@fluentcommerce/fc-connect-sdk 0.1.53 → 0.1.55

package/docs/01-TEMPLATES/versori/workflows/extraction/extraction-modes-guide.md

@@ -1,1038 +1,1038 @@
-# Extraction Modes Guide
-
-**FC Connect SDK - Choosing the Right Extraction Mode**
-
-> This guide explains the two extraction modes available in the SDK and when to use each one.
-
----
-
-## Overview
-
-The FC Connect SDK supports two extraction modes for GraphQL queries:
-
-1. **Incremental** - Extract only changed records since last run (recommended for scheduled workflows)
-2. **Date Range** - Extract records within specific date window (for ad-hoc queries, backfills, or historical data)
-
-**Note:** "Historical extraction" is simply Date Range mode with a very old start date (e.g., `updatedAfter: '1970-01-01'`). There is no separate "historical mode" - use Date Range with appropriate validation and safeguards.
-
----
-
-## Quick Comparison Table
-
-| Mode            | Use Case                                      | Safety Level | Recommended Frequency | Typical Records | Production Risk                     |
-| --------------- | --------------------------------------------- | ------------ | --------------------- | --------------- | ----------------------------------- |
-| **Incremental** | Scheduled syncs, recurring extractions        | ✅ **Safe**  | Hourly/Daily          | ~10k            | ✅ **Low** - Safe for production    |
-| **Date Range**  | Ad-hoc audits, backfills, historical dumps    | ⚠️ **Risky** | One-time only         | ~50k            | ⚠️ **Medium** - Requires validation |
-
----
-
-## Mode 1: Incremental (Recommended) ✅
-
-### What It Does
-
-Extracts only records updated since the last successful extraction run.
-
-### How It Works
-
-```typescript
-const startTime = Date.now();
-log.info('📦 Starting incremental extraction');
-
-// Load last run timestamp from state
-const lastRunTime = await kv.get(['extraction', 'products', 'lastRunTime']);
-log.info('⏱️ Last run timestamp loaded', { lastRunTime });
-
-// Query with buffered timestamp (overlap buffer prevents gaps)
-const bufferedTime = new Date(lastRunTime - 60000).toISOString(); // 60 second buffer
-
-const result = await client.graphql({
-  query: PRODUCTS_QUERY,
-  variables: {
-    retailerId: 'my-retailer',
-    updatedAfter: bufferedTime, // Only changed records
-    first: 200,
-  },
-  pagination: { maxRecords: 10000 },
-});
-
-log.info('✅ Extraction complete', {
-  recordCount: result.edges.length,
-  duration: Date.now() - startTime
-});
-
-// Save new timestamp after successful extraction
-await kv.set(['extraction', 'products', 'lastRunTime'], {
-  timestamp: new Date().toISOString(),
-});
-```
-
-### When to Use
-
-- ✅ **Scheduled extractions** (hourly, daily, every 15 minutes)
-- ✅ **Real-time inventory feeds**
-- ✅ **Order status updates**
-- ✅ **Product catalog syncs**
-- ✅ **Any recurring extraction workflow**
-
-### Safety Features
-
-- Natural rate limiting via timestamps
-- Overlap buffer prevents missed records
-- State tracking prevents reprocessing
-- Predictable record counts
-- Automatic recovery after failures
-
-### Recommended Settings
-
-| Entity                   | Frequency    | Max Records | Expected Volume    |
-| ------------------------ | ------------ | ----------- | ------------------ |
-| **Products**             | Daily        | 10,000      | 5k-10k updates/day |
-| **Orders**               | Hourly       | 5,000       | 100-500/hour       |
-| **Fulfillments**         | Hourly       | 5,000       | 100-500/hour       |
-| **Inventory Quantities** | Every 15 min | 10,000      | 500-2k per run     |
-| **Inventory Positions**  | Daily        | 20,000      | 10k-20k per day    |
-| **Virtual Positions**    | Hourly       | 10,000      | 1k-5k per hour     |
-
-### Configuration Example
-
-```json
-{
-  "extractionMode": "incremental",
-  "pageSize": 200,
-  "maxRecords": 10000,
-  "overlapBufferSeconds": 60,
-  "fallbackStartDate": "2024-01-01T00:00:00Z"
-}
-```
-
----
-
-## Mode 2: Date Range (Use Carefully) ⚠️
-
-### What It Does
-
-Extracts all records updated within a specific date window.
-
-### How It Works
-
-```typescript
-const startTime = Date.now();
-log.info('📦 Starting date range extraction');
-
-// User provides specific date range
-const startDate = ctx.activation?.getVariable('startDate'); // "2025-01-01T00:00:00Z"
-const endDate = ctx.activation?.getVariable('endDate'); // "2025-01-31T23:59:59Z"
-
-log.info('⏱️ Date range configured', { startDate, endDate });
-
-const result = await client.graphql({
-  query: PRODUCTS_QUERY,
-  variables: {
-    retailerId: 'my-retailer',
-    updatedAfter: startDate,
-    updatedBefore: endDate, // Bounded range
-    first: 200,
-  },
-  pagination: { maxRecords: 50000 },
-});
-
-log.info('✅ Extraction complete', {
-  recordCount: result.edges.length,
-  duration: Date.now() - startTime
-});
-```
-
-### When to Use
-
-- ⚠️ **One-time data audits** (e.g., "show me all orders from Q4 2024")
-- ⚠️ **Backfilling missed extractions** (e.g., extraction failed for 3 days)
-- ⚠️ **Historical reporting** (e.g., "export last month's inventory changes")
-- ⚠️ **Data quality investigations**
-
-### Risks
-
-- Can return tens of thousands of records
-- No natural rate limiting
-- Easy to exceed platform limits
-- Can timeout on large datasets
-- Risk of platform overload
-
-### Safety Guardrails Required
-
-```typescript
-// ✅ REQUIRED: Validate date range before extraction
-function validateDateRange(startDate: string, endDate: string, log: any) {
-  const start = new Date(startDate);
-  const end = new Date(endDate);
-  const daysDiff = (end.getTime() - start.getTime()) / (1000 * 60 * 60 * 24);
-
-  log.info('🔍 Validating date range', { startDate, endDate, daysDiff });
-
-  // Enforce maximum date range
-  const MAX_DAYS = 30;
-  if (daysDiff > MAX_DAYS) {
-    log.error('❌ Date range validation failed', {
-      daysDiff,
-      maxDays: MAX_DAYS,
-      recommendation: 'Split into smaller ranges or use incremental mode'
-    });
-    throw new Error(
-      `Date range too large: ${daysDiff} days (max: ${MAX_DAYS}). ` +
-        `Split into smaller ranges or use incremental mode.`
-    );
-  }
-
-  log.info('✅ Date range validation passed', { daysDiff, maxDays: MAX_DAYS });
-  return { daysDiff, isValid: true };
-}
-```
-
-### Recommended Limits
-
-| Entity           | Max Date Range | Max Records | File Splitting    |
-| ---------------- | -------------- | ----------- | ----------------- |
-| **Products**     | 30 days        | 50,000      | Required if > 25k |
-| **Orders**       | 7 days         | 20,000      | Required if > 10k |
-| **Fulfillments** | 7 days         | 20,000      | Required if > 10k |
-| **Inventory**    | 14 days        | 50,000      | Required if > 25k |
-
-### Configuration Example
-
-```json
-{
-  "extractionMode": "dateRange",
-  "startDate": "2025-01-01T00:00:00Z",
-  "endDate": "2025-01-31T23:59:59Z",
-  "pageSize": 200,
-  "maxRecords": 50000,
-  "validateDateRange": true,
-  "maxDaysAllowed": 30
-}
-```
-
-### Production Restrictions
-
-- ❌ **Never schedule** date range extractions
-- ❌ **Never use in recurring workflows**
-- ✅ **Always validate date range** before execution
-- ✅ **Always monitor execution time** and record counts
-- ✅ **Always implement file splitting** for large results
-- ✅ **Always get approval** before running on production data
-
----
-
-## File Splitting for Large Extractions
-
-### Overview
-
-When ExtractionOrchestrator fetches large datasets (e.g., 50,000 records), all records are loaded into memory. Before writing to destination (SFTP/S3), you can split these records into multiple smaller files for better manageability, parallel processing, and downstream system compatibility.
