:hedgehog: PostHog is the leading platform for building self-driving products. Our developer tools – AI observability, analytics, session replay, flags, experiments, error tracking, logs, and more – capture all the context agents need to diagnose problems, uncover opportunities, and ship fixes. Steer it all from Slack, web, desktop, or the MCP.
Python ab-testingai-analyticsanalyticscdpdata-warehouseexperimentsfeature-flagsjavascriptproduct-analyticspythonreactsession-replaysurveystypescriptweb-analytics
from functools import cached_property
from typing import Any, cast
from drf_spectacular.utils import OpenApiResponse, extend_schema
from rest_framework import status
from rest_framework.exceptions import APIException, PermissionDenied
from rest_framework.permissions import IsAuthenticated
from rest_framework.request import Request
from rest_framework.response import Response
from rest_framework.viewsets import ViewSet
from posthog.api.mixins import ValidatedRequest, validated_request
from posthog.api.routing import TeamAndOrgViewSetMixin
from posthog.models.user import User
from posthog.permissions import APIScopePermission, PostHogFeatureFlagPermission
from posthog.rate_limit import BurstRateThrottle, SustainedRateThrottle
from products.access_control.backend.facade.user_access_control import UserAccessControl
from ..sandbox import SandboxRunInProgress, load_sandbox_run, start_sandbox_run
from .serializers import SandboxQuestionSerializer, SandboxRunSerializer, SandboxRunStartedSerializer
from .settings import CanonicalTeamTokenPermission
class SandboxConflict(APIException):
status_code = status.HTTP_409_CONFLICT
default_detail = "Wait for the current answer to finish before asking another question."
default_code = "conflict"
class BusinessKnowledgeSandboxViewSet(TeamAndOrgViewSetMixin, ViewSet):
scope_object = "business_knowledge"
# Same gate as settings: resource-level none must not ride a single source grant.
requires_resource_level_access = True
serializer_class = SandboxRunSerializer
permission_classes = [
IsAuthenticated,
APIScopePermission,
PostHogFeatureFlagPermission,
CanonicalTeamTokenPermission,
]
posthog_feature_flag = "product-business-knowledge"
throttle_classes = [BurstRateThrottle, SustainedRateThrottle]
pagination_class = None
http_method_names = ["get", "post", "head", "options"]
@cached_property
def user_access_control(self) -> UserAccessControl:
team = self.team.parent_team or self.team
return UserAccessControl(user=cast(User, self.request.user), team=team, organization_id=self.organization_id)
def dangerously_get_required_scopes(self, request: Request, view: Any) -> list[str] | None:
if self.action == "retrieve":
return ["business_knowledge:read"]
# Asking starts a billable sandbox run, so it needs write like the other mutations.
if self.action == "create":
return ["business_knowledge:write"]
return None
@extend_schema(
request=SandboxQuestionSerializer,
responses={
201: OpenApiResponse(response=SandboxRunStartedSerializer, description="The sandbox run has started."),
403: OpenApiResponse(description="AI data processing is not approved for this organization."),
409: OpenApiResponse(description="This person already has a sandbox run that has not finished."),
},
summary="Ask a business knowledge sandbox question",
description="Start a sandbox agent that can search only this project's business knowledge. Returns immediately.",
)
@validated_request(
request_serializer=SandboxQuestionSerializer,
responses={201: OpenApiResponse(response=SandboxRunStartedSerializer)},
)
def create(self, request: ValidatedRequest, **kwargs: Any) -> Response:
if self.team.organization.is_ai_data_processing_approved is not True:
raise PermissionDenied("Enable AI data processing before asking another question.")
try:
created = start_sandbox_run(
team=self.team,
user_id=cast(User, request.user).id,
question=request.validated_data["question"],
)
except SandboxRunInProgress:
raise SandboxConflict()
if created.latest_run is None:
raise RuntimeError("Sandbox task was created without a run")
return Response(
SandboxRunStartedSerializer({"task_id": created.task_id, "run_id": created.latest_run.id}).data,
status=status.HTTP_201_CREATED,
)
@extend_schema(
responses={
200: OpenApiResponse(response=SandboxRunSerializer, description="Current sandbox run."),
404: OpenApiResponse(description="No sandbox run with this id for the current user."),
},
summary="Get a business knowledge sandbox run",
description="Poll a sandbox run started by the current user. A run that started before AI data processing was turned off can still be read.",
)
def retrieve(self, request: Request, pk: str | None = None, **kwargs: Any) -> Response:
payload = load_sandbox_run(
task_id=pk or "",
team_id=self.team.id,
user_id=cast(User, request.user).id,
)
if payload is None:
return Response(status=status.HTTP_404_NOT_FOUND)
return Response(SandboxRunSerializer(payload).data)