: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
"""
Django REST API endpoint for translating LLM trace message content.
Endpoint:
- POST /api/environments/:id/llm_analytics/translate/ - Translate text
"""
import time
from typing import cast
import structlog
from drf_spectacular.types import OpenApiTypes
from drf_spectacular.utils import extend_schema
from rest_framework import exceptions, serializers, status, viewsets
from rest_framework.request import Request
from rest_framework.response import Response
from posthog.api.monitoring import monitor
from posthog.api.routing import TeamAndOrgViewSetMixin
from posthog.event_usage import report_user_action
from posthog.models import User
from posthog.permissions import AccessControlPermission
from posthog.rate_limit import (
AIObservabilityTranslationBurstThrottle,
AIObservabilityTranslationDailyThrottle,
AIObservabilityTranslationSustainedThrottle,
)
from products.ai_observability.backend.api.metrics import llma_track_latency
from products.ai_observability.backend.translation.constants import DEFAULT_TARGET_LANGUAGE
from products.ai_observability.backend.translation.llm import TranslationTimeoutError, translate_text
logger = structlog.get_logger(__name__)
class TranslationTimeout(exceptions.APIException):
status_code = status.HTTP_504_GATEWAY_TIMEOUT
default_detail = (
"Translation took too long. Try again, and if it keeps failing this message may be too long to translate."
)
default_code = "translation_timeout"
class TranslateRequestSerializer(serializers.Serializer):
text = serializers.CharField(max_length=10000, help_text="The text to translate")
target_language = serializers.CharField(
max_length=10,
default=DEFAULT_TARGET_LANGUAGE,
required=False,
help_text="Target language code (default: 'en' for English)",
)
class TranslateResponseSerializer(serializers.Serializer):
translation = serializers.CharField(help_text="The translated text")
detected_language = serializers.CharField(
required=False, allow_null=True, help_text="Detected source language (if available)"
)
provider = serializers.CharField(help_text="Translation provider used")
class AIObservabilityTranslateViewSet(TeamAndOrgViewSetMixin, viewsets.GenericViewSet):
"""ViewSet for translating LLM trace message content."""
scope_object = "llm_analytics"
permission_classes = [AccessControlPermission]
def get_throttles(self):
return [
AIObservabilityTranslationBurstThrottle(),
AIObservabilityTranslationSustainedThrottle(),
AIObservabilityTranslationDailyThrottle(),
]
def _validate_feature_access(self, request: Request) -> None:
"""Validate that the user is authenticated and AI data processing is approved."""
if not request.user.is_authenticated:
raise exceptions.NotAuthenticated()
if not self.organization.is_ai_data_processing_approved:
raise exceptions.PermissionDenied(
"AI data processing must be approved by your organization before using translation"
)
@extend_schema(request=TranslateRequestSerializer, responses={200: OpenApiTypes.OBJECT})
@llma_track_latency("llma_translate")
@monitor(feature=None, endpoint="llma_translate", method="POST")
def create(self, request: Request, *args, **kwargs) -> Response:
"""Translate text to target language."""
self._validate_feature_access(request)
serializer = TranslateRequestSerializer(data=request.data)
serializer.is_valid(raise_exception=True)
text = serializer.validated_data["text"]
target_language = serializer.validated_data.get("target_language", DEFAULT_TARGET_LANGUAGE)
try:
logger.info(
"translation_requested",
target_language=target_language,
text_length=len(text),
)
start_time = time.time()
user = cast(User, request.user)
translation = translate_text(text, target_language, user_distinct_id=user.distinct_id)
duration_seconds = time.time() - start_time
logger.info(
"translation_completed",
target_language=target_language,
translation_length=len(translation),
)
report_user_action(
user,
"llma translation generated",
{
"target_language": target_language,
"text_length": len(text),
"translation_length": len(translation),
"duration_seconds": duration_seconds,
},
team=self.team,
request=self.request,
)
return Response(
{
"translation": translation,
"detected_language": None,
"provider": "openai",
},
status=status.HTTP_200_OK,
)
except TranslationTimeoutError:
# Not logger.exception: the traceback is all HTTP client plumbing, and the input length
# is what actually explains a timeout.
logger.warning(
"translation_timed_out",
target_language=target_language,
text_length=len(text),
)
raise TranslationTimeout()
except Exception as e:
logger.exception("translation_failed", error=str(e), target_language=target_language)
raise exceptions.APIException(
detail="Translation failed. Please try again.",
code="translation_error",
)