-
-### When to Split Files
-
-- ✅ **Extraction returns >10k records** - Split into manageable chunks
-- ✅ **Downstream system has file size limits** - Respect partner constraints
-- ✅ **Parallel processing desired** - Write multiple files concurrently
-- ✅ **Network reliability concerns** - Smaller files = better retry granularity
-- ✅ **Audit trail requirements** - Track processing per file
-
-### Basic File Splitting Pattern
-
-```typescript
-import { Buffer } from 'node:buffer';
-import { ExtractionOrchestrator, SftpDataSource } from '@fluentcommerce/fc-connect-sdk';
-
-const startTime = Date.now();
-log.info('📦 Starting extraction with file splitting');
-
-// STEP 1: Extract all records (loads into memory)
-const result = await orchestrator.extract({
-  query: virtualPositionsQuery,
-  variables: { retailerId: 'my-retailer', updatedAfter: lastRunTime },
-  pagination: { pageSize: 200, maxRecords: 50000 },
-});
-
-log.info('✅ Extraction complete', {
-  recordCount: result.data.length,
-  duration: Date.now() - startTime
-});
-
-// result.data contains all 50,000 records in memory
-
-// STEP 2: Configure chunking
-const RECORDS_PER_FILE = 1000; // Configurable chunk size
-const timestamp = new Date().toISOString().replace(/[:.]/g, '-');
-
-// STEP 3: Split records into chunks
-const chunks: any[][] = [];
-for (let i = 0; i < result.data.length; i += RECORDS_PER_FILE) {
-  chunks.push(result.data.slice(i, i + RECORDS_PER_FILE));
-}
-
-log.info('📂 File splitting configured', {
-  totalRecords: result.data.length,
-  recordsPerFile: RECORDS_PER_FILE,
-  filesToCreate: chunks.length,
-});
-
-// STEP 4: Generate files with configurable naming pattern
-const filePromises = chunks.map((chunk, index) => {
-  // File naming pattern: entity-timestamp-part-NNN.format
-  const partNumber = String(index + 1).padStart(3, '0'); // 001, 002, 003...
-  const filename = `virtual-positions-${timestamp}-part-${partNumber}.xml`;
-
-  // Transform chunk to desired format (XML, CSV, JSON)
-  const xmlContent = xmlBuilder.build({ records: chunk });
-
-  // Return upload promise
-  return sftp.uploadFile(
-    `/outbound/${filename}`,
-    Buffer.from(xmlContent, 'utf8'),
-    { encoding: 'utf8', overwrite: false }
-  );
-});
-
-// STEP 5: Write all files in parallel
-const uploadStartTime = Date.now();
-await Promise.all(filePromises);
-
-log.info('✅ File splitting complete', {
-  filesCreated: chunks.length,
-  totalRecords: result.data.length,
-  recordsPerFile: RECORDS_PER_FILE,
-  uploadDuration: Date.now() - uploadStartTime,
-  totalDuration: Date.now() - startTime
-});
-```
-
-### Configurable File Naming Patterns
-
-Support multiple naming patterns via configuration:
-
-```typescript
-// Configuration interface
-interface FileSplittingConfig {
-  enabled: boolean;
-  recordsPerFile: number;
-  namingPattern: 'sequential' | 'timestamp' | 'range' | 'custom';
-  customPattern?: (index: number, chunk: any[], total: number) => string;
-}
-
-// Pattern examples
-const patterns = {
-  // Pattern 1: Sequential numbering
-  sequential: (index: number) =>
-    `virtual-positions-part-${String(index + 1).padStart(3, '0')}.xml`,
-
-  // Pattern 2: Timestamp-based
-  timestamp: (index: number) =>
-    `virtual-positions-${new Date().toISOString()}-${index + 1}.xml`,
-
-  // Pattern 3: Record range
-  range: (index: number, chunk: any[]) => {
-    const start = index * chunk.length + 1;
-    const end = start + chunk.length - 1;
-    return `virtual-positions-records-${start}-${end}.xml`;
-  },
-
-  // Pattern 4: Custom with metadata
-  custom: (index: number, chunk: any[], totalChunks: number) =>
-    `VP_${new Date().toISOString().split('T')[0]}_${index + 1}_of_${totalChunks}.xml`,
-};
-```
-
-### Complete Example: Extraction with File Splitting
-
-```typescript
-import { schedule, fn, MemoryInterpreter } from '@versori/run';
-import { Buffer } from 'node:buffer';
-import {
-  ExtractionOrchestrator,
-  createClient,
-  SftpDataSource,
-  VersoriKVAdapter,
-  XMLBuilder,
-} from '@fluentcommerce/fc-connect-sdk';
-
-export const virtualPositionsExtraction = schedule(
-  'virtual-positions-hourly',
-  '0 * * * *'
-).then(
-  fn('extract', async (ctx) => {
-    const { log, activation, openKv } = ctx;
-    const startTime = Date.now();
-
-    log.info('📦 Starting virtual positions extraction');
-
-    // Configuration from activation variables
-    const FILE_SPLITTING_ENABLED = activation.getVariable('fileSplittingEnabled') !== 'false';
-    const RECORDS_PER_FILE = parseInt(activation.getVariable('recordsPerFile') || '1000', 10);
-    const FILE_NAMING_PATTERN = activation.getVariable('fileNamingPattern') || 'sequential';
-
-    try {
-      // Initialize services
-      const client = await createClient(ctx);
-      log.info('✅ Client initialized');
-
-      const kv = new VersoriKVAdapter(openKv(':project:'));
-      const orchestrator = new ExtractionOrchestrator(client, log);
-
-      // Extract records (all loaded into memory)
-      const extractionStartTime = Date.now();
-      const result = await orchestrator.extract({
-        query: virtualPositionsQuery,
-        variables: {
-          retailerId: activation.getVariable('fluentRetailerId'),
-          updatedAfter: await kv.get(['extraction', 'lastRunTime']) || '2025-01-01T00:00:00Z',
-        },
-        pagination: { pageSize: 200, maxRecords: 50000 },
-      });
-
-      log.info('✅ Extraction complete', {
-        recordCount: result.data.length,
-        fileSplittingEnabled: FILE_SPLITTING_ENABLED,
-        extractionDuration: Date.now() - extractionStartTime
-      });
-
-      // Initialize SFTP
-      const sftp = new SftpDataSource({
-        type: 'SFTP_XML',
-        connectionId: 'sftp-extractions',
-        name: 'extraction-sftp',
-        settings: {
-          host: activation.getVariable('sftpHost'),
-          port: parseInt(activation.getVariable('sftpPort') || '22', 10),
-          username: activation.getVariable('sftpUsername'),
-          password: activation.getVariable('sftpPassword'),
-          remotePath: '/outbound/',
-          requireAbsolutePaths: true,
-        },
-      }, log);
-
-      await sftp.validateConnection();
-      log.info('✅ SFTP connection validated');
-
-      // File splitting logic
-      if (FILE_SPLITTING_ENABLED && result.data.length > RECORDS_PER_FILE) {
-        const splittingStartTime = Date.now();
-
-        // Split into chunks
-        const chunks: any[][] = [];
-        for (let i = 0; i < result.data.length; i += RECORDS_PER_FILE) {
-          chunks.push(result.data.slice(i, i + RECORDS_PER_FILE));
-        }
-
-        log.info('📂 Splitting extraction into multiple files', {
-          totalRecords: result.data.length,
-          recordsPerFile: RECORDS_PER_FILE,
-          filesToCreate: chunks.length,
-        });
-
-        // Generate files
-        const timestamp = new Date().toISOString().replace(/[:.]/g, '-');
-        const xmlBuilder = new XMLBuilder();
-
-        const filePromises = chunks.map(async (chunk, index) => {
-          // Generate filename based on pattern
-          let filename: string;
-          const partNumber = String(index + 1).padStart(3, '0');
-
-          switch (FILE_NAMING_PATTERN) {
-            case 'timestamp':
-              filename = `virtual-positions-${timestamp}-${partNumber}.xml`;
-              break;
-            case 'range':
-              const start = index * RECORDS_PER_FILE + 1;
-              const end = Math.min(start + chunk.length - 1, result.data.length);
-              filename = `virtual-positions-records-${start}-${end}.xml`;
-              break;
-            case 'sequential':
-            default:
-              filename = `virtual-positions-part-${partNumber}.xml`;
-              break;
-          }
-
-          // Build XML content
-          const xmlContent = xmlBuilder.build({ virtualPositions: chunk });
-
-          // Upload to SFTP
-          await sftp.uploadFile(
-            `/outbound/${filename}`,
-            Buffer.from(xmlContent, 'utf8'),
-            { encoding: 'utf8', overwrite: false }
-          );
-
-          return { filename, recordCount: chunk.length };
-        });
-
-        // Write all files in parallel
-        const uploadedFiles = await Promise.all(filePromises);
-
-        log.info('✅ File splitting complete', {
-          filesCreated: uploadedFiles.length,
-          totalRecords: result.data.length,
-          files: uploadedFiles,
-          splittingDuration: Date.now() - splittingStartTime,
-          totalDuration: Date.now() - startTime
-        });
-
-        return {
-          success: true,
-          extractionMode: 'incremental',
-          totalRecords: result.data.length,
-          filesCreated: uploadedFiles.length,
-          files: uploadedFiles,
-          duration: Date.now() - startTime
-        };
-      } else {
-        // Single file (no splitting)
-        const timestamp = new Date().toISOString().replace(/[:.]/g, '-');
