type_name
+
+    @staticmethod
+    def read_from_json(json_data: list[dict[str, Any]], key: str) -> list[Any]:
+        """
+        From a list of dictionaries, return a list of values for the given key from each dictionary. Skips null values.
+
+        :param json_data: The JSON data to read from.
+        :param key: The key to read the values from.
+        :return: A list of values corresponding to the given key in the JSON data.
+        """
+        ret_val: list[Any] = []
+        for entry in json_data:
+            if key in entry:
+                value = entry[key]
+                if isinstance(value, list):
+                    ret_val.extend(value)
+                elif value is not None:
+                    ret_val.append(value)
+        return ret_val
+
+    @staticmethod
+    def flatten_text(text_data: list[str]) -> list[str]:
+        """
+        From a list of strings that come from a text input, flatten the list by splitting each string on newlines and
+        stripping whitespace. Empty lines will be removed.
+
+        :param text_data: The text data to flatten.
+        :return: A flattened list of strings.
+        """
+        ret_val: list[str] = []
+        for entry in text_data:
+            lines: list[str] = [x.strip() for x in entry.splitlines() if x.strip()]
+            ret_val.extend(lines)
+        return ret_val