feat(ui): add chat-driven dataset review flow

Move dataset review clarification into the assistant workspace and
rework the review page into a chat-centric layout with execution rails.

Add session-scoped assistant actions for mappings, semantic fields,
and SQL preview generation. Introduce optimistic locking for dataset
review mutations, propagate session versions through API responses,
and mask imported filter values before assistant exposure.

Refresh tests, i18n, and spec artifacts to match the new workflow.

BREAKING CHANGE: dataset review mutation endpoints now require the
X-Session-Version header, and clarification is no longer handled
through ClarificationDialog-based flows
This commit is contained in:
2026-03-26 13:33:12 +03:00
parent 1ae9e03ff7
commit d17f75a65d
74 changed files with 6122 additions and 2970 deletions

View File

@@ -59,6 +59,8 @@ class ClarificationQuestionPayload:
priority: int
state: QuestionState
options: list[dict[str, object]] = field(default_factory=list)
# [/DEF:ClarificationQuestionPayload:Class]
@@ -71,6 +73,8 @@ class ClarificationStateResult:
current_question: Optional[ClarificationQuestionPayload]
session: DatasetReviewSession
changed_findings: List[ValidationFinding] = field(default_factory=list)
# [/DEF:ClarificationStateResult:Class]
@@ -84,6 +88,8 @@ class ClarificationAnswerCommand:
answer_kind: AnswerKind
answer_value: Optional[str]
user: User
# [/DEF:ClarificationAnswerCommand:Class]
@@ -102,6 +108,7 @@ class ClarificationEngine:
# @PURPOSE: Bind repository dependency for clarification persistence operations.
def __init__(self, repository: DatasetReviewSessionRepository) -> None:
self.repository = repository
# [/DEF:ClarificationEngine_init:Function]
# [DEF:build_question_payload:Function]
@@ -127,10 +134,17 @@ class ClarificationEngine:
return None
active_questions = [
question for question in clarification_session.questions
question
for question in clarification_session.questions
if question.state == QuestionState.OPEN
]
active_questions.sort(key=lambda item: (-int(item.priority), item.created_at, item.question_id))
active_questions.sort(
key=lambda item: (
-int(item.priority),
item.created_at,
item.question_id,
)
)
if not active_questions:
clarification_session.current_question_id = None
@@ -184,7 +198,11 @@ class ClarificationEngine:
}
for option in sorted(
selected_question.options,
key=lambda item: (item.display_order, item.label, item.option_id),
key=lambda item: (
item.display_order,
item.label,
item.option_id,
),
)
],
)
@@ -197,6 +215,7 @@ class ClarificationEngine:
},
)
return payload
# [/DEF:build_question_payload:Function]
# [DEF:record_answer:Function]
@@ -208,7 +227,9 @@ class ClarificationEngine:
# @POST: Answer row is persisted before current-question pointer advances; skipped/expert-review items remain unresolved and visible.
# @SIDE_EFFECT: Inserts answer row, mutates question/session states, updates clarification findings, and commits.
# @DATA_CONTRACT: Input[ClarificationAnswerCommand] -> Output[ClarificationStateResult]
def record_answer(self, command: ClarificationAnswerCommand) -> ClarificationStateResult:
def record_answer(
self, command: ClarificationAnswerCommand
) -> ClarificationStateResult:
with belief_scope("ClarificationEngine.record_answer"):
session = command.session
clarification_session = self._get_latest_clarification_session(session)
@@ -223,18 +244,27 @@ class ClarificationEngine:
if question is None:
logger.explore(
"Cannot record clarification answer for foreign or missing question",
extra={"session_id": session.session_id, "question_id": command.question_id},
extra={
"session_id": session.session_id,
"question_id": command.question_id,
},
)
raise ValueError("Clarification question not found")
if question.answer is not None:
logger.explore(
"Rejected duplicate clarification answer submission",
extra={"session_id": session.session_id, "question_id": command.question_id},
extra={
"session_id": session.session_id,
"question_id": command.question_id,
},
)
raise ValueError("Clarification question already answered")
if clarification_session.current_question_id and clarification_session.current_question_id != question.question_id:
if (
clarification_session.current_question_id
and clarification_session.current_question_id != question.question_id
):
logger.explore(
"Rejected answer for non-active clarification question",