-        const filename = `virtual-positions-${timestamp}.xml`;
-        const xmlBuilder = new XMLBuilder();
-        const xmlContent = xmlBuilder.build({ virtualPositions: result.data });
-
-        await sftp.uploadFile(
-          `/outbound/${filename}`,
-          Buffer.from(xmlContent, 'utf8'),
-          { encoding: 'utf8', overwrite: false }
-        );
-
-        log.info('✅ Single file extraction complete', {
-          filename,
-          recordCount: result.data.length,
-          duration: Date.now() - startTime
-        });
-
-        return {
-          success: true,
-          extractionMode: 'incremental',
-          totalRecords: result.data.length,
-          filesCreated: 1,
-          files: [{ filename, recordCount: result.data.length }],
-          duration: Date.now() - startTime
-        };
-      }
-    } catch (error: any) {
-      log.error('❌ Extraction failed', {
-        message: error instanceof Error ? error.message : String(error),
-        stack: error instanceof Error ? error.stack : undefined,
-        errorType: error instanceof Error ? error.constructor.name : 'Error',
-        duration: Date.now() - startTime
-      });
-
-      return {
-        success: false,
-        error: error.message,
-        duration: Date.now() - startTime
-      };
-    } finally {
-      if (sftp) {
-        await sftp.dispose();
-        log.info('✅ SFTP connection disposed');
-      }
-    }
-  })
-);
-
-// Export with MemoryInterpreter for Versori platform
-export const interpreter = new MemoryInterpreter({
-  workflows: [virtualPositionsExtraction]
-});
-```
-
-### Date Range Splitting for Very Large Datasets
-
-For extractions spanning multiple years (e.g., historical dumps), split the date range and run multiple smaller extractions:
-
-```typescript
-// Split large date range into manageable chunks
-async function extractHistoricalData(
-  startDate: string,  // "2020-01-01"
-  endDate: string,    // "2025-01-01"
-  chunkMonths: number = 1,  // Extract 1 month at a time
-  log: any
-) {
-  const overallStartTime = Date.now();
-  const results = [];
-  let currentStart = new Date(startDate);
-  const finalEnd = new Date(endDate);
-
-  log.info('📦 Starting historical data extraction', {
-    startDate,
-    endDate,
-    chunkMonths,
-    estimatedChunks: Math.ceil((finalEnd.getTime() - currentStart.getTime()) / (chunkMonths * 30 * 24 * 60 * 60 * 1000))
-  });
-
-  while (currentStart < finalEnd) {
-    const chunkStartTime = Date.now();
-
-    // Calculate chunk end date
-    const currentEnd = new Date(currentStart);
-    currentEnd.setMonth(currentEnd.getMonth() + chunkMonths);
-
-    // Ensure we don't exceed final end date
-    if (currentEnd > finalEnd) {
-      currentEnd.setTime(finalEnd.getTime());
-    }
-
-    log.info('📂 Extracting date range chunk', {
-      chunkNumber: results.length + 1,
-      startDate: currentStart.toISOString(),
-      endDate: currentEnd.toISOString(),
-    });
-
-    // Extract for this date range
-    const result = await orchestrator.extract({
-      query: virtualPositionsQuery,
-      variables: {
-        retailerId: 'my-retailer',
-        updatedAfter: currentStart.toISOString(),
-        updatedBefore: currentEnd.toISOString(),
-      },
-      pagination: { pageSize: 200, maxRecords: 50000 },
-    });
-
-    log.info('✅ Chunk extraction complete', {
-      chunkNumber: results.length + 1,
-      recordCount: result.data.length,
-      chunkDuration: Date.now() - chunkStartTime
-    });
-
-    // Split this chunk into files if needed
-    if (result.data.length > RECORDS_PER_FILE) {
-      const chunks: any[][] = [];
-      for (let i = 0; i < result.data.length; i += RECORDS_PER_FILE) {
-        chunks.push(result.data.slice(i, i + RECORDS_PER_FILE));
-      }
-
-      // Write chunks in parallel
-      const timestamp = currentStart.toISOString().split('T')[0];
-      await Promise.all(
-        chunks.map((chunk, index) => {
-          const filename = `historical-${timestamp}-part-${String(index + 1).padStart(3, '0')}.xml`;
-          return sftp.uploadFile(`/outbound/${filename}`, Buffer.from(xmlBuilder.build(chunk), 'utf8'));
-        })
-      );
-    }
-
-    results.push({
-      startDate: currentStart.toISOString(),
-      endDate: currentEnd.toISOString(),
-      recordCount: result.data.length,
-      duration: Date.now() - chunkStartTime
-    });
-
-    // Move to next chunk
-    currentStart = new Date(currentEnd);
-
-    // Rate limiting - wait between chunks
-    log.info('⏱️ Rate limiting delay (5 seconds)');
-    await new Promise(resolve => setTimeout(resolve, 5000)); // 5 second delay
-  }
-
-  const totalRecords = results.reduce((sum, r) => sum + r.recordCount, 0);
-  const totalDuration = Date.now() - overallStartTime;
-
-  log.info('✅ Historical extraction complete', {
-    totalChunks: results.length,
-    totalRecords,
-    totalDuration,
-    averageRecordsPerChunk: Math.round(totalRecords / results.length)
-  });
-
-  return results;
-}
-
-// Usage: Extract 5 years of data, 1 month at a time
-const historicalResults = await extractHistoricalData('2020-01-01', '2025-01-01', 1, log);
-```
-
-### Configuration Variables for File Splitting
-
-Add these to Versori activation variables:
-
-```bash
-# File Splitting Configuration
-FILE_SPLITTING_ENABLED=true              # Enable/disable file splitting
-RECORDS_PER_FILE=1000                    # Records per file (default: 1000)
-FILE_NAMING_PATTERN=sequential           # Naming pattern: sequential | timestamp | range
-MAX_RECORDS_PER_EXTRACTION=50000         # Maximum records to extract in one run
-
-# Date Range Splitting (for large historical extractions)
-DATE_RANGE_CHUNK_MONTHS=1                # Months per extraction chunk (default: 1)
-RATE_LIMIT_DELAY_MS=5000                 # Delay between chunks (milliseconds)
-```
-
-### Benefits of File Splitting
-
-- ✅ **Memory efficiency** - Process large datasets without memory overflow
-- ✅ **Parallel writes** - Multiple files written concurrently (faster)
-- ✅ **Better error recovery** - Retry individual files vs entire extraction
-- ✅ **Downstream compatibility** - Honor partner file size limits
-- ✅ **Audit granularity** - Track processing per file
-- ✅ **Network resilience** - Smaller files = better upload success rate
-
-### Limitations
-
-- All records must fit in memory during extraction (ExtractionOrchestrator loads all pages)
-- File splitting happens post-extraction (not during pagination)
-- Parallel writes limited by available memory and network bandwidth
-- SFTP connection pool size may limit concurrency (default: 10 connections)
-
----
-
-## Historical Data Extraction via Date Range
-
-### Overview
-
-There is **no separate "historical mode"** in the SDK. To extract historical data (e.g., all records from 2020 onwards), use **Date Range mode** with a very old start date and appropriate safeguards.
-
-### How to Extract Historical Data
-
-```typescript
-// Historical extraction is just Date Range with old start date
-await client.graphql({
-  query: PRODUCTS_QUERY,
-  variables: {
-    retailerId: 'my-retailer',
-    updatedAfter: '1970-01-01T00:00:00Z',  // Very old date = "all records"
-    updatedBefore: new Date().toISOString(), // Up to now
-    first: 200,
-  },
-  pagination: { maxRecords: 50000 },
-});
-```
-
-### Required Safeguards for Historical Extraction
-
-1. **Date Range Splitting** - Split into smaller chunks (monthly/quarterly)
-2. **File Splitting** - Split large results into multiple files
-3. **Rate Limiting** - Add delays between chunks
-4. **Validation** - Verify date ranges before execution
-5. **Monitoring** - Track progress and alert on anomalies
-6. **Approval** - Get sign-off before running on production
-
-### Recommended Approach: Chunked Date Range Extraction
-
-```bash
-#!/bin/bash
-# Safe historical extraction via chunked date ranges
-
-START_DATE="2020-01-01"
-END_DATE="2025-01-01"
-
-# Extract one month at a time
-current=$START_DATE
-while [[ "$current" < "$END_DATE" ]]; do
-  # Calculate month end
-  monthEnd=$(date -d "$current + 1 month" +%Y-%m-%d)
-
-  echo "Extracting $current to $monthEnd..."
-
-  # Trigger date range extraction
-  curl -X POST https://versori-webhook.com/extract \
-    -H "Content-Type: application/json" \
-    -d "{
-      \"extractionMode\": \"dateRange\",
-      \"startDate\": \"${current}T00:00:00Z\",
-      \"endDate\": \"${monthEnd}T00:00:00Z\",
-      \"fileSplittingEnabled\": true,
-      \"recordsPerFile\": 1000
-    }"
-
-  # Rate limiting - wait 60 seconds between chunks
-  sleep 60
-
-  # Move to next month
-  current=$monthEnd
-done
-
-echo "Historical extraction complete"
-```
-
-### Migration to Incremental After Historical Load
-
-After completing a one-time historical extraction, switch to incremental mode for ongoing syncs:
-
-```json
-{
-  "extractionMode": "incremental",
-  "fallbackStartDate": "2025-01-22T00:00:00Z",
-  "pageSize": 200,
-  "maxRecords": 10000
-}
-```
-
----
-
-## Decision Tree: Which Mode to Use?
