import { inngest } from '../client';
import { extractFramesAtIntervals, VideoUnavailableError } from '@/features/discovery/services/frame-extraction.service';
import { detectObjectsInFrame, VisionRateLimitError } from '@/features/discovery/services/ai-vision.service';
import { db } from '@/lib/db';
import { detectedObjects, youtubeVideos } from '@/lib/db/schema';
import * as Sentry from '@sentry/nextjs';
import { NonRetriableError, RetryAfterError } from 'inngest';
import { eq } from 'drizzle-orm';
import sharp from 'sharp';
import { uploadThumbnail } from '@/features/discovery/services/storage.service';

export const detectObjects = inngest.createFunction(
    { id: 'detect-objects', retries: 3 },
    { event: 'youtube/video.detect-objects' },
    async ({ event, step }) => {
        const { videoId, videoUrl } = event.data;

        if (!videoId || !videoUrl) {
            throw new NonRetriableError('Missing videoId or videoUrl');
        }

        try {
            // New: Mark as in_progress
            await step.run('update-scan-status-started', async () => {
                await db.update(youtubeVideos)
                    .set({ scanStatus: 'in_progress' })
                    .where(eq(youtubeVideos.id, videoId));
            });

            // Step 1: Process Video (Extract & Detect)
            const detections = await step.run('analyze-video-content', async () => {
                try {
                    // Calculate adaptive frame interval based on video duration
                    // Short videos: more frequent sampling for better coverage
                    // Long videos: less frequent to manage API costs
                    const calculateFrameInterval = (durationSeconds: number | null | undefined): number => {
                        const duration = durationSeconds || 300; // Default to 5 minutes if unknown
                        if (duration <= 120) return 5;      // 0-2 min: every 5 seconds
                        if (duration <= 300) return 10;     // 2-5 min: every 10 seconds
                        if (duration <= 600) return 15;     // 5-10 min: every 15 seconds
                        return 30;                           // 10+ min: every 30 seconds
                    };

                    // Get video metadata to determine duration
                    const videoRecord = await db.query.youtubeVideos.findFirst({ where: eq(youtubeVideos.id, videoId) });
                    const durationSeconds = videoRecord?.duration ? parseInt(videoRecord.duration, 10) : 300; // Default to 5 minutes if unknown
                    const frameInterval = calculateFrameInterval(durationSeconds);

                    console.log(`Video duration: ${durationSeconds}s, using frame interval: ${frameInterval}s`);

                    // Extract frames at adaptive intervals
                    const frames = await extractFramesAtIntervals(videoUrl, frameInterval);

                    const allResults: any[] = [];

                    // Process frames in parallel batches of 3 for better performance
                    const BATCH_SIZE = 3;
                    const HEARTBEAT_INTERVAL = 5; // Update heartbeat every 5 frames

                    for (let i = 0; i < frames.length; i += BATCH_SIZE) {
                        const batch = frames.slice(i, i + BATCH_SIZE);

                        // Process batch frames in parallel
                        const batchResults = await Promise.allSettled(
                            batch.map(async (frame) => {
                                try {
                                    const results = await detectObjectsInFrame(frame.frameBuffer);

                                    // Process detections to add thumbnails
                                    const processedResults = await Promise.all(
                                        results.map(async (result) => {
                                            let thumbnailUrl = null;

                                            if (result.boundingBox) {
                                                try {
                                                    // Get base image dimensions to convert normalized coordinates if needed
                                                    const meta = await sharp(frame.frameBuffer).metadata();
                                                    const imgW = meta.width || 1280;
                                                    const imgH = meta.height || 720;

                                                    let { x, y, width, height } = result.boundingBox;

                                                    // Detect coordinate format and normalize to pixel coordinates
                                                    const maxCoord = Math.max(x, y, width, height);

                                                    if (maxCoord <= 1) {
                                                        // Normalized 0-1 coordinates
                                                        x = x * imgW;
                                                        y = y * imgH;
                                                        width = width * imgW;
                                                        height = height * imgH;
                                                    } else if (maxCoord <= 1000 && maxCoord > imgW && maxCoord > imgH) {
                                                        // Gemini 0-1000 scale
                                                        x = (x / 1000) * imgW;
                                                        y = (y / 1000) * imgH;
                                                        width = (width / 1000) * imgW;
                                                        height = (height / 1000) * imgH;
                                                    }
                                                    // else: already in pixel coordinates (HuggingFace format)

                                                    // Add 20% padding around the object for better context
                                                    const paddingX = width * 0.2;
                                                    const paddingY = height * 0.2;

                                                    const paddedX = x - paddingX;
                                                    const paddedY = y - paddingY;
                                                    const paddedWidth = width + (2 * paddingX);
                                                    const paddedHeight = height + (2 * paddingY);

                                                    // Clamp to image bounds
                                                    const left = Math.max(0, Math.floor(paddedX));
                                                    const top = Math.max(0, Math.floor(paddedY));
                                                    const right = Math.min(imgW, Math.ceil(paddedX + paddedWidth));
                                                    const bottom = Math.min(imgH, Math.ceil(paddedY + paddedHeight));

                                                    const finalWidth = right - left;
                                                    const finalHeight = bottom - top;

                                                    // Validate dimensions (minimum 20x20 pixels)
                                                    if (finalWidth >= 20 && finalHeight >= 20) {
                                                        console.log(`Cropping ${result.name}: [${left},${top}] ${finalWidth}x${finalHeight} from ${imgW}x${imgH}`);

