EMI-Backend/routes/liveCaptions.js

var express = require('express');
var router = express.Router();

const sessionChecker = require("../middleware/sessionChecker.js");

const MAX_BUFFER_SIZE = 300;
const DEFAULT_INITIAL_LIMIT = 40;
const MAX_INITIAL_LIMIT = 120;
const CAPTION_META_KEYS = new Set(["sequence", "createdAt", "original"]);

const liveCaptionState = {
    startedAt: Date.now(),
    latestSequence: 0,
    captions: [],
};

const normalizeLang = (lang = "") => {
    const value = String(lang || "").trim().toLowerCase();
    if (!value) return "";
    const base = value.split(",")[0].split("-")[0].trim();
    return base || value;
};

const normalizeTranslations = (translations) => {
    if (!translations || typeof translations !== "object" || Array.isArray(translations)) return {};
    const normalized = {};
    for (const [langKey, translatedText] of Object.entries(translations)) {
        const lang = normalizeLang(langKey);
        const text = typeof translatedText === "string" ? translatedText.trim() : "";
        if (!lang || !text) continue;
        normalized[lang] = text;
    }
    return normalized;
};

const buildTranslationsFromFlatPayload = (payload) => {
    const ignoredKeys = new Set(["original", "sourceLang", "translations"]);
    const normalized = {};
    for (const [key, value] of Object.entries(payload || {})) {
        if (ignoredKeys.has(key)) continue;
        const lang = normalizeLang(key);
        const text = typeof value === "string" ? value.trim() : "";
        if (!lang || !text) continue;
        normalized[lang] = text;
    }
    return normalized;
};

const inferSourceLangFromTranslations = (original, translations) => {
    const normalizedOriginal = String(original || "").trim();
    if (!normalizedOriginal) return "original";
    for (const [lang, text] of Object.entries(translations || {})) {
        if (String(text || "").trim() === normalizedOriginal) return normalizeLang(lang);
    }
    return "original";
};

const getAvailableLanguages = () => {
    const langs = new Set();
    for (const caption of liveCaptionState.captions) {
        Object.keys(caption || {}).forEach((key) => {
            if (CAPTION_META_KEYS.has(key)) return;
            const lang = normalizeLang(key);
            if (lang) langs.add(lang);
        });
    }
    return Array.from(langs).filter(Boolean).sort();
};

router.get("/stream", sessionChecker, async (req, res) => {
    try {
        const sinceSequence = Number.parseInt(req.query?.sinceSequence, 10);
        const requestedLimit = Number.parseInt(req.query?.limit, 10);
        const initialLimit = Number.isFinite(requestedLimit)
            ? Math.max(1, Math.min(requestedLimit, MAX_INITIAL_LIMIT))
            : DEFAULT_INITIAL_LIMIT;

        let captions = [];
        if (Number.isFinite(sinceSequence) && sinceSequence >= 0) {
            captions = liveCaptionState.captions.filter((item) => item.sequence > sinceSequence);
        } else {
            captions = liveCaptionState.captions.slice(-initialLimit);
        }

        return res.json({
            status: "ok",
            latestSequence: liveCaptionState.latestSequence,
            startedAt: new Date(liveCaptionState.startedAt).toISOString(),
            availableLanguages: getAvailableLanguages(),
            captions,
        });
    } catch (error) {
        console.error("Error getting live captions stream", error);
        return res.status(500).json({
            status: "Internal server error",
            latestSequence: liveCaptionState.latestSequence,
            captions: [],
            availableLanguages: [],
        });
    }
});

router.post("/ingest", async (req, res) => {
    try {
        // TODO: Add basic auth/API key validation before production roll-out.
        const original = typeof req.body?.original === "string" ? req.body.original.trim() : "";
        const mapFromNested = normalizeTranslations(req.body?.translations);
        const mapFromFlat = buildTranslationsFromFlatPayload(req.body);
        const translations = { ...mapFromNested, ...mapFromFlat };
        const sourceLang = normalizeLang(req.body?.sourceLang || inferSourceLangFromTranslations(original, translations));

        if (!original) {
            return res.status(400).json({ status: "Original text is required" });
        }
        if (sourceLang && sourceLang !== "original" && !translations[sourceLang]) {
            translations[sourceLang] = original;
        }

        const sequence = liveCaptionState.latestSequence + 1;
        const caption = {
            sequence,
            createdAt: new Date().toISOString(),
            original,
            ...translations,
        };

        liveCaptionState.latestSequence = sequence;
        liveCaptionState.captions.push(caption);
        if (liveCaptionState.captions.length > MAX_BUFFER_SIZE) {
            liveCaptionState.captions.splice(0, liveCaptionState.captions.length - MAX_BUFFER_SIZE);
        }

        return res.json({
            status: "ok",
            caption,
            latestSequence: liveCaptionState.latestSequence,
            availableLanguages: getAvailableLanguages(),
        });
    } catch (error) {
        console.error("Error ingesting live captions", error);
        return res.status(500).json({ status: "Internal server error" });
    }
});

router.post("/reset", async (_, res) => {
    try {
        // TODO: Add admin authorization before exposing this endpoint.
        liveCaptionState.startedAt = Date.now();
        liveCaptionState.latestSequence = 0;
        liveCaptionState.captions = [];
        return res.json({ status: "ok" });
    } catch (error) {
        console.error("Error resetting live captions state", error);
        return res.status(500).json({ status: "Internal server error" });
    }
});

module.exports = router;