-
-```
-Start Here
-    │
-    ├─ Need recurring extractions? ─────────► Use INCREMENTAL ✅
-    │                                          (hourly/daily/every 15 min)
-    │                                          Tracks state, auto-recovery
-    │
-    └─ Need specific date range? ────────────► Use DATE RANGE ⚠️
-         │                                      (one-time, validate range)
-         │
-         ├─ Date range < 30 days? ──────────► Single DATE RANGE run
-         │                                     + file splitting if >10k records
-         │
-         └─ Date range > 30 days? ──────────► Split into monthly chunks
-              │                               + file splitting per chunk
-              │                               + rate limiting between chunks
-              │
-              └─ Historical (all data)? ────► DATE RANGE with old start date
-                                              + chunked approach (monthly)
-                                              + approval required
-```
-
----
-
-## Monitoring & Alerts
-
-Set up alerts for extraction volumes:
-
-```typescript
-// In extraction workflow
-const recordCount = edges.length;
-const ALERT_THRESHOLD = 50000;
-
-log.info('📊 Checking extraction volume', { recordCount, threshold: ALERT_THRESHOLD });
-
-if (recordCount > ALERT_THRESHOLD) {
-  log.error('❌ Extraction volume exceeded threshold', {
-    recordCount,
-    threshold: ALERT_THRESHOLD,
-    percentageOver: Math.round(((recordCount - ALERT_THRESHOLD) / ALERT_THRESHOLD) * 100),
-    mode: extractionMode,
-    recommendation: 'Switch to incremental mode or reduce date range',
-  });
-
-  // Send alert to monitoring system
-  await sendAlert({
-    severity: 'high',
-    message: `Extraction returned ${recordCount} records (threshold: ${ALERT_THRESHOLD})`,
-    mode: extractionMode,
-  });
-
-  log.info('✅ Alert sent to monitoring system');
-} else {
-  log.info('✅ Extraction volume within acceptable limits', {
-    recordCount,
-    threshold: ALERT_THRESHOLD,
-    percentageUsed: Math.round((recordCount / ALERT_THRESHOLD) * 100)
-  });
-}
-```
-
----
-
-## Summary & Best Practices
-
-### ✅ DO
-
-- Use **incremental mode** for all scheduled extractions
-- Validate date ranges before running **dateRange mode**
-- Implement **file splitting** for large extractions (>10k records)
-- Use **parallel writes** with `Promise.all()` for split files
-- Monitor extraction volumes and set alerts
-- Test on staging before production
-- Use overlap buffer (60s) to prevent gaps in incremental mode
-- Track state with VersoriKV
-- Split large historical extractions into monthly chunks
-- Add rate limiting between extraction chunks
-
-### ❌ DON'T
-
-- Schedule **dateRange** extractions (incremental only)
-- Use date ranges > 30 days without chunking
-- Skip validation checks for date ranges
-- Ignore volume alerts (>50k records)
-- Forget to implement file splitting for large results
-- Run historical extractions without approval and monitoring
-
----
-
-## ExtractionOrchestrator (SDK v0.1.27+)
-
-The SDK includes **ExtractionOrchestrator** - a high-level service that simplifies extraction workflows with built-in mode handling, pagination, and output management.
-
-### Why Use ExtractionOrchestrator?
-
-**Instead of manually implementing:**
-
-- Mode detection (incremental/dateRange)
-- Pagination loops
-- Path-based field extraction
-- Output formatting (CSV/JSON/Parquet)
-- S3/SFTP uploads
-- Error handling
-
-**ExtractionOrchestrator handles it all:**
-
-```typescript
-import { ExtractionOrchestrator, createClient } from '@fluentcommerce/fc-connect-sdk';
-
-const startTime = Date.now();
-log.info('📦 Initializing ExtractionOrchestrator');
-
-const client = await createClient(ctx);
-const orchestrator = new ExtractionOrchestrator(client, log);
-
-log.info('✅ Orchestrator initialized');
-
-// Both modes supported with single interface
-const result = await orchestrator.extract({
-  query: virtualPositionsQuery,
-  variables: { retailerId: 'my-retailer' },
-
-  // Mode: 'incremental' (scheduled) or 'dateRange' (ad-hoc)
-  extractionMode: 'incremental',
-  stateKey: 'virtual-positions-extraction',
-
-  // Pagination handled automatically
-  pagination: {
-    pageSize: 200,
-    maxRecords: 10000,
-  },
-
-  // Output format and destination
-  outputFormat: 'csv',
-  outputDestination: {
-    type: 's3',
-    bucket: 'my-extracts',
-    key: 'virtual-positions/hourly/{{timestamp}}.csv',
-  },
-
-  // Field extraction from nested paths
-  fieldPaths: {
-    position_ref: 'ref',
-    quantity: 'quantity',
-    location_ref: 'locationLink.ref',
-    location_name: 'locationLink.name',
-  },
-});
-
-log.info('✅ Extraction and upload complete', {
-  recordCount: result.recordCount,
-  outputFile: result.outputFile,
-  duration: Date.now() - startTime
-});
-```
-
-### Features
-
-- **Auto-pagination**: Handles cursor-based pagination automatically
-- **Mode support**: Both modes (incremental, dateRange)
-- **State management**: Tracks last run timestamps for incremental mode
-- **Path extraction**: Extracts nested fields from GraphQL responses
-- **Multi-format**: Outputs CSV, JSON, or Parquet
-- **Validation**: Built-in query and response validation
-- **Error recovery**: Graceful failure handling with detailed logs
-- **File splitting**: Post-extraction chunking for large datasets (add manually, see File Splitting section above)
-
-### Example: Incremental Extraction with ExtractionOrchestrator
-
-```typescript
-import { schedule, fn, MemoryInterpreter } from '@versori/run';
-import { Buffer } from 'node:buffer';  // Required for Deno/Versori runtime
-import {
-  ExtractionOrchestrator,
-  createClient,
-  VersoriKVAdapter,
-} from '@fluentcommerce/fc-connect-sdk';
-
-export const hourlyExtraction = schedule('hourly-virtual-positions', '0 * * * *').then(
-  fn('extract', async (ctx) => {
-    const { log, openKv, env } = ctx;
-    const startTime = Date.now();
-
-    log.info('📦 Starting hourly virtual positions extraction');
-
-    const client = await createClient(ctx);
-    log.info('✅ Client initialized');
-
-    const kv = new VersoriKVAdapter(openKv(':project:'));
-    const orchestrator = new ExtractionOrchestrator(client, log);
-
-    const result = await orchestrator.extract({
-      query: `
-        query GetVirtualPositions($retailerId: String!, $updatedAfter: String, $first: Int, $after: String) {
-          virtualPositions(retailerId: $retailerId, updatedAfter: $updatedAfter, first: $first, after: $after) {
-            edges {
-              node {
-                ref
-                quantity
-                productRef
-                locationLink { ref name }
-                updatedOn
-              }
-              cursor
-            }
-            pageInfo {
-              hasNextPage
-              # Note: Fluent doesn't return endCursor - cursors are in edges[].cursor
-            }
-          }
-        }
-      `,
-      variables: { retailerId: env.FLUENT_RETAILER_ID },
-      extractionMode: 'incremental',
-      stateAdapter: kv,
-      stateKey: 'hourly-extraction',
-      fallbackStartDate: '2025-01-01T00:00:00Z',
-      pagination: { pageSize: 200, maxRecords: 10000 },
-      outputFormat: 'csv',
-      outputDestination: {
-        type: 's3',
-        bucket: env.S3_BUCKET,
-        key: `virtual-positions/{{date}}/{{timestamp}}.csv`,
-        config: {
-          accessKeyId: env.AWS_ACCESS_KEY_ID,
-          secretAccessKey: env.AWS_SECRET_ACCESS_KEY,
-          region: env.AWS_REGION,
-        },
-      },
-    });
-
-    log.info('✅ Extraction complete', {
-      success: result.success,
-      recordCount: result.recordCount,
-      outputFile: result.outputFile,
-      duration: Date.now() - startTime
-    });
-
-    return {
-      success: result.success,
-      recordCount: result.recordCount,
-      outputFile: result.outputFile,
-      duration: Date.now() - startTime
-    };
-  })
-);
-
-// Export with MemoryInterpreter for Versori platform
-export const interpreter = new MemoryInterpreter({
-  workflows: [hourlyExtraction]
-});
-```
-
-### When to Use ExtractionOrchestrator
-
-- **✅ Use for**: New extraction workflows, scheduled extractions, standard use cases
-- **⚠️ Manual approach**: Complex transformations, custom business logic, non-standard outputs
-
-See [ExtractionOrchestrator API Reference](../../../../02-CORE-GUIDES/extraction/modules/02-core-guides-extraction-08-extraction-orchestrator.md) for complete documentation.
-
----
-
-## See Also
-
-- [CLI Validation Workflow](../../../../02-CORE-GUIDES/api-reference/modules/api-reference-11-cli-tools.md) - Validate queries and mappings
-- [Production Safety Guide](../../../../02-CORE-GUIDES/ingestion/modules/02-core-guides-ingestion-09-best-practices.md) - General safety practices
-- [GraphQL Query Examples](./graphql-queries/) - Sample extraction queries
-- [Universal Mapping Guide](../../../../02-CORE-GUIDES/advanced-services/advanced-services-readme.md) - Field mapping documentation
-- [ExtractionOrchestrator Examples](../../../../03-PATTERN-GUIDES/examples/test-data/03-PATTERN-GUIDES-readme.md) - Complete working examples
+# Extraction Modes Guide
+
+**FC Connect SDK - Choosing the Right Extraction Mode**
+
+> This guide explains the two extraction modes available in the SDK and when to use each one.
