edsl 0.1.38.dev4__py3-none-any.whl → 0.1.39__py3-none-any.whl

edsl/jobs/interviews/Interview.py

@@ -2,58 +2,44 @@
 
 from __future__ import annotations
 import asyncio
-from typing import Any, Type, List, Generator, Optional, Union
+from typing import Any, Type, List, Generator, Optional, Union, TYPE_CHECKING
 import copy
+from dataclasses import dataclass
 
-from tenacity import (
-    retry,
-    stop_after_attempt,
-    wait_exponential,
-    retry_if_exception_type,
-    RetryError,
-)
-
-from edsl import CONFIG
-from edsl.surveys.base import EndOfSurvey
-from edsl.exceptions import QuestionAnswerValidationError
-from edsl.exceptions import QuestionAnswerValidationError
-from edsl.data_transfer_models import AgentResponseDict, EDSLResultObjectInput
-
-from edsl.jobs.buckets.ModelBuckets import ModelBuckets
-from edsl.jobs.Answers import Answers
-from edsl.jobs.tasks.QuestionTaskCreator import QuestionTaskCreator
-from edsl.jobs.tasks.TaskCreators import TaskCreators
+# from edsl.jobs.Answers import Answers
+from edsl.jobs.data_structures import Answers
 from edsl.jobs.interviews.InterviewStatusLog import InterviewStatusLog
+from edsl.jobs.interviews.InterviewStatusDictionary import InterviewStatusDictionary
 from edsl.jobs.interviews.InterviewExceptionCollection import (
     InterviewExceptionCollection,
 )
-
-# from edsl.jobs.interviews.InterviewStatusMixin import InterviewStatusMixin
-
-from edsl.surveys.base import EndOfSurvey
-from edsl.jobs.buckets.ModelBuckets import ModelBuckets
 from edsl.jobs.interviews.InterviewExceptionEntry import InterviewExceptionEntry
-from edsl.jobs.tasks.task_status_enum import TaskStatus
-from edsl.jobs.tasks.QuestionTaskCreator import QuestionTaskCreator
-
-
-from edsl import Agent, Survey, Scenario, Cache
-from edsl.language_models import LanguageModel
-from edsl.questions import QuestionBase
-from edsl.agents.InvigilatorBase import InvigilatorBase
-
-from edsl.exceptions.language_models import LanguageModelNoResponseError
+from edsl.jobs.buckets.ModelBuckets import ModelBuckets
+from edsl.jobs.AnswerQuestionFunctionConstructor import (
+    AnswerQuestionFunctionConstructor,
+)
+from edsl.jobs.InterviewTaskManager import InterviewTaskManager
+from edsl.jobs.FetchInvigilator import FetchInvigilator
+from edsl.jobs.RequestTokenEstimator import RequestTokenEstimator
 
-from edsl.jobs.interviews.InterviewStatusLog import InterviewStatusLog
-from edsl.jobs.tokens.InterviewTokenUsage import InterviewTokenUsage
-from edsl.jobs.interviews.InterviewStatusDictionary import InterviewStatusDictionary
 
+if TYPE_CHECKING:
+    from edsl.agents.Agent import Agent
+    from edsl.surveys.Survey import Survey
+    from edsl.scenarios.Scenario import Scenario
+    from edsl.data.Cache import Cache
+    from edsl.language_models.LanguageModel import LanguageModel
+    from edsl.jobs.tokens.InterviewTokenUsage import InterviewTokenUsage
+    from edsl.agents.InvigilatorBase import InvigilatorBase
+    from edsl.language_models.key_management.KeyLookup import KeyLookup
 
-from edsl import CONFIG
 
-EDSL_BACKOFF_START_SEC = float(CONFIG.get("EDSL_BACKOFF_START_SEC"))
-EDSL_BACKOFF_MAX_SEC = float(CONFIG.get("EDSL_BACKOFF_MAX_SEC"))
-EDSL_MAX_ATTEMPTS = int(CONFIG.get("EDSL_MAX_ATTEMPTS"))
+@dataclass
+class InterviewRunningConfig:
+    cache: Optional["Cache"] = (None,)
+    skip_retry: bool = (False,)  # COULD BE SET WITH CONFIG
+    raise_validation_errors: bool = (True,)
+    stop_on_exception: bool = (False,)
 