                                                        const thumbnailBuffer = await sharp(frame.frameBuffer)
                                                            .extract({
                                                                left,
                                                                top,
                                                                width: finalWidth,
                                                                height: finalHeight
                                                            })
                                                            .resize(300, 300, {
                                                                fit: 'contain',
                                                                background: { r: 255, g: 255, b: 255, alpha: 1 }
                                                            })
                                                            .toFormat('jpeg', { quality: 85 })
                                                            .toBuffer();

                                                        const fileName = `${videoId}/${frame.timestamp}_${result.name.replace(/[^a-z0-9]/gi, '_').toLowerCase()}_${Date.now()}.jpg`;
                                                        thumbnailUrl = await uploadThumbnail(thumbnailBuffer, fileName);
                                                        console.log(`✓ Thumbnail saved: ${fileName}`);
                                                    } else {
                                                        throw new Error(`Region too small (${finalWidth}x${finalHeight})`);
                                                    }
                                                } catch (cropError) {
                                                    console.warn(`⚠ Thumbnail generation issue for ${result.name} (falling back to full frame):`, cropError);

                                                    try {
                                                        const fallbackBuffer = await sharp(frame.frameBuffer)
                                                            .resize(300, 300, {
                                                                fit: 'contain',
                                                                background: { r: 255, g: 255, b: 255, alpha: 1 }
                                                            })
                                                            .toFormat('jpeg', { quality: 85 })
                                                            .toBuffer();

                                                        const fallbackName = `${videoId}/${frame.timestamp}_${result.name.replace(/[^a-z0-9]/gi, '_').toLowerCase()}_full_${Date.now()}.jpg`;
                                                        thumbnailUrl = await uploadThumbnail(fallbackBuffer, fallbackName);
                                                        console.log(`✓ Fallback thumbnail saved: ${fallbackName}`);
                                                    } catch (fallbackError) {
                                                        console.error(`✗ Fallback thumbnail generation failed for ${result.name}:`, fallbackError);
                                                    }
                                                }
                                            }

                                            return {
                                                ...result,
                                                timestamp: frame.timestamp,
                                                thumbnailUrl
                                            };
                                        })
                                    );

                                    return processedResults;
                                } catch (err) {
                                    // Handle rate limiting
                                    if (err instanceof VisionRateLimitError) {
                                        throw new RetryAfterError(
                                            `Vision API rate limited`,
                                            `${err.retryAfter || 60}s`
                                        );
                                    }
                                    // Log but continue processing other frames
                                    Sentry.captureException(err, {
                                        tags: {
                                            source: 'vision-detection',
                                            videoId,
                                            frameTimestamp: frame.timestamp,
                                        },
                                    });
                                    return []; // Return empty array for failed frames
                                }
                            })
                        );

                        // Collect successful results from the batch
                        for (const result of batchResults) {
                            if (result.status === 'fulfilled') {
                                allResults.push(...result.value);
                            }
                        }

                        // Periodic heartbeat update (every HEARTBEAT_INTERVAL frames, not per frame)
                        if (i % HEARTBEAT_INTERVAL === 0) {
                            await db.update(youtubeVideos)
                                .set({ updatedAt: new Date() })
                                .where(eq(youtubeVideos.id, videoId));
                        }
                    }

                    return allResults;
                } catch (err) {
                    // Propagate rate limit errors
                    if (err instanceof RetryAfterError) throw err;
                    throw err;
                }
            });

            // Step 2: Save Results (filter out Person category - only save products)
            const savedObjectIds = await step.run('save-detections', async () => {
                // Filter out Person detections - only interested in products
                const productDetections = detections.filter((d: any) => d.category !== 'Person');

                if (productDetections.length === 0) return [];

                const records = productDetections.map((d: any) => ({
                    videoId,
                    objectName: d.name,
                    category: d.category,
                    confidenceScore: d.confidence,
                    frameTimestamp: d.timestamp,
                    detectionMetadata: d.boundingBox ? { boundingBox: d.boundingBox } : null,
                    thumbnailUrl: d.thumbnailUrl,
                    status: (d.confidence < 0.7 ? 'flagged' : 'pending_review') as 'flagged' | 'pending_review'
                }));

                const results = await db.insert(detectedObjects).values(records).returning({ id: detectedObjects.id });
                return results.map(r => r.id);
            });

            // Step 3: Trigger Marketplace Match (single batch event for all objects)
            if (savedObjectIds.length > 0) {
                await step.run('trigger-marketplace-match', async () => {
                    await inngest.send({
                        name: 'discovery/objects.match-marketplace',
                        data: {
                            detectedObjectIds: savedObjectIds,
                            videoId
                        }
                    });
                });
            }

            await step.run('update-scan-status-finished', async () => {
                await db.update(youtubeVideos)
                    .set({ scanStatus: 'awaiting_approval' })
                    .where(eq(youtubeVideos.id, videoId));
            });

            return { success: true, count: detections.length };

        } catch (error) {
            // New: Mark as failed
            await step.run('update-scan-status-failed', async () => {
                await db.update(youtubeVideos)
                    .set({ scanStatus: 'failed' })
                    .where(eq(youtubeVideos.id, videoId));
            });

            if (error instanceof VideoUnavailableError) {
                console.warn(`Skipping video ${videoId}: ${error.message}`);
                await db.update(youtubeVideos)
                    .set({ availabilityStatus: 'private' })
                    .where(eq(youtubeVideos.id, videoId));
                throw new NonRetriableError(error.message);
            }
            Sentry.captureException(error, {
                tags: {
                    source: 'inngest',
                    function: 'detect-objects',
                    videoId
                }
            });
            throw error;
        }
    }
);