+
+---
+
+## Overview
+
+The FC Connect SDK supports two extraction modes for GraphQL queries:
+
+1. **Incremental** - Extract only changed records since last run (recommended for scheduled workflows)
+2. **Date Range** - Extract records within specific date window (for ad-hoc queries, backfills, or historical data)
+
+**Note:** "Historical extraction" is simply Date Range mode with a very old start date (e.g., `updatedAfter: '1970-01-01'`). There is no separate "historical mode" - use Date Range with appropriate validation and safeguards.
+
+---
+
+## Quick Comparison Table
+
+| Mode            | Use Case                                      | Safety Level | Recommended Frequency | Typical Records | Production Risk                     |
+| --------------- | --------------------------------------------- | ------------ | --------------------- | --------------- | ----------------------------------- |
+| **Incremental** | Scheduled syncs, recurring extractions        | ✅ **Safe**  | Hourly/Daily          | ~10k            | ✅ **Low** - Safe for production    |
+| **Date Range**  | Ad-hoc audits, backfills, historical dumps    | ⚠️ **Risky** | One-time only         | ~50k            | ⚠️ **Medium** - Requires validation |
+
+---
+
+## Mode 1: Incremental (Recommended) ✅
+
+### What It Does
+
+Extracts only records updated since the last successful extraction run.
+
+### How It Works
+
+```typescript
+const startTime = Date.now();
+log.info('📦 Starting incremental extraction');
+
+// Load last run timestamp from state
+const lastRunTime = await kv.get(['extraction', 'products', 'lastRunTime']);
+log.info('⏱️ Last run timestamp loaded', { lastRunTime });
+
+// Query with buffered timestamp (overlap buffer prevents gaps)
+const bufferedTime = new Date(lastRunTime - 60000).toISOString(); // 60 second buffer
+
+const result = await client.graphql({
+  query: PRODUCTS_QUERY,
+  variables: {
+    retailerId: 'my-retailer',
+    updatedAfter: bufferedTime, // Only changed records
+    first: 200,
+  },
+  pagination: { maxRecords: 10000 },
+});
+
+log.info('✅ Extraction complete', {
+  recordCount: result.edges.length,
+  duration: Date.now() - startTime
+});
+
+// Save new timestamp after successful extraction
+await kv.set(['extraction', 'products', 'lastRunTime'], {
+  timestamp: new Date().toISOString(),
+});
+```
+
+### When to Use
+
+- ✅ **Scheduled extractions** (hourly, daily, every 15 minutes)
+- ✅ **Real-time inventory feeds**
+- ✅ **Order status updates**
+- ✅ **Product catalog syncs**
+- ✅ **Any recurring extraction workflow**
+
+### Safety Features
+
+- Natural rate limiting via timestamps
+- Overlap buffer prevents missed records
+- State tracking prevents reprocessing
+- Predictable record counts
+- Automatic recovery after failures
+
+### Recommended Settings
+
+| Entity                   | Frequency    | Max Records | Expected Volume    |
+| ------------------------ | ------------ | ----------- | ------------------ |
+| **Products**             | Daily        | 10,000      | 5k-10k updates/day |
+| **Orders**               | Hourly       | 5,000       | 100-500/hour       |
+| **Fulfillments**         | Hourly       | 5,000       | 100-500/hour       |
+| **Inventory Quantities** | Every 15 min | 10,000      | 500-2k per run     |
+| **Inventory Positions**  | Daily        | 20,000      | 10k-20k per day    |
+| **Virtual Positions**    | Hourly       | 10,000      | 1k-5k per hour     |
+
+### Configuration Example
+
+```json
+{
+  "extractionMode": "incremental",
+  "pageSize": 200,
+  "maxRecords": 10000,
+  "overlapBufferSeconds": 60,
+  "fallbackStartDate": "2024-01-01T00:00:00Z"
+}
+```
+
+---
+
+## Mode 2: Date Range (Use Carefully) ⚠️
+
+### What It Does
+
+Extracts all records updated within a specific date window.
+
+### How It Works
+
+```typescript
+const startTime = Date.now();
+log.info('📦 Starting date range extraction');
+
+// User provides specific date range
+const startDate = ctx.activation?.getVariable('startDate'); // "2025-01-01T00:00:00Z"
+const endDate = ctx.activation?.getVariable('endDate'); // "2025-01-31T23:59:59Z"
+
+log.info('⏱️ Date range configured', { startDate, endDate });
+
+const result = await client.graphql({
+  query: PRODUCTS_QUERY,
+  variables: {
+    retailerId: 'my-retailer',
+    updatedAfter: startDate,
+    updatedBefore: endDate, // Bounded range
+    first: 200,
+  },
+  pagination: { maxRecords: 50000 },
+});
+
+log.info('✅ Extraction complete', {
+  recordCount: result.edges.length,
+  duration: Date.now() - startTime
+});
+```
+
+### When to Use
+
+- ⚠️ **One-time data audits** (e.g., "show me all orders from Q4 2024")
+- ⚠️ **Backfilling missed extractions** (e.g., extraction failed for 3 days)
+- ⚠️ **Historical reporting** (e.g., "export last month's inventory changes")
+- ⚠️ **Data quality investigations**
+
+### Risks
+
+- Can return tens of thousands of records
+- No natural rate limiting
+- Easy to exceed platform limits
+- Can timeout on large datasets
+- Risk of platform overload
+
+### Safety Guardrails Required
+
+```typescript
+// ✅ REQUIRED: Validate date range before extraction
+function validateDateRange(startDate: string, endDate: string, log: any) {
+  const start = new Date(startDate);
+  const end = new Date(endDate);
+  const daysDiff = (end.getTime() - start.getTime()) / (1000 * 60 * 60 * 24);
+
+  log.info('🔍 Validating date range', { startDate, endDate, daysDiff });
+
+  // Enforce maximum date range
+  const MAX_DAYS = 30;
+  if (daysDiff > MAX_DAYS) {
+    log.error('❌ Date range validation failed', {
+      daysDiff,
+      maxDays: MAX_DAYS,
+      recommendation: 'Split into smaller ranges or use incremental mode'
+    });
+    throw new Error(
+      `Date range too large: ${daysDiff} days (max: ${MAX_DAYS}). ` +
+        `Split into smaller ranges or use incremental mode.`
+    );
+  }
+
+  log.info('✅ Date range validation passed', { daysDiff, maxDays: MAX_DAYS });
+  return { daysDiff, isValid: true };
+}
+```
+
+### Recommended Limits
+
+| Entity           | Max Date Range | Max Records | File Splitting    |
+| ---------------- | -------------- | ----------- | ----------------- |
+| **Products**     | 30 days        | 50,000      | Required if > 25k |
+| **Orders**       | 7 days         | 20,000      | Required if > 10k |
+| **Fulfillments** | 7 days         | 20,000      | Required if > 10k |
+| **Inventory**    | 14 days        | 50,000      | Required if > 25k |
+
+### Configuration Example
+
+```json
+{
+  "extractionMode": "dateRange",
+  "startDate": "2025-01-01T00:00:00Z",
+  "endDate": "2025-01-31T23:59:59Z",
+  "pageSize": 200,
+  "maxRecords": 50000,
+  "validateDateRange": true,
+  "maxDaysAllowed": 30
+}
+```
+
+### Production Restrictions
+
+- ❌ **Never schedule** date range extractions
+- ❌ **Never use in recurring workflows**
+- ✅ **Always validate date range** before execution
+- ✅ **Always monitor execution time** and record counts
+- ✅ **Always implement file splitting** for large results
+- ✅ **Always get approval** before running on production data
+
+---
+
+## File Splitting for Large Extractions
+
+### Overview
+
+When ExtractionOrchestrator fetches large datasets (e.g., 50,000 records), all records are loaded into memory. Before writing to destination (SFTP/S3), you can split these records into multiple smaller files for better manageability, parallel processing, and downstream system compatibility.