 
 class Interview:
@@ -70,11 +56,10 @@ class Interview:
         survey: Survey,
         scenario: Scenario,
         model: Type["LanguageModel"],
-        debug: Optional[bool] = False,
         iteration: int = 0,
+        indices: dict = None,  # explain?
         cache: Optional["Cache"] = None,
-        sidecar_model: Optional["LanguageModel"] = None,
-        skip_retry: bool = False,
+        skip_retry: bool = False,  # COULD BE SET WITH CONFIG
         raise_validation_errors: bool = True,
     ):
         """Initialize the Interview instance.
@@ -83,13 +68,12 @@ class Interview:
         :param survey: the survey being administered to the agent.
         :param scenario: the scenario that populates the survey questions.
         :param model: the language model used to answer the questions.
-        :param debug: if True, run without calls to the language model.
+        # :param debug: if True, run without calls to the language model.
         :param iteration: the iteration number of the interview.
         :param cache: the cache used to store the answers.
-        :param sidecar_model: a sidecar model used to answer questions.
 
         >>> i = Interview.example()
-        >>> i.task_creators
+        >>> i.task_manager.task_creators
         {}
 
         >>> i.exceptions
@@ -104,22 +88,27 @@ class Interview:
 
         """
         self.agent = agent
-        self.survey = copy.deepcopy(survey)
+        self.survey = copy.deepcopy(survey)  # why do we need to deepcopy the survey?
         self.scenario = scenario
         self.model = model
-        self.debug = debug
         self.iteration = iteration
-        self.cache = cache
-        self.answers: dict[
-            str, str
-        ] = Answers()  # will get filled in as interview progresses
-        self.sidecar_model = sidecar_model
 
-        # Trackers
-        self.task_creators = TaskCreators()  # tracks the task creators
+        self.answers = Answers()  # will get filled in as interview progresses
+
+        self.task_manager = InterviewTaskManager(
+            survey=self.survey,
+            iteration=iteration,
+        )
+
         self.exceptions = InterviewExceptionCollection()
 
-        self._task_status_log_dict = InterviewStatusLog()
+        self.running_config = InterviewRunningConfig(
+            cache=cache,
+            skip_retry=skip_retry,
+            raise_validation_errors=raise_validation_errors,
+        )
+
+        self.cache = cache
         self.skip_retry = skip_retry
         self.raise_validation_errors = raise_validation_errors
 
@@ -131,6 +120,9 @@ class Interview:
 
         self.failed_questions = []
 
+        self.indices = indices
+        self.initial_hash = hash(self)
+
     @property
     def has_exceptions(self) -> bool:
         """Return True if there are exceptions."""
@@ -142,30 +134,26 @@ class Interview:
 
         The keys are the question names; the values are the lists of status log changes for each task.
         """
-        for task_creator in self.task_creators.values():
-            self._task_status_log_dict[
-                task_creator.question.question_name
-            ] = task_creator.status_log
-        return self._task_status_log_dict
+        return self.task_manager.task_status_logs
 
     @property
     def token_usage(self) -> InterviewTokenUsage:
         """Determine how many tokens were used for the interview."""
-        return self.task_creators.token_usage
+        return self.task_manager.token_usage  # task_creators.token_usage
 
     @property
     def interview_status(self) -> InterviewStatusDictionary:
         """Return a dictionary mapping task status codes to counts."""
-        return self.task_creators.interview_status
+        # return self.task_creators.interview_status
+        return self.task_manager.interview_status
 
-    # region: Serialization
     def to_dict(self, include_exceptions=True, add_edsl_version=True) -> dict[str, Any]:
         """Return a dictionary representation of the Interview instance.
         This is just for hashing purposes.
 
         >>> i = Interview.example()
         >>> hash(i)
-        1217840301076717434
+        193593189022259693
         """
         d = {
             "agent": self.agent.to_dict(add_edsl_version=add_edsl_version),
@@ -177,23 +165,34 @@ class Interview:
         }
         if include_exceptions:
             d["exceptions"] = self.exceptions.to_dict()
+        if hasattr(self, "indices"):
+            d["indices"] = self.indices
         return d
 
