Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype to production without rebuilding the stack.
Python agentagentic-aiagentic-frameworkagentic-workflowaiautomationclaudedeepseekgenaigptllmlow-codemcpnextjsno-codeopenaipythonskillsworkflow
"""Shared suggested-question capability for an admitted app and message actor.
Callers own access admission, supported app modes, quotas and HTTP errors.
The app owner tenant is independent of an account's active workspace.
"""
from collections.abc import Callable, Mapping
from contextlib import AbstractContextManager
from dataclasses import dataclass
from typing import Literal, Protocol
from services.entities.message_entities import MessageActor
from services.errors.message import SuggestedQuestionsAfterAnswerDisabledError
type SuggestedQuestionsInvokeFrom = Literal["explore", "debugger", "web-app", "service-api"]
class MessageSuggestedQuestions(Protocol):
def get_suggested_questions(
self,
*,
app_id: str,
app_owner_tenant_id: str,
expected_app_mode: str,
actor: MessageActor,
invoke_from: SuggestedQuestionsInvokeFrom,
message_id: str,
) -> list[str]: ...
@dataclass(frozen=True, slots=True)
class SuggestedQuestionsContext:
"""An authorized message and its selected configuration, without ORM state."""
app_id: str
tenant_id: str
app_mode: str
message_id: str
conversation_id: str
actor: MessageActor
invoke_from: SuggestedQuestionsInvokeFrom
config: Mapping[str, object]
class SuggestedQuestionsContextQuery[HistoryT](Protocol):
def prepare(
self,
*,
app_id: str,
app_owner_tenant_id: str,
expected_app_mode: str,
actor: MessageActor,
invoke_from: SuggestedQuestionsInvokeFrom,
message_id: str,
) -> SuggestedQuestionsContext | None:
"""Return message configuration, or None when there is no selected workflow."""
...
def load_history(self, *, context: SuggestedQuestionsContext) -> HistoryT:
"""Return detached history after closing its query session."""
...
class SuggestedQuestionsGeneration[HistoryT](Protocol):
def prepare(
self,
*,
context: SuggestedQuestionsContext,
instruction_prompt: str | None,
model_config: object | None,
) -> AbstractContextManager[Callable[[HistoryT], list[str]] | None]:
"""Isolate one generation attempt and resolve its history model.
Yield None when no history model is available; otherwise yield a
generator accepting detached history. None prompt/model inputs select
the built-in prompt and default model.
"""
...
class MessageSuggestedQuestionsService[HistoryT]:
"""Apply the feature policy shared by every message suggested-question endpoint."""
def __init__(
self,
*,
queries: SuggestedQuestionsContextQuery[HistoryT],
generator: SuggestedQuestionsGeneration[HistoryT],
) -> None:
self._queries: SuggestedQuestionsContextQuery[HistoryT] = queries
self._generator: SuggestedQuestionsGeneration[HistoryT] = generator
def get_suggested_questions(
self,
*,
app_id: str,
app_owner_tenant_id: str,
expected_app_mode: str,
actor: MessageActor,
invoke_from: SuggestedQuestionsInvokeFrom,
message_id: str,
) -> list[str]:
context = self._queries.prepare(
app_id=app_id,
app_owner_tenant_id=app_owner_tenant_id,
expected_app_mode=expected_app_mode,
actor=actor,
invoke_from=invoke_from,
message_id=message_id,
)
if context is None:
return []
enabled = context.config.get("enabled", False)
# Agent and workflow features use truthiness; legacy model configs have
# historically disabled this feature only for an explicit False value.
disabled = not enabled if context.app_mode in {"agent", "advanced-chat"} else enabled is False
if disabled:
raise SuggestedQuestionsAfterAnswerDisabledError()
prompt = context.config.get("prompt")
instruction_prompt = prompt if isinstance(prompt, str) and prompt.strip() else None
with self._generator.prepare(
context=context,
instruction_prompt=instruction_prompt,
model_config=context.config.get("model"),
) as generate:
if generate is None:
return []
history = self._queries.load_history(context=context)
return generate(history)