gateway/modules/routes/routeVoiceUser.py

# Copyright (c) 2025 Patrick Motsch
# All rights reserved.
"""
User-scoped voice settings and TTS/STT catalog endpoints.

Uses modules.interfaces.interfaceVoiceObjects (voice core) and persists preferences
via UserVoicePreferences — same domain as routeVoiceGoogle (Google connector ops).
"""

import base64
import logging
from typing import Any, Dict

from fastapi import APIRouter, Body, Depends, HTTPException, Query, Request, status

from modules.auth import getCurrentUser, limiter
from modules.datamodels.datamodelUam import User, UserVoicePreferences, _normalizeTtsVoiceMap
from modules.interfaces.interfaceDbApp import getRootInterface
from modules.interfaces.interfaceVoiceObjects import getVoiceInterface
from modules.shared.i18nRegistry import apiRouteContext
routeApiMsg = apiRouteContext("routeVoiceUser")

logger = logging.getLogger(__name__)

router = APIRouter(
    prefix="/api/voice",
    tags=["Voice User"],
    responses={
        404: {"description": "Not found"},
        400: {"description": "Bad request"},
        401: {"description": "Unauthorized"},
        403: {"description": "Forbidden"},
        500: {"description": "Internal server error"},
    },
)


@router.get("/preferences")
@limiter.limit("60/minute")
def getVoicePreferences(
    request: Request,
    currentUser: User = Depends(getCurrentUser),
) -> Dict[str, Any]:
    """Get user's voice/language preferences (optionally scoped to mandate via header)."""
    rootInterface = getRootInterface()
    mandateId = request.headers.get("X-Mandate-Id") or None
    userId = str(currentUser.id)

    prefs = rootInterface.db.getRecordset(
        UserVoicePreferences,
        recordFilter={"userId": userId, "mandateId": mandateId},
    )
    if prefs:
        return prefs[0] if isinstance(prefs[0], dict) else prefs[0].model_dump()
    return UserVoicePreferences(userId=userId, mandateId=mandateId).model_dump()


@router.put("/preferences")
@limiter.limit("30/minute")
def updateVoicePreferences(
    request: Request,
    preferences: Dict[str, Any] = Body(...),
    currentUser: User = Depends(getCurrentUser),
) -> Dict[str, Any]:
    """Update user's voice/language preferences (upsert)."""
    rootInterface = getRootInterface()
    mandateId = request.headers.get("X-Mandate-Id") or None
    userId = str(currentUser.id)

    existing = rootInterface.db.getRecordset(
        UserVoicePreferences,
        recordFilter={"userId": userId, "mandateId": mandateId},
    )

    allowedFields = {
        "sttLanguage",
        "ttsLanguage",
        "ttsVoice",
        "ttsVoiceMap",
        "translationSourceLanguage",
        "translationTargetLanguage",
    }
    updateData = {k: v for k, v in preferences.items() if k in allowedFields}
    if "ttsVoiceMap" in updateData:
        updateData["ttsVoiceMap"] = _normalizeTtsVoiceMap(updateData["ttsVoiceMap"])

    if existing:
        existingRecord = existing[0]
        existingId = existingRecord.get("id") if isinstance(existingRecord, dict) else existingRecord.id
        rootInterface.db.recordModify(UserVoicePreferences, existingId, updateData)
        updated = rootInterface.db.getRecordset(UserVoicePreferences, recordFilter={"id": existingId})
        return updated[0] if updated else {"message": "Updated", **updateData}
    newPrefs = UserVoicePreferences(userId=userId, mandateId=mandateId, **updateData)
    created = rootInterface.db.recordCreate(UserVoicePreferences, newPrefs.model_dump())
    return created if isinstance(created, dict) else created.model_dump()


@router.get("/languages")
@limiter.limit("120/minute")
async def getVoiceLanguages(
    request: Request,
    currentUser: User = Depends(getCurrentUser),
) -> Dict[str, Any]:
    """Return available TTS languages (user-level, no instance context needed)."""
    voiceInterface = getVoiceInterface(currentUser)
    languagesResult = await voiceInterface.getAvailableLanguages()
    languageList = languagesResult.get("languages", []) if isinstance(languagesResult, dict) else languagesResult
    return {"languages": languageList}