extra={
@@ -243,9 +273,13 @@ class ClarificationEngine:
"current_question_id": clarification_session.current_question_id,
},
)
raise ValueError("Only the active clarification question can be answered")
raise ValueError(
"Only the active clarification question can be answered"
)
normalized_answer_value = self._normalize_answer_value(command.answer_kind, command.answer_value, question)
normalized_answer_value = self._normalize_answer_value(
command.answer_kind, command.answer_value, question
)
logger.reason(
"Persisting clarification answer before state advancement",
@@ -260,7 +294,9 @@ class ClarificationEngine:
answer_kind=command.answer_kind,
answer_value=normalized_answer_value,
answered_by_user_id=command.user.id,
impact_summary=self._build_impact_summary(question, command.answer_kind, normalized_answer_value),
impact_summary=self._build_impact_summary(
question, command.answer_kind, normalized_answer_value
),
)
self.repository.db.add(persisted_answer)
self.repository.db.flush()
@@ -284,15 +320,25 @@ class ClarificationEngine:
question.updated_at = datetime.utcnow()
self.repository.db.flush()
clarification_session.resolved_count = self._count_resolved_questions(clarification_session)
clarification_session.remaining_count = self._count_remaining_questions(clarification_session)
clarification_session.summary_delta = self.summarize_progress(clarification_session)
clarification_session.resolved_count = self._count_resolved_questions(
clarification_session
)
clarification_session.remaining_count = self._count_remaining_questions(
clarification_session
)
clarification_session.summary_delta = self.summarize_progress(
clarification_session
)
clarification_session.updated_at = datetime.utcnow()
next_question = self._select_next_open_question(clarification_session)
clarification_session.current_question_id = next_question.question_id if next_question else None
clarification_session.current_question_id = (
next_question.question_id if next_question else None
)
clarification_session.status = (
ClarificationStatus.ACTIVE if next_question else ClarificationStatus.COMPLETED
ClarificationStatus.ACTIVE
if next_question
else ClarificationStatus.COMPLETED
)
if clarification_session.status == ClarificationStatus.COMPLETED:
clarification_session.completed_at = datetime.utcnow()
@@ -304,7 +350,7 @@ class ClarificationEngine:
if clarification_session.current_question_id
else SessionPhase.REVIEW
)
session.last_activity_at = datetime.utcnow()
self.repository.bump_session_version(session)
self.repository.db.commit()
self.repository.db.refresh(session)
@@ -326,6 +372,7 @@ class ClarificationEngine:
session=session,
changed_findings=[changed_finding] if changed_finding else [],
)
# [/DEF:record_answer:Function]
# [DEF:summarize_progress:Function]
@@ -336,6 +383,7 @@ class ClarificationEngine:
resolved = self._count_resolved_questions(clarification_session)
remaining = self._count_remaining_questions(clarification_session)
return f"{resolved} resolved, {remaining} unresolved"
# [/DEF:summarize_progress:Function]
# [DEF:_get_latest_clarification_session:Function]
@@ -353,6 +401,7 @@ class ClarificationEngine:
reverse=True,
)
return ordered_sessions[0]
# [/DEF:_get_latest_clarification_session:Function]
# [DEF:_find_question:Function]
@@ -367,6 +416,7 @@ class ClarificationEngine:
if question.question_id == question_id:
return question
return None
# [/DEF:_find_question:Function]
# [DEF:_select_next_open_question:Function]
@@ -377,35 +427,46 @@ class ClarificationEngine:
clarification_session: ClarificationSession,
) -> Optional[ClarificationQuestion]:
open_questions = [
question for question in clarification_session.questions
question
for question in clarification_session.questions
if question.state == QuestionState.OPEN
]
if not open_questions:
return None
open_questions.sort(key=lambda item: (-int(item.priority), item.created_at, item.question_id))
open_questions.sort(
key=lambda item: (-int(item.priority), item.created_at, item.question_id)
)
return open_questions[0]
# [/DEF:_select_next_open_question:Function]
# [DEF:_count_resolved_questions:Function]
# @COMPLEXITY: 2
# @PURPOSE: Count questions whose answers fully resolved the ambiguity.
def _count_resolved_questions(self, clarification_session: ClarificationSession) -> int:
def _count_resolved_questions(
self, clarification_session: ClarificationSession