+
+### When to Split Files
+
+- ✅ **Extraction returns >10k records** - Split into manageable chunks
+- ✅ **Downstream system has file size limits** - Respect partner constraints
+- ✅ **Parallel processing desired** - Write multiple files concurrently
+- ✅ **Network reliability concerns** - Smaller files = better retry granularity
+- ✅ **Audit trail requirements** - Track processing per file
+
+### Basic File Splitting Pattern
+
+```typescript
+import { Buffer } from 'node:buffer';
+import { ExtractionOrchestrator, SftpDataSource } from '@fluentcommerce/fc-connect-sdk';
+
+const startTime = Date.now();
+log.info('📦 Starting extraction with file splitting');
+
+// STEP 1: Extract all records (loads into memory)
+const result = await orchestrator.extract({
+  query: virtualPositionsQuery,
+  variables: { retailerId: 'my-retailer', updatedAfter: lastRunTime },
+  pagination: { pageSize: 200, maxRecords: 50000 },
+});
+
+log.info('✅ Extraction complete', {
+  recordCount: result.data.length,
+  duration: Date.now() - startTime
+});
+
+// result.data contains all 50,000 records in memory
+
+// STEP 2: Configure chunking
+const RECORDS_PER_FILE = 1000; // Configurable chunk size
+const timestamp = new Date().toISOString().replace(/[:.]/g, '-');
+
+// STEP 3: Split records into chunks
+const chunks: any[][] = [];
+for (let i = 0; i < result.data.length; i += RECORDS_PER_FILE) {
+  chunks.push(result.data.slice(i, i + RECORDS_PER_FILE));
+}
+
+log.info('📂 File splitting configured', {
+  totalRecords: result.data.length,
+  recordsPerFile: RECORDS_PER_FILE,
+  filesToCreate: chunks.length,
+});
+
+// STEP 4: Generate files with configurable naming pattern
+const filePromises = chunks.map((chunk, index) => {
+  // File naming pattern: entity-timestamp-part-NNN.format
+  const partNumber = String(index + 1).padStart(3, '0'); // 001, 002, 003...
+  const filename = `virtual-positions-${timestamp}-part-${partNumber}.xml`;
+
+  // Transform chunk to desired format (XML, CSV, JSON)
+  const xmlContent = xmlBuilder.build({ records: chunk });
+
+  // Return upload promise
+  return sftp.uploadFile(
+    `/outbound/${filename}`,
+    Buffer.from(xmlContent, 'utf8'),
+    { encoding: 'utf8', overwrite: false }
+  );
+});
+
+// STEP 5: Write all files in parallel
+const uploadStartTime = Date.now();
+await Promise.all(filePromises);
+
+log.info('✅ File splitting complete', {
+  filesCreated: chunks.length,
+  totalRecords: result.data.length,
+  recordsPerFile: RECORDS_PER_FILE,
+  uploadDuration: Date.now() - uploadStartTime,
+  totalDuration: Date.now() - startTime
+});
+```
+
+### Configurable File Naming Patterns
+
+Support multiple naming patterns via configuration:
+
+```typescript
+// Configuration interface
+interface FileSplittingConfig {
+  enabled: boolean;
+  recordsPerFile: number;
+  namingPattern: 'sequential' | 'timestamp' | 'range' | 'custom';
+  customPattern?: (index: number, chunk: any[], total: number) => string;
+}
+
+// Pattern examples
+const patterns = {
+  // Pattern 1: Sequential numbering
+  sequential: (index: number) =>
+    `virtual-positions-part-${String(index + 1).padStart(3, '0')}.xml`,
+
+  // Pattern 2: Timestamp-based
+  timestamp: (index: number) =>
+    `virtual-positions-${new Date().toISOString()}-${index + 1}.xml`,
+
+  // Pattern 3: Record range
+  range: (index: number, chunk: any[]) => {
+    const start = index * chunk.length + 1;
+    const end = start + chunk.length - 1;
+    return `virtual-positions-records-${start}-${end}.xml`;
+  },
+
+  // Pattern 4: Custom with metadata
+  custom: (index: number, chunk: any[], totalChunks: number) =>
+    `VP_${new Date().toISOString().split('T')[0]}_${index + 1}_of_${totalChunks}.xml`,
+};
+```
+
+### Complete Example: Extraction with File Splitting
+
+```typescript
+import { schedule, fn, MemoryInterpreter } from '@versori/run';
+import { Buffer } from 'node:buffer';
+import {
+  ExtractionOrchestrator,
+  createClient,
+  SftpDataSource,
+  VersoriKVAdapter,
+  XMLBuilder,
+} from '@fluentcommerce/fc-connect-sdk';
+
+export const virtualPositionsExtraction = schedule(
+  'virtual-positions-hourly',
+  '0 * * * *'
+).then(
+  fn('extract', async (ctx) => {
+    const { log, activation, openKv } = ctx;
+    const startTime = Date.now();
+
+    log.info('📦 Starting virtual positions extraction');
+
+    // Configuration from activation variables
+    const FILE_SPLITTING_ENABLED = activation.getVariable('fileSplittingEnabled') !== 'false';
+    const RECORDS_PER_FILE = parseInt(activation.getVariable('recordsPerFile') || '1000', 10);
+    const FILE_NAMING_PATTERN = activation.getVariable('fileNamingPattern') || 'sequential';
+
+    try {
+      // Initialize services
+      const client = await createClient(ctx);
+      log.info('✅ Client initialized');
+
+      const kv = new VersoriKVAdapter(openKv(':project:'));
+      const orchestrator = new ExtractionOrchestrator(client, log);
+
+      // Extract records (all loaded into memory)
+      const extractionStartTime = Date.now();
+      const result = await orchestrator.extract({
+        query: virtualPositionsQuery,
+        variables: {
+          retailerId: activation.getVariable('fluentRetailerId'),
+          updatedAfter: await kv.get(['extraction', 'lastRunTime']) || '2025-01-01T00:00:00Z',
+        },
+        pagination: { pageSize: 200, maxRecords: 50000 },
+      });
+
+      log.info('✅ Extraction complete', {
+        recordCount: result.data.length,
+        fileSplittingEnabled: FILE_SPLITTING_ENABLED,
+        extractionDuration: Date.now() - extractionStartTime
+      });
+
+      // Initialize SFTP
+      const sftp = new SftpDataSource({
+        type: 'SFTP_XML',
+        connectionId: 'sftp-extractions',
+        name: 'extraction-sftp',
+        settings: {
+          host: activation.getVariable('sftpHost'),
+          port: parseInt(activation.getVariable('sftpPort') || '22', 10),
+          username: activation.getVariable('sftpUsername'),
+          password: activation.getVariable('sftpPassword'),
+          remotePath: '/outbound/',
+          requireAbsolutePaths: true,
+        },
+      }, log);
+
+      await sftp.validateConnection();
+      log.info('✅ SFTP connection validated');
+
+      // File splitting logic
+      if (FILE_SPLITTING_ENABLED && result.data.length > RECORDS_PER_FILE) {
+        const splittingStartTime = Date.now();
+
+        // Split into chunks
+        const chunks: any[][] = [];
+        for (let i = 0; i < result.data.length; i += RECORDS_PER_FILE) {
+          chunks.push(result.data.slice(i, i + RECORDS_PER_FILE));
+        }
+
+        log.info('📂 Splitting extraction into multiple files', {
+          totalRecords: result.data.length,
+          recordsPerFile: RECORDS_PER_FILE,
+          filesToCreate: chunks.length,
+        });
+
+        // Generate files
+        const timestamp = new Date().toISOString().replace(/[:.]/g, '-');
+        const xmlBuilder = new XMLBuilder();
+
+        const filePromises = chunks.map(async (chunk, index) => {
+          // Generate filename based on pattern
+          let filename: string;
+          const partNumber = String(index + 1).padStart(3, '0');
+
+          switch (FILE_NAMING_PATTERN) {
+            case 'timestamp':
+              filename = `virtual-positions-${timestamp}-${partNumber}.xml`;
+              break;
+            case 'range':
+              const start = index * RECORDS_PER_FILE + 1;
+              const end = Math.min(start + chunk.length - 1, result.data.length);
+              filename = `virtual-positions-records-${start}-${end}.xml`;
+              break;
+            case 'sequential':
+            default:
+              filename = `virtual-positions-part-${partNumber}.xml`;
+              break;
+          }
+
+          // Build XML content
+          const xmlContent = xmlBuilder.build({ virtualPositions: chunk });
+
+          // Upload to SFTP
+          await sftp.uploadFile(
+            `/outbound/${filename}`,
+            Buffer.from(xmlContent, 'utf8'),
+            { encoding: 'utf8', overwrite: false }
+          );
+
+          return { filename, recordCount: chunk.length };
+        });
+
+        // Write all files in parallel
+        const uploadedFiles = await Promise.all(filePromises);
+
+        log.info('✅ File splitting complete', {
+          filesCreated: uploadedFiles.length,
+          totalRecords: result.data.length,
+          files: uploadedFiles,
+          splittingDuration: Date.now() - splittingStartTime,
+          totalDuration: Date.now() - startTime
+        });
+
+        return {
+          success: true,
+          extractionMode: 'incremental',
+          totalRecords: result.data.length,
+          filesCreated: uploadedFiles.length,
+          files: uploadedFiles,
+          duration: Date.now() - startTime
+        };
+      } else {
+        // Single file (no splitting)
+        const timestamp = new Date().toISOString().replace(/[:.]/g, '-');