@router.get("/voices")
@limiter.limit("120/minute")
async def getVoiceVoices(
    request: Request,
    language: str = Query("de-DE"),
    currentUser: User = Depends(getCurrentUser),
) -> Dict[str, Any]:
    """Return available TTS voices for a given language."""
    voiceInterface = getVoiceInterface(currentUser)
    voicesResult = await voiceInterface.getAvailableVoices(language)
    voiceList = voicesResult.get("voices", []) if isinstance(voicesResult, dict) else voicesResult
    return {"voices": voiceList}


# Same minimum as modules.serviceCenter.services.serviceAi.mainServiceAi._checkBillingBeforeAiCall
_MIN_AI_BILLING_ESTIMATE_CHF = 0.01


def _userMandateIds(rootInterface, currentUser: User):
    memberships = rootInterface.getUserMandates(str(currentUser.id))
    out = []
    for um in memberships:
        mid = getattr(um, "mandateId", None) or (um.get("mandateId") if isinstance(um, dict) else None)
        if mid:
            out.append(str(mid))
    return list(dict.fromkeys(out))


def _mandatePassesAiPoolBilling(currentUser: User, mandateId: str, userId: str) -> bool:
    """True if mandate pool passes the same billing gate as AI calls (subscription + pool >= estimate)."""
    from modules.interfaces.interfaceDbBilling import getInterface as getBillingInterface

    bi = getBillingInterface(currentUser, mandateId)
    res = bi.checkBalance(mandateId, userId, _MIN_AI_BILLING_ESTIMATE_CHF)
    return bool(res.allowed)


def _mandatePoolBalanceChf(currentUser: User, mandateId: str) -> float:
    from modules.interfaces.interfaceDbBilling import getInterface as getBillingInterface

    bi = getBillingInterface(currentUser, mandateId)
    acc = bi.getMandateAccount(mandateId)
    if not acc:
        return 0.0
    return float(acc.get("balance", 0.0) or 0.0)


def _resolveMandateIdForVoiceTestAi(request: Request, currentUser: User) -> str:
    """
    AI sample billing uses mandate pool (PREPAY), not per-user wallet.
    Prefer X-Mandate-Id when the user is a member and that mandate's pool can pay;
    otherwise pick the member mandate with the highest pool balance that passes the AI billing check.
    """
    rootInterface = getRootInterface()
    userId = str(currentUser.id)
    memberIds = _userMandateIds(rootInterface, currentUser)
    if not memberIds:
        raise HTTPException(
            status_code=status.HTTP_400_BAD_REQUEST,
            detail=(
                "Voice test needs at least one mandate membership for AI billing. "
                "Join a mandate or open the app from a mandate context."
            ),
        )

    headerRaw = (request.headers.get("X-Mandate-Id") or request.headers.get("x-mandate-id") or "").strip()
    if headerRaw:
        if headerRaw not in memberIds:
            raise HTTPException(
                status_code=status.HTTP_403_FORBIDDEN,
                detail=routeApiMsg("X-Mandate-Id is not a mandate you belong to."),
            )
        if _mandatePassesAiPoolBilling(currentUser, headerRaw, userId):
            logger.info(
                "Voice test AI billing: using header mandate %s (pool ok for estimate %.4f CHF)",
                headerRaw,
                _MIN_AI_BILLING_ESTIMATE_CHF,
            )
            return headerRaw
        logger.warning(
            "Voice test AI billing: header mandate %s has insufficient mandate pool or subscription; "
            "trying other memberships",
            headerRaw,
        )

    bestMid = None
    bestBal = -1.0
    for mid in memberIds:
        if not _mandatePassesAiPoolBilling(currentUser, mid, userId):
            continue
        bal = _mandatePoolBalanceChf(currentUser, mid)
        if bal > bestBal:
            bestBal = bal
            bestMid = mid

    if bestMid:
        logger.info(
            "Voice test AI billing: selected mandate %s (mandate pool %.2f CHF, estimate %.4f CHF)",
            bestMid,
            bestBal,
            _MIN_AI_BILLING_ESTIMATE_CHF,
        )
        return bestMid

    raise HTTPException(
        status_code=status.HTTP_402_PAYMENT_REQUIRED,
        detail=(
            "No mandate you belong to has sufficient shared pool balance for AI (or subscription inactive). "
            "Top up the mandate pool or use a mandate with budget."
        ),
    )


