/**
 * Server-side helpers for AlloyDB vector memory operations
 * 
 * Handles CRUD operations on knowledge_chunks and semantic search.
 */

import { getAlloyDbClient, executeQuery, getPooledClient } from '@/lib/db/alloydb';
import type {
  KnowledgeChunk,
  KnowledgeChunkRow,
  KnowledgeChunkSearchResult,
  VectorSearchOptions,
  CreateKnowledgeChunkInput,
  BatchCreateKnowledgeChunksInput,
} from '@/lib/types/vector-memory';

/**
 * Convert database row (snake_case) to TypeScript object (camelCase)
 */
function rowToKnowledgeChunk(row: KnowledgeChunkRow): KnowledgeChunk {
  return {
    id: row.id,
    projectId: row.project_id,
    knowledgeItemId: row.knowledge_item_id,
    chunkIndex: row.chunk_index,
    content: row.content,
    sourceType: row.source_type,
    importance: row.importance,
    createdAt: row.created_at,
    updatedAt: row.updated_at,
  };
}

/**
 * Retrieve relevant knowledge chunks using vector similarity search
 * 
 * @param projectId - Firestore project ID to filter by
 * @param queryEmbedding - Vector embedding of the query (e.g., user's question)
 * @param options - Search options (limit, filters, etc.)
 * @returns Array of chunks ordered by similarity (most relevant first)
 * 
 * @example
 * ```typescript
 * const embedding = await embedText("What's the MVP scope?");
 * const chunks = await retrieveRelevantChunks('proj123', embedding, { limit: 10, minSimilarity: 0.7 });
 * ```
 */
export async function retrieveRelevantChunks(
  projectId: string,
  queryEmbedding: number[],
  options: VectorSearchOptions = {}
): Promise<KnowledgeChunkSearchResult[]> {
  const {
    limit = 10,
    minSimilarity,
    sourceTypes,
    importanceLevels,
  } = options;

  try {
    // Build the query with optional filters
    let queryText = `
      SELECT 
        id, 
        project_id, 
        knowledge_item_id, 
        chunk_index, 
        content, 
        source_type, 
        importance, 
        created_at, 
        updated_at,
        1 - (embedding <=> $1::vector) AS similarity
      FROM knowledge_chunks
      WHERE project_id = $2
    `;

    const params: any[] = [JSON.stringify(queryEmbedding), projectId];
    let paramIndex = 3;

    // Filter by source types
    if (sourceTypes && sourceTypes.length > 0) {
      queryText += ` AND source_type = ANY($${paramIndex})`;
      params.push(sourceTypes);
      paramIndex++;
    }

    // Filter by importance levels
    if (importanceLevels && importanceLevels.length > 0) {
      queryText += ` AND importance = ANY($${paramIndex})`;
      params.push(importanceLevels);
      paramIndex++;
    }

    // Filter by minimum similarity
    if (minSimilarity !== undefined) {
      queryText += ` AND (1 - (embedding <=> $1::vector)) >= $${paramIndex}`;
      params.push(minSimilarity);
      paramIndex++;
    }

    // Order by similarity and limit
    queryText += ` ORDER BY embedding <=> $1::vector LIMIT $${paramIndex}`;
    params.push(limit);

    const result = await executeQuery<KnowledgeChunkRow & { similarity: number }>(
      queryText,
      params
    );

    return result.rows.map((row) => ({
      ...rowToKnowledgeChunk(row),
      similarity: row.similarity,
    }));
  } catch (error) {
    console.error('[Vector Memory] Failed to retrieve relevant chunks:', error);
    throw new Error(
      `Failed to retrieve chunks: ${error instanceof Error ? error.message : String(error)}`
    );
  }
}