     @classmethod
     def from_dict(cls, d: dict[str, Any]) -> "Interview":
         """Return an Interview instance from a dictionary."""
+
+        from edsl.agents.Agent import Agent
+        from edsl.surveys.Survey import Survey
+        from edsl.scenarios.Scenario import Scenario
+        from edsl.language_models.LanguageModel import LanguageModel
+
         agent = Agent.from_dict(d["agent"])
         survey = Survey.from_dict(d["survey"])
         scenario = Scenario.from_dict(d["scenario"])
         model = LanguageModel.from_dict(d["model"])
         iteration = d["iteration"]
-        interview = cls(
-            agent=agent,
-            survey=survey,
-            scenario=scenario,
-            model=model,
-            iteration=iteration,
-        )
+        params = {
+            "agent": agent,
+            "survey": survey,
+            "scenario": scenario,
+            "model": model,
+            "iteration": iteration,
+        }
+        if "indices" in d:
+            params["indices"] = d["indices"]
+        interview = cls(**params)
         if "exceptions" in d:
             exceptions = InterviewExceptionCollection.from_dict(d["exceptions"])
             interview.exceptions = exceptions
@@ -211,311 +210,13 @@ class Interview:
         """
         return hash(self) == hash(other)
 
-    # endregion
-
-    # region: Creating tasks
-    @property
-    def dag(self) -> "DAG":
-        """Return the directed acyclic graph for the survey.
-
-        The DAG, or directed acyclic graph, is a dictionary that maps question names to their dependencies.
-        It is used to determine the order in which questions should be answered.
-        This reflects both agent 'memory' considerations and 'skip' logic.
-        The 'textify' parameter is set to True, so that the question names are returned as strings rather than integer indices.
-
-        >>> i = Interview.example()
-        >>> i.dag == {'q2': {'q0'}, 'q1': {'q0'}}
-        True
-        """
-        return self.survey.dag(textify=True)
-
-    def _build_question_tasks(
-        self,
-        model_buckets: ModelBuckets,
-    ) -> list[asyncio.Task]:
-        """Create a task for each question, with dependencies on the questions that must be answered before this one can be answered.
-
-        :param debug: whether to use debug mode, in which case `InvigilatorDebug` is used.
-        :param model_buckets: the model buckets used to track and control usage rates.
-        """
-        tasks = []
-        for question in self.survey.questions:
-            tasks_that_must_be_completed_before = list(
-                self._get_tasks_that_must_be_completed_before(
-                    tasks=tasks, question=question
-                )
-            )
-            question_task = self._create_question_task(
-                question=question,
-                tasks_that_must_be_completed_before=tasks_that_must_be_completed_before,
-                model_buckets=model_buckets,
-                iteration=self.iteration,
-            )
-            tasks.append(question_task)
-        return tuple(tasks)
-
-    def _get_tasks_that_must_be_completed_before(
-        self, *, tasks: list[asyncio.Task], question: "QuestionBase"
-    ) -> Generator[asyncio.Task, None, None]:
-        """Return the tasks that must be completed before the given question can be answered.
-
-        :param tasks: a list of tasks that have been created so far.
-        :param question: the question for which we are determining dependencies.
-
-        If a question has no dependencies, this will be an empty list, [].
-        """
-        parents_of_focal_question = self.dag.get(question.question_name, [])
-        for parent_question_name in parents_of_focal_question:
-            yield tasks[self.to_index[parent_question_name]]
-
-    def _create_question_task(
-        self,
-        *,
-        question: QuestionBase,
-        tasks_that_must_be_completed_before: list[asyncio.Task],
-        model_buckets: ModelBuckets,
-        iteration: int = 0,
-    ) -> asyncio.Task:
-        """Create a task that depends on the passed-in dependencies that are awaited before the task is run.
-
-        :param question: the question to be answered. This is the question we are creating a task for.
-        :param tasks_that_must_be_completed_before: the tasks that must be completed before the focal task is run.
-        :param model_buckets: the model buckets used to track and control usage rates.
-        :param debug: whether to use debug mode, in which case `InvigilatorDebug` is used.
-        :param iteration: the iteration number for the interview.
-
-        The task is created by a `QuestionTaskCreator`, which is responsible for creating the task and managing its dependencies.
-        It is passed a reference to the function that will be called to answer the question.
-        It is passed a list "tasks_that_must_be_completed_before" that are awaited before the task is run.
-        These are added as a dependency to the focal task.
-        """
-        task_creator = QuestionTaskCreator(
-            question=question,
-            answer_question_func=self._answer_question_and_record_task,
-            token_estimator=self._get_estimated_request_tokens,
-            model_buckets=model_buckets,
-            iteration=iteration,
-        )
-        for task in tasks_that_must_be_completed_before:
-            task_creator.add_dependency(task)
-
-        self.task_creators.update(
-            {question.question_name: task_creator}
-        )  # track this task creator
-        return task_creator.generate_task()
-
-    def _get_estimated_request_tokens(self, question) -> float:
-        """Estimate the number of tokens that will be required to run the focal task."""
-        from edsl.scenarios.FileStore import FileStore
-
-        invigilator = self._get_invigilator(question=question)
-        # TODO: There should be a way to get a more accurate estimate.
-        combined_text = ""
-        file_tokens = 0
-        for prompt in invigilator.get_prompts().values():
-            if hasattr(prompt, "text"):
-                combined_text += prompt.text
-            elif isinstance(prompt, str):
-                combined_text += prompt
-            elif isinstance(prompt, list):
-                for file in prompt:
-                    if isinstance(file, FileStore):
-                        file_tokens += file.size * 0.25
-            else:
-                raise ValueError(f"Prompt is of type {type(prompt)}")
-        return len(combined_text) / 4.0 + file_tokens
-
-    async def _answer_question_and_record_task(
-        self,
-        *,
-        question: "QuestionBase",
-        task=None,
-    ) -> "AgentResponseDict":
-        """Answer a question and records the task."""
-
-        had_language_model_no_response_error = False
-
-        @retry(
-            stop=stop_after_attempt(EDSL_MAX_ATTEMPTS),
-            wait=wait_exponential(
-                multiplier=EDSL_BACKOFF_START_SEC, max=EDSL_BACKOFF_MAX_SEC
-            ),
-            retry=retry_if_exception_type(LanguageModelNoResponseError),
-            reraise=True,
-        )
-        async def attempt_answer():
-            nonlocal had_language_model_no_response_error
-
-            invigilator = self._get_invigilator(question)
-
-            if self._skip_this_question(question):