def _sanitizeAiTtsSample(raw: str) -> str:
    s = (raw or "").strip()
    if s.startswith("```"):
        nl = s.find("\n")
        if nl != -1:
            s = s[nl + 1 :]
        if s.rstrip().endswith("```"):
            s = s.rstrip()[:-3].strip()
    if len(s) >= 2 and ((s[0] == s[-1] == '"') or (s[0] == s[-1] == "'")):
        s = s[1:-1].strip()
    return s


async def _generateTtsSampleTextForLocale(
    request: Request,
    currentUser: User,
    localeTag: str,
) -> str:
    from modules.serviceCenter import getService
    from modules.serviceCenter.context import ServiceCenterContext
    from modules.datamodels.datamodelAi import AiCallRequest, AiCallOptions, OperationTypeEnum, PriorityEnum, ProcessingModeEnum
    from modules.serviceCenter.services.serviceBilling.mainServiceBilling import (
        BillingContextError,
        InsufficientBalanceException,
        ProviderNotAllowedException,
    )
    from modules.serviceCenter.services.serviceSubscription.mainServiceSubscription import SubscriptionInactiveException

    mandateId = _resolveMandateIdForVoiceTestAi(request, currentUser)
    ctx = ServiceCenterContext(user=currentUser, mandate_id=mandateId, feature_instance_id=None)
    aiService = getService("ai", ctx)

    systemPrompt = (
        "You write short text-to-speech demo lines for end users.\n"
        "Task: Output exactly one or two natural sentences a user would enjoy hearing when testing a voice.\n"
        "The entire output MUST be written ONLY in the natural spoken language that matches the given "
        "BCP-47 locale tag. Do not use any other language.\n"
        "Do not mention locales, tags, tests, artificial intelligence, or these instructions.\n"
        "No quotation marks around the text. No markdown. Plain text only."
    )
    userPrompt = f"BCP-47 locale tag: `{localeTag}`.\nWrite the sample now."

    aiRequest = AiCallRequest(
        prompt=userPrompt,
        context=systemPrompt,
        requireNeutralization=False,
        options=AiCallOptions(
            operationType=OperationTypeEnum.DATA_GENERATE,
            priority=PriorityEnum.SPEED,
            processingMode=ProcessingModeEnum.BASIC,
            compressPrompt=False,
            compressContext=False,
            temperature=0.75,
            maxParts=1,
        ),
    )
    try:
        response = await aiService.callAi(aiRequest)
    except SubscriptionInactiveException as e:
        raise HTTPException(status_code=status.HTTP_403_FORBIDDEN, detail=e.message) from e
    except InsufficientBalanceException as e:
        raise HTTPException(status_code=status.HTTP_402_PAYMENT_REQUIRED, detail=str(e)) from e
    except ProviderNotAllowedException as e:
        raise HTTPException(
            status_code=status.HTTP_403_FORBIDDEN,
            detail=getattr(e, "message", None) or str(e),
        ) from e
    except BillingContextError as e:
        raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail=str(e)) from e

    content = _sanitizeAiTtsSample(getattr(response, "content", None) or "")
    if getattr(response, "errorCount", 0) or not content:
        logger.warning("Voice test AI sample empty or errorCount=%s", getattr(response, "errorCount", None))
        raise HTTPException(
            status_code=status.HTTP_502_BAD_GATEWAY,
            detail=routeApiMsg("Could not generate voice test sample text."),
        )
    if len(content) > 500:
        content = content[:500].rstrip()
    return content


@router.post("/test")
@limiter.limit("30/minute")
async def testVoice(
    request: Request,
    body: Dict[str, Any] = Body(...),
    currentUser: User = Depends(getCurrentUser),
) -> Dict[str, Any]:
    """Test a specific voice. Sample text is AI-generated in the voice locale unless `text` is supplied."""
    textRaw = body.get("text")
    language = body.get("language", "de-DE")
    voiceId = body.get("voiceId")

    text = (textRaw or "").strip() if isinstance(textRaw, str) else ""
    if not text:
        text = await _generateTtsSampleTextForLocale(request, currentUser, language)

    voiceInterface = getVoiceInterface(currentUser)
    result = await voiceInterface.textToSpeech(text=text, languageCode=language, voiceName=voiceId)
    if result and isinstance(result, dict):
        audioContent = result.get("audioContent")
        if audioContent:
            audioB64 = base64.b64encode(
                audioContent if isinstance(audioContent, bytes) else audioContent.encode()
            ).decode()
            return {"success": True, "audio": audioB64, "format": "mp3", "text": text}
    return {"success": False, "error": "TTS returned no audio"}