/**
 * Create a single knowledge chunk
 */
export async function createKnowledgeChunk(
  input: CreateKnowledgeChunkInput
): Promise<KnowledgeChunk> {
  const {
    projectId,
    knowledgeItemId,
    chunkIndex,
    content,
    embedding,
    sourceType = null,
    importance = null,
  } = input;

  try {
    const queryText = `
      INSERT INTO knowledge_chunks (
        project_id, 
        knowledge_item_id, 
        chunk_index, 
        content, 
        embedding, 
        source_type, 
        importance
      )
      VALUES ($1, $2, $3, $4, $5::vector, $6, $7)
      RETURNING 
        id, 
        project_id, 
        knowledge_item_id, 
        chunk_index, 
        content, 
        source_type, 
        importance, 
        created_at, 
        updated_at
    `;

    const result = await executeQuery<KnowledgeChunkRow>(queryText, [
      projectId,
      knowledgeItemId,
      chunkIndex,
      content,
      JSON.stringify(embedding),
      sourceType,
      importance,
    ]);

    if (result.rows.length === 0) {
      throw new Error('Failed to insert knowledge chunk');
    }

    return rowToKnowledgeChunk(result.rows[0]);
  } catch (error) {
    console.error('[Vector Memory] Failed to create knowledge chunk:', error);
    throw new Error(
      `Failed to create chunk: ${error instanceof Error ? error.message : String(error)}`
    );
  }
}

/**
 * Batch create multiple knowledge chunks efficiently
 * 
 * Uses a transaction to ensure atomicity.
 */
export async function batchCreateKnowledgeChunks(
  input: BatchCreateKnowledgeChunksInput
): Promise<KnowledgeChunk[]> {
  const { projectId, knowledgeItemId, chunks } = input;

  if (chunks.length === 0) {
    return [];
  }

  const client = await getPooledClient();

  try {
    await client.query('BEGIN');

    const createdChunks: KnowledgeChunk[] = [];

    for (const chunk of chunks) {
      const queryText = `
        INSERT INTO knowledge_chunks (
          project_id, 
          knowledge_item_id, 
          chunk_index, 
          content, 
          embedding, 
          source_type, 
          importance
        )
        VALUES ($1, $2, $3, $4, $5::vector, $6, $7)
        RETURNING 
          id, 
          project_id, 
          knowledge_item_id, 
          chunk_index, 
          content, 
          source_type, 
          importance, 
          created_at, 
          updated_at
      `;

      const result = await client.query<KnowledgeChunkRow>(queryText, [
        projectId,
        knowledgeItemId,
        chunk.chunkIndex,
        chunk.content,
        JSON.stringify(chunk.embedding),
        chunk.sourceType ?? null,
        chunk.importance ?? null,
      ]);

      if (result.rows.length > 0) {
        createdChunks.push(rowToKnowledgeChunk(result.rows[0]));
      }
    }

    await client.query('COMMIT');

    console.log(
      `[Vector Memory] Batch created ${createdChunks.length} chunks for knowledge_item ${knowledgeItemId}`
    );

    return createdChunks;
  } catch (error) {
    await client.query('ROLLBACK');
    console.error('[Vector Memory] Failed to batch create chunks:', error);
    throw new Error(
      `Failed to batch create chunks: ${error instanceof Error ? error.message : String(error)}`
    );
  } finally {
    client.release();
  }
}

/**
 * Delete all chunks for a specific knowledge_item
 * 
 * Used when regenerating chunks or removing a knowledge_item.
 */
export async function deleteChunksForKnowledgeItem(
  knowledgeItemId: string
): Promise<number> {
  try {
    const queryText = `
      DELETE FROM knowledge_chunks 
      WHERE knowledge_item_id = $1
      RETURNING id
    `;

    const result = await executeQuery(queryText, [knowledgeItemId]);

    console.log(
      `[Vector Memory] Deleted ${result.rowCount ?? 0} chunks for knowledge_item ${knowledgeItemId}`
    );

    return result.rowCount ?? 0;
  } catch (error) {
    console.error('[Vector Memory] Failed to delete chunks:', error);
    throw new Error(
      `Failed to delete chunks: ${error instanceof Error ? error.message : String(error)}`
    );
  }
}

/**
 * Delete all chunks for a specific project
 * 
 * Used when cleaning up or resetting a project.
 */
export async function deleteChunksForProject(projectId: string): Promise<number> {
  try {
    const queryText = `
      DELETE FROM knowledge_chunks 
      WHERE project_id = $1
      RETURNING id
    `;

    const result = await executeQuery(queryText, [projectId]);

    console.log(
      `[Vector Memory] Deleted ${result.rowCount ?? 0} chunks for project ${projectId}`
    );

    return result.rowCount ?? 0;
  } catch (error) {
    console.error('[Vector Memory] Failed to delete project chunks:', error);
    throw new Error(
      `Failed to delete project chunks: ${error instanceof Error ? error.message : String(error)}`
    );
  }
}