-                return invigilator.get_failed_task_result(
-                    failure_reason="Question skipped."
-                )
-
-            try:
-                response: EDSLResultObjectInput = (
-                    await invigilator.async_answer_question()
-                )
-                if response.validated:
-                    self.answers.add_answer(response=response, question=question)
-                    self._cancel_skipped_questions(question)
-                else:
-                    # When a question is not validated, it is not added to the answers.
-                    # this should also cancel and dependent children questions.
-                    # Is that happening now?
-                    if (
-                        hasattr(response, "exception_occurred")
-                        and response.exception_occurred
-                    ):
-                        raise response.exception_occurred
-
-            except QuestionAnswerValidationError as e:
-                self._handle_exception(e, invigilator, task)
-                return invigilator.get_failed_task_result(
-                    failure_reason="Question answer validation failed."
-                )
-
-            except asyncio.TimeoutError as e:
-                self._handle_exception(e, invigilator, task)
-                had_language_model_no_response_error = True
-                raise LanguageModelNoResponseError(
-                    f"Language model timed out for question '{question.question_name}.'"
-                )
-
-            except Exception as e:
-                self._handle_exception(e, invigilator, task)
-
-            if "response" not in locals():
-                had_language_model_no_response_error = True
-                raise LanguageModelNoResponseError(
-                    f"Language model did not return a response for question '{question.question_name}.'"
-                )
-
-            # if it gets here, it means the no response error was fixed
-            if (
-                question.question_name in self.exceptions
-                and had_language_model_no_response_error
-            ):
-                self.exceptions.record_fixed_question(question.question_name)
-
-            return response
-
-        try:
-            return await attempt_answer()
-        except RetryError as retry_error:
-            # All retries have failed for LanguageModelNoResponseError
-            original_error = retry_error.last_attempt.exception()
-            self._handle_exception(
-                original_error, self._get_invigilator(question), task
-            )
-            raise original_error  # Re-raise the original error after handling
-
-    def _get_invigilator(self, question: QuestionBase) -> InvigilatorBase:
-        """Return an invigilator for the given question.
-
-        :param question: the question to be answered
-        :param debug: whether to use debug mode, in which case `InvigilatorDebug` is used.
-        """
-        invigilator = self.agent.create_invigilator(
-            question=question,
-            scenario=self.scenario,
-            model=self.model,
-            debug=False,
-            survey=self.survey,
-            memory_plan=self.survey.memory_plan,
-            current_answers=self.answers,
-            iteration=self.iteration,
-            cache=self.cache,
-            sidecar_model=self.sidecar_model,
-            raise_validation_errors=self.raise_validation_errors,
-        )
-        """Return an invigilator for the given question."""
-        return invigilator
-
-    def _skip_this_question(self, current_question: "QuestionBase") -> bool:
-        """Determine if the current question should be skipped.
-
-        :param current_question: the question to be answered.
-        """
-        current_question_index = self.to_index[current_question.question_name]
-
-        answers = self.answers | self.scenario | self.agent["traits"]
-        skip = self.survey.rule_collection.skip_question_before_running(
-            current_question_index, answers
-        )
-        return skip
-
-    def _handle_exception(
-        self, e: Exception, invigilator: "InvigilatorBase", task=None
-    ):
-        import copy
-
-        # breakpoint()
-
-        answers = copy.copy(self.answers)
-        exception_entry = InterviewExceptionEntry(
-            exception=e,
-            invigilator=invigilator,
-            answers=answers,
-        )
-        if task:
-            task.task_status = TaskStatus.FAILED
-        self.exceptions.add(invigilator.question.question_name, exception_entry)
-
-        if self.raise_validation_errors:
-            if isinstance(e, QuestionAnswerValidationError):
-                raise e
-
-        if hasattr(self, "stop_on_exception"):
-            stop_on_exception = self.stop_on_exception
-        else:
-            stop_on_exception = False
-
-        if stop_on_exception:
-            raise e
-
-    def _cancel_skipped_questions(self, current_question: QuestionBase) -> None:
-        """Cancel the tasks for questions that are skipped.
-
-        :param current_question: the question that was just answered.
-
-        It first determines the next question, given the current question and the current answers.
-        If the next question is the end of the survey, it cancels all remaining tasks.
-        If the next question is after the current question, it cancels all tasks between the current question and the next question.
-        """
-        current_question_index: int = self.to_index[current_question.question_name]
-
-        next_question: Union[
-            int, EndOfSurvey
-        ] = self.survey.rule_collection.next_question(
-            q_now=current_question_index,
-            answers=self.answers | self.scenario | self.agent["traits"],
-        )
-
-        next_question_index = next_question.next_q
-
-        def cancel_between(start, end):
-            """Cancel the tasks between the start and end indices."""
-            for i in range(start, end):
-                self.tasks[i].cancel()
-
-        if next_question_index == EndOfSurvey:
-            cancel_between(current_question_index + 1, len(self.survey.questions))
-            return
-
-        if next_question_index > (current_question_index + 1):
-            cancel_between(current_question_index + 1, next_question_index)
-
-    # endregion
-
-    # region: Conducting the interview
     async def async_conduct_interview(
         self,
-        model_buckets: Optional[ModelBuckets] = None,
-        stop_on_exception: bool = False,
-        sidecar_model: Optional["LanguageModel"] = None,
-        raise_validation_errors: bool = True,
+        run_config: Optional["RunConfig"] = None,
+        #     model_buckets: Optional[ModelBuckets] = None,
+        #     stop_on_exception: bool = False,
+        #     raise_validation_errors: bool = True,
+        #     key_lookup: Optional[KeyLookup] = None,
     ) -> tuple["Answers", List[dict[str, Any]]]:
         """
         Conduct an Interview asynchronously.
@@ -524,7 +225,6 @@ class Interview:
         :param model_buckets: a dictionary of token buckets for the model.
         :param debug: run without calls to LLM.
         :param stop_on_exception: if True, stops the interview if an exception is raised.
-        :param sidecar_model: a sidecar model used to answer questions.
 