+        const filename = `virtual-positions-${timestamp}.xml`;
+        const xmlBuilder = new XMLBuilder();
+        const xmlContent = xmlBuilder.build({ virtualPositions: result.data });
+
+        await sftp.uploadFile(
+          `/outbound/${filename}`,
+          Buffer.from(xmlContent, 'utf8'),
+          { encoding: 'utf8', overwrite: false }
+        );
+
+        log.info('✅ Single file extraction complete', {
+          filename,
+          recordCount: result.data.length,
+          duration: Date.now() - startTime
+        });
+
+        return {
+          success: true,
+          extractionMode: 'incremental',
+          totalRecords: result.data.length,
+          filesCreated: 1,
+          files: [{ filename, recordCount: result.data.length }],
+          duration: Date.now() - startTime
+        };
+      }
+    } catch (error: any) {
+      log.error('❌ Extraction failed', {
+        message: error instanceof Error ? error.message : String(error),
+        stack: error instanceof Error ? error.stack : undefined,
+        errorType: error instanceof Error ? error.constructor.name : 'Error',
+        duration: Date.now() - startTime
+      });
+
+      return {
+        success: false,
+        error: error.message,
+        duration: Date.now() - startTime
+      };
+    } finally {
+      if (sftp) {
+        await sftp.dispose();
+        log.info('✅ SFTP connection disposed');
+      }
+    }
+  })
+);
+
+// Export with MemoryInterpreter for Versori platform
+export const interpreter = new MemoryInterpreter({
+  workflows: [virtualPositionsExtraction]
+});
+```
+
+### Date Range Splitting for Very Large Datasets
+
+For extractions spanning multiple years (e.g., historical dumps), split the date range and run multiple smaller extractions:
+
+```typescript
+// Split large date range into manageable chunks
+async function extractHistoricalData(
+  startDate: string,  // "2020-01-01"
+  endDate: string,    // "2025-01-01"
+  chunkMonths: number = 1,  // Extract 1 month at a time
+  log: any
+) {
+  const overallStartTime = Date.now();
+  const results = [];
+  let currentStart = new Date(startDate);
+  const finalEnd = new Date(endDate);
+
+  log.info('📦 Starting historical data extraction', {
+    startDate,
+    endDate,
+    chunkMonths,
+    estimatedChunks: Math.ceil((finalEnd.getTime() - currentStart.getTime()) / (chunkMonths * 30 * 24 * 60 * 60 * 1000))
+  });
+
+  while (currentStart < finalEnd) {
+    const chunkStartTime = Date.now();
+
+    // Calculate chunk end date
+    const currentEnd = new Date(currentStart);
+    currentEnd.setMonth(currentEnd.getMonth() + chunkMonths);
+
+    // Ensure we don't exceed final end date
+    if (currentEnd > finalEnd) {
+      currentEnd.setTime(finalEnd.getTime());
+    }
+
+    log.info('📂 Extracting date range chunk', {
+      chunkNumber: results.length + 1,
+      startDate: currentStart.toISOString(),
+      endDate: currentEnd.toISOString(),
+    });
+
+    // Extract for this date range
+    const result = await orchestrator.extract({
+      query: virtualPositionsQuery,
+      variables: {
+        retailerId: 'my-retailer',
+        updatedAfter: currentStart.toISOString(),
+        updatedBefore: currentEnd.toISOString(),
+      },
+      pagination: { pageSize: 200, maxRecords: 50000 },
+    });
+
+    log.info('✅ Chunk extraction complete', {
+      chunkNumber: results.length + 1,
+      recordCount: result.data.length,
+      chunkDuration: Date.now() - chunkStartTime
+    });
+
+    // Split this chunk into files if needed
+    if (result.data.length > RECORDS_PER_FILE) {
+      const chunks: any[][] = [];
+      for (let i = 0; i < result.data.length; i += RECORDS_PER_FILE) {
+        chunks.push(result.data.slice(i, i + RECORDS_PER_FILE));
+      }
+
+      // Write chunks in parallel
+      const timestamp = currentStart.toISOString().split('T')[0];
+      await Promise.all(
+        chunks.map((chunk, index) => {
+          const filename = `historical-${timestamp}-part-${String(index + 1).padStart(3, '0')}.xml`;
+          return sftp.uploadFile(`/outbound/${filename}`, Buffer.from(xmlBuilder.build(chunk), 'utf8'));
+        })
+      );
+    }
+
+    results.push({
+      startDate: currentStart.toISOString(),
+      endDate: currentEnd.toISOString(),
+      recordCount: result.data.length,
+      duration: Date.now() - chunkStartTime
+    });
+
+    // Move to next chunk
+    currentStart = new Date(currentEnd);
+
+    // Rate limiting - wait between chunks
+    log.info('⏱️ Rate limiting delay (5 seconds)');
+    await new Promise(resolve => setTimeout(resolve, 5000)); // 5 second delay
+  }
+
+  const totalRecords = results.reduce((sum, r) => sum + r.recordCount, 0);
+  const totalDuration = Date.now() - overallStartTime;
+
+  log.info('✅ Historical extraction complete', {
+    totalChunks: results.length,
+    totalRecords,
+    totalDuration,
+    averageRecordsPerChunk: Math.round(totalRecords / results.length)
+  });
+
+  return results;
+}
+
+// Usage: Extract 5 years of data, 1 month at a time
+const historicalResults = await extractHistoricalData('2020-01-01', '2025-01-01', 1, log);
+```
+
+### Configuration Variables for File Splitting
+
+Add these to Versori activation variables:
+
+```bash
+# File Splitting Configuration
+FILE_SPLITTING_ENABLED=true              # Enable/disable file splitting
+RECORDS_PER_FILE=1000                    # Records per file (default: 1000)
+FILE_NAMING_PATTERN=sequential           # Naming pattern: sequential | timestamp | range
+MAX_RECORDS_PER_EXTRACTION=50000         # Maximum records to extract in one run
+
+# Date Range Splitting (for large historical extractions)
+DATE_RANGE_CHUNK_MONTHS=1                # Months per extraction chunk (default: 1)
+RATE_LIMIT_DELAY_MS=5000                 # Delay between chunks (milliseconds)
+```
+
+### Benefits of File Splitting
+
+- ✅ **Memory efficiency** - Process large datasets without memory overflow
+- ✅ **Parallel writes** - Multiple files written concurrently (faster)
+- ✅ **Better error recovery** - Retry individual files vs entire extraction
+- ✅ **Downstream compatibility** - Honor partner file size limits
+- ✅ **Audit granularity** - Track processing per file
+- ✅ **Network resilience** - Smaller files = better upload success rate
+
+### Limitations
+
+- All records must fit in memory during extraction (ExtractionOrchestrator loads all pages)
+- File splitting happens post-extraction (not during pagination)
+- Parallel writes limited by available memory and network bandwidth
+- SFTP connection pool size may limit concurrency (default: 10 connections)
+
+---
+
+## Historical Data Extraction via Date Range
+
+### Overview
+
+There is **no separate "historical mode"** in the SDK. To extract historical data (e.g., all records from 2020 onwards), use **Date Range mode** with a very old start date and appropriate safeguards.
+
+### How to Extract Historical Data
+
+```typescript
+// Historical extraction is just Date Range with old start date
+await client.graphql({
+  query: PRODUCTS_QUERY,
+  variables: {
+    retailerId: 'my-retailer',
+    updatedAfter: '1970-01-01T00:00:00Z',  // Very old date = "all records"
+    updatedBefore: new Date().toISOString(), // Up to now
+    first: 200,
+  },
+  pagination: { maxRecords: 50000 },
+});
+```
+
+### Required Safeguards for Historical Extraction
+
+1. **Date Range Splitting** - Split into smaller chunks (monthly/quarterly)
+2. **File Splitting** - Split large results into multiple files
+3. **Rate Limiting** - Add delays between chunks
+4. **Validation** - Verify date ranges before execution
+5. **Monitoring** - Track progress and alert on anomalies
+6. **Approval** - Get sign-off before running on production
+
+### Recommended Approach: Chunked Date Range Extraction
+
+```bash
+#!/bin/bash
+# Safe historical extraction via chunked date ranges
+
+START_DATE="2020-01-01"
+END_DATE="2025-01-01"
+
+# Extract one month at a time
+current=$START_DATE
+while [[ "$current" < "$END_DATE" ]]; do
+  # Calculate month end
+  monthEnd=$(date -d "$current + 1 month" +%Y-%m-%d)
+
+  echo "Extracting $current to $monthEnd..."
+
+  # Trigger date range extraction
+  curl -X POST https://versori-webhook.com/extract \
+    -H "Content-Type: application/json" \
+    -d "{
+      \"extractionMode\": \"dateRange\",
+      \"startDate\": \"${current}T00:00:00Z\",
+      \"endDate\": \"${monthEnd}T00:00:00Z\",
+      \"fileSplittingEnabled\": true,
+      \"recordsPerFile\": 1000
+    }"
+
+  # Rate limiting - wait 60 seconds between chunks
+  sleep 60
+
+  # Move to next month
+  current=$monthEnd
+done
+
+echo "Historical extraction complete"
+```
+
+### Migration to Incremental After Historical Load
+
+After completing a one-time historical extraction, switch to incremental mode for ongoing syncs:
+
+```json
+{
+  "extractionMode": "incremental",
+  "fallbackStartDate": "2025-01-22T00:00:00Z",
+  "pageSize": 200,
+  "maxRecords": 10000
+}
+```
+
+---
+
+## Decision Tree: Which Mode to Use?
