: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 __future__ import annotations
from typing import Any
from uuid import UUID
from drf_spectacular.utils import extend_schema
from rest_framework import serializers, viewsets
from rest_framework.request import Request
from rest_framework.response import Response
from posthog.api.routing import TeamAndOrgViewSetMixin
from posthog.paging import PageRequest
from posthog.permissions import TeamMemberLightManagementPermission
from products.access_control.backend.facade.user_access_control import UserAccessControl
from products.access_control.backend.presentation.access_control import AccessControlViewSetMixin
from products.conversations.backend.ai.ticket_context import ACCOUNT_TARGET_TYPE, MAX_AI_CONTEXT_ACCOUNT_PROPERTY_IDS
from products.customer_analytics.backend.facade.api import (
get_custom_property_definition_target_type,
list_custom_property_definitions,
)
# One bounded read, not a paging loop: each facade call repeats the count, the workflow
# reference scan and the source enrichment, so paging a large team costs several full scans
# for a picker that only ever shows a short list.
MAX_ACCOUNT_PROPERTY_OPTIONS = 500
class AIContextAccountPropertySerializer(serializers.Serializer):
id = serializers.UUIDField(help_text="Customer analytics account property definition id.")
name = serializers.CharField(help_text="Display name of the account property.")
def list_account_property_options(team_id: int, user_access_control: UserAccessControl) -> list[dict[str, object]]:
page = list_custom_property_definitions(
team_id,
page=PageRequest(limit=MAX_ACCOUNT_PROPERTY_OPTIONS),
user_access_control=user_access_control,
exclude_group_targets=True,
target_type=ACCOUNT_TARGET_TYPE,
)
return [{"id": item.id, "name": item.name} for item in page.results if item.id is not None and item.name]
class AIContextAccountPropertiesViewSet(TeamAndOrgViewSetMixin, AccessControlViewSetMixin, viewsets.GenericViewSet):
# Same authorization as the Customer analytics definition list this reads from: the names
# and ids are account data, so project membership alone is not enough.
scope_object = "account"
permission_classes = [TeamMemberLightManagementPermission]
pagination_class = None
serializer_class = AIContextAccountPropertySerializer
@extend_schema(
responses=AIContextAccountPropertySerializer(many=True),
description=(
"Account-target Customer analytics properties that can be included in AI reply context. "
f"Capped at the first {MAX_ACCOUNT_PROPERTY_OPTIONS} properties by name."
),
)
def list(self, request: Request, *args: Any, **kwargs: Any) -> Response:
serializer = self.get_serializer(
list_account_property_options(self.team.id, self.user_access_control),
many=True,
)
return Response(serializer.data)
def validate_ai_context_conversations_settings(value: dict, *, team_id: int | None) -> dict:
"""Normalize ai_context_account_property_ids. Raises DRF ValidationError.
``team_id`` is ``None`` while a team is being created. The shape checks still run, but a team
that does not exist yet owns no property definition, so every id is dropped."""
if "ai_context_account_property_ids" not in value:
return value
raw = value.get("ai_context_account_property_ids")
if raw is None:
value["ai_context_account_property_ids"] = []
return value
if not isinstance(raw, list):
raise serializers.ValidationError({"ai_context_account_property_ids": "Must be a list of property ids."})
ids: list[str] = []
seen: set[str] = set()
for item in raw:
if not isinstance(item, str):
raise serializers.ValidationError(
{"ai_context_account_property_ids": "Each property id must be a UUID string."}
)
try:
parsed = str(UUID(item))
except ValueError as exc:
raise serializers.ValidationError(
{"ai_context_account_property_ids": "Each property id must be a UUID string."}
) from exc
if parsed in seen:
continue
seen.add(parsed)
ids.append(parsed)
if len(ids) >= MAX_AI_CONTEXT_ACCOUNT_PROPERTY_IDS:
break
if not ids or team_id is None:
value["ai_context_account_property_ids"] = []
return value
try:
kept = [
item for item in ids if get_custom_property_definition_target_type(team_id, item) == ACCOUNT_TARGET_TYPE
]
except Exception as exc:
raise serializers.ValidationError(
{"ai_context_account_property_ids": "Couldn't validate account properties."}
) from exc
value["ai_context_account_property_ids"] = kept
return value