         Example usage:
 
@@ -538,37 +238,68 @@ class Interview:
         >>> i.exceptions
         {'q0': ...
         >>> i = Interview.example()
-        >>> result, _ = asyncio.run(i.async_conduct_interview(stop_on_exception = True))
+        >>> from edsl.jobs.Jobs import RunConfig, RunParameters, RunEnvironment
+        >>> run_config = RunConfig(parameters = RunParameters(), environment = RunEnvironment())
+        >>> run_config.parameters.stop_on_exception = True
+        >>> result, _ = asyncio.run(i.async_conduct_interview(run_config))
         Traceback (most recent call last):
         ...
         asyncio.exceptions.CancelledError
         """
-        self.sidecar_model = sidecar_model
-        self.stop_on_exception = stop_on_exception
+        from edsl.jobs.Jobs import RunConfig, RunParameters, RunEnvironment
+
+        if run_config is None:
+            run_config = RunConfig(
+                parameters=RunParameters(),
+                environment=RunEnvironment(),
+            )
+        self.stop_on_exception = run_config.parameters.stop_on_exception
 
         # if no model bucket is passed, create an 'infinity' bucket with no rate limits
+        bucket_collection = run_config.environment.bucket_collection
+
+        if bucket_collection:
+            model_buckets = bucket_collection.get(self.model)
+        else:
+            model_buckets = None
+
         if model_buckets is None or hasattr(self.agent, "answer_question_directly"):
             model_buckets = ModelBuckets.infinity_bucket()
 
