vibn-frontend/app/api/projects/[projectId]/knowledge/batch-extract/route.ts

import { NextResponse } from 'next/server';
import { getAdminDb } from '@/lib/firebase/admin';
import { runChatExtraction } from '@/lib/ai/chat-extractor';
import { GeminiLlmClient } from '@/lib/ai/gemini-client';
import { createChatExtraction } from '@/lib/server/chat-extraction';
import { FieldValue } from 'firebase-admin/firestore';
import type { ProjectPhaseScores } from '@/lib/types/project-artifacts';
import type { KnowledgeItem } from '@/lib/types/knowledge';

export const maxDuration = 300; // 5 minutes for batch processing

interface BatchExtractionResult {
  knowledgeItemId: string;
  success: boolean;
  error?: string;
}

export async function POST(
  request: Request,
  context: { params?: Promise<{ projectId?: string }> | { projectId?: string } } = {},
) {
  try {
    // Await params if it's a Promise (Next.js 15+)
    const params = context.params instanceof Promise ? await context.params : context.params;
    
    const url = new URL(request.url);
    const pathSegments = url.pathname.split('/');
    const projectsIndex = pathSegments.indexOf('projects');
    const projectIdFromPath =
      projectsIndex !== -1 ? pathSegments[projectsIndex + 1] : undefined;

    const projectId =
      (params?.projectId ?? projectIdFromPath ?? url.searchParams.get('projectId') ?? '').trim();
    
    if (!projectId) {
      return NextResponse.json({ error: 'Missing projectId' }, { status: 400 });
    }

    const adminDb = getAdminDb();
    
    // Get all knowledge_items for this project
    const knowledgeSnapshot = await adminDb
      .collection('projects')
      .doc(projectId)
      .collection('knowledge_items')
      .get();

    if (knowledgeSnapshot.empty) {
      return NextResponse.json({ 
        message: 'No knowledge items to extract',
        results: []
      });
    }

    const knowledgeItems = knowledgeSnapshot.docs.map(doc => ({
      id: doc.id,
      ...doc.data()
    })) as KnowledgeItem[];

    // Get existing extractions to avoid re-processing
    const extractionsSnapshot = await adminDb
      .collection('projects')
      .doc(projectId)
      .collection('chat_extractions')
      .get();

    const processedKnowledgeIds = new Set(
      extractionsSnapshot.docs.map(doc => doc.data().knowledgeItemId)
    );

    // Filter to only unprocessed items
    const itemsToProcess = knowledgeItems.filter(
      item => !processedKnowledgeIds.has(item.id)
    );

    if (itemsToProcess.length === 0) {
      return NextResponse.json({
        message: 'All knowledge items already extracted',
        results: []
      });
    }

    console.log(`[batch-extract] Processing ${itemsToProcess.length} knowledge items for project ${projectId}`);

    const llm = new GeminiLlmClient();
    const results: BatchExtractionResult[] = [];
    let successCount = 0;
    let lastSuccessfulExtraction = null;

    // Process each item
    for (const knowledgeItem of itemsToProcess) {
      try {
        console.log(`[batch-extract] Extracting from knowledgeItemId=${knowledgeItem.id}`);
        
        const extractionData = await runChatExtraction(knowledgeItem, llm);
        const overallCompletion = extractionData.summary_scores.overall_completion ?? 0;
        const overallConfidence = extractionData.summary_scores.overall_confidence ?? 0;

        const extraction = await createChatExtraction({
          projectId,
          knowledgeItemId: knowledgeItem.id,
          data: extractionData,
          overallCompletion,
          overallConfidence,
        });

        lastSuccessfulExtraction = extraction;
        successCount++;

        results.push({
          knowledgeItemId: knowledgeItem.id,
          success: true
        });

        console.log(`[batch-extract] Successfully extracted from knowledgeItemId=${knowledgeItem.id}`);

        // Also chunk and embed this item (fire-and-forget)
        (async () => {
          try {
            const { writeKnowledgeChunksForItem } = await import('@/lib/server/vector-memory');
            await writeKnowledgeChunksForItem({
              id: knowledgeItem.id,
              projectId: knowledgeItem.projectId,
              content: knowledgeItem.content,
              sourceMeta: knowledgeItem.sourceMeta,
            });
          } catch (chunkError) {
            console.error(`[batch-extract] Failed to chunk item ${knowledgeItem.id}:`, chunkError);
          }
        })();
      } catch (error) {
        console.error(`[batch-extract] Failed to extract from knowledgeItemId=${knowledgeItem.id}:`, error);
        results.push({
          knowledgeItemId: knowledgeItem.id,
          success: false,
          error: error instanceof Error ? error.message : String(error)
        });
      }
    }

    // Update project phase if we had any successful extractions
    if (successCount > 0 && lastSuccessfulExtraction) {
      const projectRef = adminDb.collection('projects').doc(projectId);
      const snapshot = await projectRef.get();
      const docData = snapshot.data() ?? {};
      const existingScores = (docData.phaseScores ?? {}) as ProjectPhaseScores;
      const phaseHistory = Array.isArray(docData.phaseHistory) ? [...docData.phaseHistory] : [];
      
      phaseHistory.push({
        phase: 'extractor',
        status: 'completed',
        knowledgeItemId: 'batch_extraction',
        timestamp: new Date().toISOString(),
      });

      // Use the last extraction's scores as representative
      const lastData = lastSuccessfulExtraction.data as { summary_scores?: { overall_completion?: number; overall_confidence?: number } };
      existingScores.extractor = {
        knowledgeItemId: 'batch_extraction',
        overallCompletion: lastData.summary_scores?.overall_completion ?? 0,
        overallConfidence: lastData.summary_scores?.overall_confidence ?? 0,
        updatedAt: new Date().toISOString(),
      };

      await projectRef.set(
        {
          currentPhase: 'analyzed',
          phaseScores: existingScores,
          phaseStatus: 'in_progress',
          phaseHistory,
          updatedAt: FieldValue.serverTimestamp(),
        },
        { merge: true },
      );

      console.log(`[batch-extract] Updated project phase to 'analyzed' for project ${projectId}`);
    }

    return NextResponse.json({
      message: `Processed ${itemsToProcess.length} items: ${successCount} succeeded, ${results.filter(r => !r.success).length} failed`,
      results,
      successCount,
      totalProcessed: itemsToProcess.length
    });
  } catch (error) {
    console.error('[batch-extract] Batch extraction failed:', error);
    return NextResponse.json(
      {
        error: 'Failed to batch extract knowledge items',
        details: error instanceof Error ? error.message : String(error),
      },
      { status: 500 },
    );
  }
}