) -> int:
return sum(
1
for question in clarification_session.questions
if question.state == QuestionState.ANSWERED
)
# [/DEF:_count_resolved_questions:Function]
# [DEF:_count_remaining_questions:Function]
# @COMPLEXITY: 2
# @PURPOSE: Count questions still unresolved or deferred after clarification interaction.
def _count_remaining_questions(self, clarification_session: ClarificationSession) -> int:
def _count_remaining_questions(
self, clarification_session: ClarificationSession
) -> int:
return sum(
1
for question in clarification_session.questions
if question.state in {QuestionState.OPEN, QuestionState.SKIPPED, QuestionState.EXPERT_REVIEW}
if question.state
in {QuestionState.OPEN, QuestionState.SKIPPED, QuestionState.EXPERT_REVIEW}
)
# [/DEF:_count_remaining_questions:Function]
# [DEF:_normalize_answer_value:Function]
@@ -417,18 +478,28 @@ class ClarificationEngine:
answer_value: Optional[str],
question: ClarificationQuestion,
) -> Optional[str]:
normalized_answer_value = str(answer_value).strip() if answer_value is not None else None
if answer_kind in {AnswerKind.SELECTED, AnswerKind.CUSTOM} and not normalized_answer_value:
raise ValueError("answer_value is required for selected or custom clarification answers")
normalized_answer_value = (
str(answer_value).strip() if answer_value is not None else None
)
if (
answer_kind in {AnswerKind.SELECTED, AnswerKind.CUSTOM}
and not normalized_answer_value
):
raise ValueError(
"answer_value is required for selected or custom clarification answers"
)
if answer_kind == AnswerKind.SELECTED:
allowed_values = {option.value for option in question.options}
if normalized_answer_value not in allowed_values:
raise ValueError("answer_value must match one of the current clarification options")
raise ValueError(
"answer_value must match one of the current clarification options"
)
if answer_kind == AnswerKind.SKIPPED:
return normalized_answer_value or "skipped"
if answer_kind == AnswerKind.EXPERT_REVIEW:
return normalized_answer_value or "expert_review"
return normalized_answer_value
# [/DEF:_normalize_answer_value:Function]
# [DEF:_build_impact_summary:Function]
@@ -445,6 +516,7 @@ class ClarificationEngine:
if answer_kind == AnswerKind.EXPERT_REVIEW:
return f"Clarification for {question.topic_ref} was deferred for expert review."
return f"Clarification for {question.topic_ref} recorded as '{answer_value}'."
# [/DEF:_build_impact_summary:Function]
# [DEF:_upsert_clarification_finding:Function]
@@ -461,8 +533,10 @@ class ClarificationEngine:
caused_by_ref = f"clarification:{question.question_id}"
existing = next(
(
finding for finding in session.findings
if finding.area == FindingArea.CLARIFICATION and finding.caused_by_ref == caused_by_ref
finding
for finding in session.findings
if finding.area == FindingArea.CLARIFICATION
and finding.caused_by_ref == caused_by_ref
),
None,
)
@@ -478,7 +552,9 @@ class ClarificationEngine:
else:
resolution_state = ResolutionState.EXPERT_REVIEW
resolved_at = None
message = f"Clarification for '{question.topic_ref}' requires expert review."
message = (
f"Clarification for '{question.topic_ref}' requires expert review."
)
if existing is None:
existing = ValidationFinding(
@@ -513,6 +589,7 @@ class ClarificationEngine:
existing.title = "Clarification requires expert review"
return existing
# [/DEF:_upsert_clarification_finding:Function]
# [DEF:_derive_readiness_state:Function]
@@ -532,12 +609,15 @@ class ClarificationEngine:
return ReadinessState.CLARIFICATION_NEEDED
return ReadinessState.REVIEW_READY
# [/DEF:_derive_readiness_state:Function]
# [DEF:_derive_recommended_action:Function]
# @COMPLEXITY: 2
# @PURPOSE: Recompute next-action guidance after clarification mutations.
def _derive_recommended_action(self, session: DatasetReviewSession) -> RecommendedAction:
def _derive_recommended_action(
self, session: DatasetReviewSession
) -> RecommendedAction:
clarification_session = self._get_latest_clarification_session(session)
if clarification_session is None:
return session.recommended_action
@@ -546,7 +626,10 @@ class ClarificationEngine:
if clarification_session.remaining_count > 0:
return RecommendedAction.START_CLARIFICATION
return RecommendedAction.REVIEW_DOCUMENTATION
# [/DEF:_derive_recommended_action:Function]
# [/DEF:ClarificationEngine:Class]
# [/DEF:ClarificationEngine:Module]
# [/DEF:ClarificationEngine:Module]