+        # was "self.tasks" - is that necessary?
+        self.tasks = self.task_manager.build_question_tasks(
+            answer_func=AnswerQuestionFunctionConstructor(
+                self, key_lookup=run_config.environment.key_lookup
+            )(),
+            token_estimator=RequestTokenEstimator(self),
+            model_buckets=model_buckets,
+        )
+
         ## This is the key part---it creates a task for each question,
         ## with dependencies on the questions that must be answered before this one can be answered.
-        self.tasks = self._build_question_tasks(model_buckets=model_buckets)
 
-        ## 'Invigilators' are used to administer the survey
-        self.invigilators = [
-            self._get_invigilator(question) for question in self.survey.questions
-        ]
+        ## 'Invigilators' are used to administer the survey.
+        fetcher = FetchInvigilator(
+            interview=self,
+            current_answers=self.answers,
+            key_lookup=run_config.environment.key_lookup,
+        )
+        self.invigilators = [fetcher(question) for question in self.survey.questions]
         await asyncio.gather(
-            *self.tasks, return_exceptions=not stop_on_exception
-        )  # not stop_on_exception)
+            *self.tasks, return_exceptions=not run_config.parameters.stop_on_exception
+        )
         self.answers.replace_missing_answers_with_none(self.survey)
-        valid_results = list(self._extract_valid_results())
+        valid_results = list(
+            self._extract_valid_results(self.tasks, self.invigilators, self.exceptions)
+        )
         return self.answers, valid_results
 
-    # endregion
-
-    # region: Extracting results and recording errors
-    def _extract_valid_results(self) -> Generator["Answers", None, None]:
+    @staticmethod
+    def _extract_valid_results(
+        tasks: List["asyncio.Task"],
+        invigilators: List["InvigilatorBase"],
+        exceptions: InterviewExceptionCollection,
+    ) -> Generator["Answers", None, None]:
         """Extract the valid results from the list of results.
 
         It iterates through the tasks and invigilators, and yields the results of the tasks that are done.
@@ -577,16 +308,10 @@ class Interview:
 
         >>> i = Interview.example()
         >>> result, _ = asyncio.run(i.async_conduct_interview())
-        >>> results = list(i._extract_valid_results())
-        >>> len(results) == len(i.survey)
-        True
         """
-        assert len(self.tasks) == len(self.invigilators)
-
-        for task, invigilator in zip(self.tasks, self.invigilators):
-            if not task.done():
-                raise ValueError(f"Task {task.get_name()} is not done.")
+        assert len(tasks) == len(invigilators)
 
+        def handle_task(task, invigilator):
             try:
                 result = task.result()
             except asyncio.CancelledError as e:  # task was cancelled
@@ -601,18 +326,22 @@ class Interview:
                     exception=e,
                     invigilator=invigilator,
                 )
-                self.exceptions.add(task.get_name(), exception_entry)
+                exceptions.add(task.get_name(), exception_entry)
+            return result
 
-            yield result
+        for task, invigilator in zip(tasks, invigilators):
+            if not task.done():
+                raise ValueError(f"Task {task.get_name()} is not done.")
 
-    # endregion
+            yield handle_task(task, invigilator)
 
-    # region: Magic methods
     def __repr__(self) -> str:
         """Return a string representation of the Interview instance."""
         return f"Interview(agent = {repr(self.agent)}, survey = {repr(self.survey)}, scenario = {repr(self.scenario)}, model = {repr(self.model)})"
 
-    def duplicate(self, iteration: int, cache: "Cache") -> Interview:
+    def duplicate(
+        self, iteration: int, cache: "Cache", randomize_survey: Optional[bool] = True
+    ) -> Interview:
         """Duplicate the interview, but with a new iteration number and cache.
 
         >>> i = Interview.example()
@@ -621,14 +350,20 @@ class Interview:
         True
 
         """
+        if randomize_survey:
+            new_survey = self.survey.draw()
+        else:
+            new_survey = self.survey
+
         return Interview(
             agent=self.agent,
-            survey=self.survey,
+            survey=new_survey,
             scenario=self.scenario,
             model=self.model,
             iteration=iteration,
-            cache=cache,
-            skip_retry=self.skip_retry,
+            cache=self.running_config.cache,
+            skip_retry=self.running_config.skip_retry,
+            indices=self.indices,
         )
 
     @classmethod