/**
 * Get chunk count for a knowledge_item
 */
export async function getChunkCountForKnowledgeItem(
  knowledgeItemId: string
): Promise<number> {
  try {
    const result = await executeQuery<{ count: string }>(
      'SELECT COUNT(*) as count FROM knowledge_chunks WHERE knowledge_item_id = $1',
      [knowledgeItemId]
    );

    return parseInt(result.rows[0]?.count ?? '0', 10);
  } catch (error) {
    console.error('[Vector Memory] Failed to get chunk count:', error);
    return 0;
  }
}

/**
 * Get chunk count for a project
 */
export async function getChunkCountForProject(projectId: string): Promise<number> {
  try {
    const result = await executeQuery<{ count: string }>(
      'SELECT COUNT(*) as count FROM knowledge_chunks WHERE project_id = $1',
      [projectId]
    );

    return parseInt(result.rows[0]?.count ?? '0', 10);
  } catch (error) {
    console.error('[Vector Memory] Failed to get project chunk count:', error);
    return 0;
  }
}

/**
 * Regenerate knowledge_chunks for a single knowledge_item
 * 
 * This is the main pipeline that:
 * 1. Chunks the knowledge_item.content
 * 2. Generates embeddings for each chunk
 * 3. Deletes existing chunks for this item
 * 4. Inserts new chunks into AlloyDB
 * 
 * @param knowledgeItem - The knowledge item to process
 * 
 * @example
 * ```typescript
 * const knowledgeItem = await getKnowledgeItem(projectId, itemId);
 * await writeKnowledgeChunksForItem(knowledgeItem);
 * ```
 */
export async function writeKnowledgeChunksForItem(
  knowledgeItem: {
    id: string;
    projectId: string;
    content: string;
    sourceMeta?: { sourceType?: string; importance?: 'primary' | 'supporting' | 'irrelevant' };
  }
): Promise<void> {
  const { chunkText } = await import('@/lib/ai/chunking');
  const { embedTextBatch } = await import('@/lib/ai/embeddings');

  try {
    console.log(
      `[Vector Memory] Starting chunking pipeline for knowledge_item ${knowledgeItem.id}`
    );

    // Step 1: Chunk the content
    const textChunks = chunkText(knowledgeItem.content, {
      maxTokens: 800,
      overlapChars: 200,
      preserveParagraphs: true,
    });

    if (textChunks.length === 0) {
      console.warn(
        `[Vector Memory] No chunks generated for knowledge_item ${knowledgeItem.id} - content may be empty`
      );
      return;
    }

    console.log(
      `[Vector Memory] Generated ${textChunks.length} chunks for knowledge_item ${knowledgeItem.id}`
    );

    // Step 2: Generate embeddings for all chunks
    const chunkTexts = textChunks.map((chunk) => chunk.text);
    const embeddings = await embedTextBatch(chunkTexts, {
      delayMs: 50, // Small delay to avoid rate limiting
      skipEmpty: true,
    });

    if (embeddings.length !== textChunks.length) {
      throw new Error(
        `Embedding count mismatch: got ${embeddings.length}, expected ${textChunks.length}`
      );
    }

    // Step 3: Delete existing chunks for this knowledge_item
    await deleteChunksForKnowledgeItem(knowledgeItem.id);

    // Step 4: Insert new chunks
    const chunksToInsert = textChunks.map((chunk, index) => ({
      chunkIndex: chunk.index,
      content: chunk.text,
      embedding: embeddings[index],
      sourceType: knowledgeItem.sourceMeta?.sourceType ?? null,
      importance: knowledgeItem.sourceMeta?.importance ?? null,
    }));

    await batchCreateKnowledgeChunks({
      projectId: knowledgeItem.projectId,
      knowledgeItemId: knowledgeItem.id,
      chunks: chunksToInsert,
    });

    console.log(
      `[Vector Memory] Successfully processed ${chunksToInsert.length} chunks for knowledge_item ${knowledgeItem.id}`
    );
  } catch (error) {
    console.error(
      `[Vector Memory] Failed to write chunks for knowledge_item ${knowledgeItem.id}:`,
      error
    );
    throw new Error(
      `Failed to write chunks: ${error instanceof Error ? error.message : String(error)}`
    );
  }
}