+
+```
+Start Here
+    │
+    ├─ Need recurring extractions? ─────────► Use INCREMENTAL ✅
+    │                                          (hourly/daily/every 15 min)
+    │                                          Tracks state, auto-recovery
+    │
+    └─ Need specific date range? ────────────► Use DATE RANGE ⚠️
+         │                                      (one-time, validate range)
+         │
+         ├─ Date range < 30 days? ──────────► Single DATE RANGE run
+         │                                     + file splitting if >10k records
+         │
+         └─ Date range > 30 days? ──────────► Split into monthly chunks
+              │                               + file splitting per chunk
+              │                               + rate limiting between chunks
+              │
+              └─ Historical (all data)? ────► DATE RANGE with old start date
+                                              + chunked approach (monthly)
+                                              + approval required
+```
+
+---
+
+## Monitoring & Alerts
+
+Set up alerts for extraction volumes:
+
+```typescript
+// In extraction workflow
+const recordCount = edges.length;
+const ALERT_THRESHOLD = 50000;
+
+log.info('📊 Checking extraction volume', { recordCount, threshold: ALERT_THRESHOLD });
+
+if (recordCount > ALERT_THRESHOLD) {
+  log.error('❌ Extraction volume exceeded threshold', {
+    recordCount,
+    threshold: ALERT_THRESHOLD,
+    percentageOver: Math.round(((recordCount - ALERT_THRESHOLD) / ALERT_THRESHOLD) * 100),
+    mode: extractionMode,
+    recommendation: 'Switch to incremental mode or reduce date range',
+  });
+
+  // Send alert to monitoring system
+  await sendAlert({
+    severity: 'high',
+    message: `Extraction returned ${recordCount} records (threshold: ${ALERT_THRESHOLD})`,
+    mode: extractionMode,
+  });
+
+  log.info('✅ Alert sent to monitoring system');
+} else {
+  log.info('✅ Extraction volume within acceptable limits', {
+    recordCount,
+    threshold: ALERT_THRESHOLD,
+    percentageUsed: Math.round((recordCount / ALERT_THRESHOLD) * 100)
+  });
+}
+```
+
+---
+
+## Summary & Best Practices
+
+### ✅ DO
+
+- Use **incremental mode** for all scheduled extractions
+- Validate date ranges before running **dateRange mode**
+- Implement **file splitting** for large extractions (>10k records)
+- Use **parallel writes** with `Promise.all()` for split files
+- Monitor extraction volumes and set alerts
+- Test on staging before production
+- Use overlap buffer (60s) to prevent gaps in incremental mode
+- Track state with VersoriKV
+- Split large historical extractions into monthly chunks
+- Add rate limiting between extraction chunks
+
+### ❌ DON'T
+
+- Schedule **dateRange** extractions (incremental only)
+- Use date ranges > 30 days without chunking
+- Skip validation checks for date ranges
+- Ignore volume alerts (>50k records)
+- Forget to implement file splitting for large results
+- Run historical extractions without approval and monitoring
+
+---
+
+## ExtractionOrchestrator (SDK v0.1.27+)
+
+The SDK includes **ExtractionOrchestrator** - a high-level service that simplifies extraction workflows with built-in mode handling, pagination, and output management.
+
+### Why Use ExtractionOrchestrator?
+
+**Instead of manually implementing:**
+
+- Mode detection (incremental/dateRange)
+- Pagination loops
+- Path-based field extraction
+- Output formatting (CSV/JSON/Parquet)
+- S3/SFTP uploads
+- Error handling
+
+**ExtractionOrchestrator handles it all:**
+
+```typescript
+import { ExtractionOrchestrator, createClient } from '@fluentcommerce/fc-connect-sdk';
+
+const startTime = Date.now();
+log.info('📦 Initializing ExtractionOrchestrator');
+
+const client = await createClient(ctx);
+const orchestrator = new ExtractionOrchestrator(client, log);
+
+log.info('✅ Orchestrator initialized');
+
+// Both modes supported with single interface
+const result = await orchestrator.extract({
+  query: virtualPositionsQuery,
+  variables: { retailerId: 'my-retailer' },
+
+  // Mode: 'incremental' (scheduled) or 'dateRange' (ad-hoc)
+  extractionMode: 'incremental',
+  stateKey: 'virtual-positions-extraction',
+
+  // Pagination handled automatically
+  pagination: {
+    pageSize: 200,
+    maxRecords: 10000,
+  },
+
+  // Output format and destination
+  outputFormat: 'csv',
+  outputDestination: {
+    type: 's3',
+    bucket: 'my-extracts',
+    key: 'virtual-positions/hourly/{{timestamp}}.csv',
+  },
+
+  // Field extraction from nested paths
+  fieldPaths: {
+    position_ref: 'ref',
+    quantity: 'quantity',
+    location_ref: 'locationLink.ref',
+    location_name: 'locationLink.name',
+  },
+});
+
+log.info('✅ Extraction and upload complete', {
+  recordCount: result.recordCount,
+  outputFile: result.outputFile,
+  duration: Date.now() - startTime
+});
+```
+
+### Features
+
+- **Auto-pagination**: Handles cursor-based pagination automatically
+- **Mode support**: Both modes (incremental, dateRange)
+- **State management**: Tracks last run timestamps for incremental mode
+- **Path extraction**: Extracts nested fields from GraphQL responses
+- **Multi-format**: Outputs CSV, JSON, or Parquet
+- **Validation**: Built-in query and response validation
+- **Error recovery**: Graceful failure handling with detailed logs
+- **File splitting**: Post-extraction chunking for large datasets (add manually, see File Splitting section above)
+
+### Example: Incremental Extraction with ExtractionOrchestrator
+
+```typescript
+import { schedule, fn, MemoryInterpreter } from '@versori/run';
+import { Buffer } from 'node:buffer';  // Required for Deno/Versori runtime
+import {
+  ExtractionOrchestrator,
+  createClient,
+  VersoriKVAdapter,
+} from '@fluentcommerce/fc-connect-sdk';
+
+export const hourlyExtraction = schedule('hourly-virtual-positions', '0 * * * *').then(
+  fn('extract', async (ctx) => {
+    const { log, openKv, env } = ctx;
+    const startTime = Date.now();
+
+    log.info('📦 Starting hourly virtual positions extraction');
+
+    const client = await createClient(ctx);
+    log.info('✅ Client initialized');
+
+    const kv = new VersoriKVAdapter(openKv(':project:'));
+    const orchestrator = new ExtractionOrchestrator(client, log);
+
+    const result = await orchestrator.extract({
+      query: `
+        query GetVirtualPositions($retailerId: String!, $updatedAfter: String, $first: Int, $after: String) {
+          virtualPositions(retailerId: $retailerId, updatedAfter: $updatedAfter, first: $first, after: $after) {
+            edges {
+              node {
+                ref
+                quantity
+                productRef
+                locationLink { ref name }
+                updatedOn
+              }
+              cursor
+            }
+            pageInfo {
+              hasNextPage
+              # Note: Fluent doesn't return endCursor - cursors are in edges[].cursor
+            }
+          }
+        }
+      `,
+      variables: { retailerId: env.FLUENT_RETAILER_ID },
+      extractionMode: 'incremental',
+      stateAdapter: kv,
+      stateKey: 'hourly-extraction',
+      fallbackStartDate: '2025-01-01T00:00:00Z',
+      pagination: { pageSize: 200, maxRecords: 10000 },
+      outputFormat: 'csv',
+      outputDestination: {
+        type: 's3',
+        bucket: env.S3_BUCKET,
+        key: `virtual-positions/{{date}}/{{timestamp}}.csv`,
+        config: {
+          accessKeyId: env.AWS_ACCESS_KEY_ID,
+          secretAccessKey: env.AWS_SECRET_ACCESS_KEY,
+          region: env.AWS_REGION,
+        },
+      },
+    });
+
+    log.info('✅ Extraction complete', {
+      success: result.success,
+      recordCount: result.recordCount,
+      outputFile: result.outputFile,
+      duration: Date.now() - startTime
+    });
+
+    return {
+      success: result.success,
+      recordCount: result.recordCount,
+      outputFile: result.outputFile,
+      duration: Date.now() - startTime
+    };
+  })
+);
+
+// Export with MemoryInterpreter for Versori platform
+export const interpreter = new MemoryInterpreter({
+  workflows: [hourlyExtraction]
+});
+```
+
+### When to Use ExtractionOrchestrator
+
+- **✅ Use for**: New extraction workflows, scheduled extractions, standard use cases
+- **⚠️ Manual approach**: Complex transformations, custom business logic, non-standard outputs
+
+See [ExtractionOrchestrator API Reference](../../../../02-CORE-GUIDES/extraction/modules/02-core-guides-extraction-08-extraction-orchestrator.md) for complete documentation.
+
+---
+
+## See Also
+
+- [CLI Validation Workflow](../../../../02-CORE-GUIDES/api-reference/modules/api-reference-11-cli-tools.md) - Validate queries and mappings
+- [Production Safety Guide](../../../../02-CORE-GUIDES/ingestion/modules/02-core-guides-ingestion-09-best-practices.md) - General safety practices
+- [GraphQL Query Examples](./graphql-queries/) - Sample extraction queries
+- [Universal Mapping Guide](../../../../02-CORE-GUIDES/advanced-services/advanced-services-readme.md) - Field mapping documentation
+- [ExtractionOrchestrator Examples](../../../../03-PATTERN-GUIDES/examples/test-data/03-PATTERN-GUIDES-readme.md) - Complete working examples