ETL CoreStream

Reactive stream-based ETL engine for massive data workflows

Process gigabyte-scale files without freezing the UI, exhausting memory, or blocking user interaction.

Built for resumable imports, async validation pipelines, real-time editing, and fully modular orchestration.

⚠️ Disclaimer — under active development

ETL CoreStream is currently under heavy development and is not stable yet. Until version 1.0.0, APIs, behaviors, and internal architecture may change without notice.

Expect breaking changes while the project evolves toward a stable v1 release.

📚 Table of Contents

• Features
• Why ETL CoreStream?
• Installation
• Quick Start
• Layout-Driven Architecture
• Local Transform Pipelines
• Local Validation Pipelines
• Async Global Validation Pipelines
• Async Global Transform Pipelines
• Persistent & Resumable Workflows
• Export System
• Processing Pipeline
• Architecture Overview
• Fully Modular Architecture
• Reactive by Design
• Performance Characteristics
• The 1GB Challenge
• Capability Comparison
• Deployment Flexibility
• Documentation & Examples
• Official Adapters
• Use Cases
• Design Philosophy
• Open Source
• Current Status
• License

✨ Core Features

• Stream-first architecture
• Constant memory footprint
• Massive file processing
• Resumable ETL sessions
• Real-time editable datasets
• Persistent recoverable workflows
• Async backend-aware validation pipelines
• Async transforms & mapping
• Layout-driven processing
• Header remapping & aliases
• Headless & framework agnostic
• Signals + RxJS reactive state
• Fully modular dependency injection
• Replaceable orchestrator architecture
• Custom importers/exporters
• Export streams support
• Browser-first architecture
• Open source & extensible

🚀 Why ETL CoreStream?

Traditional browser ETL solutions usually:

• Load entire files into memory
• Freeze the UI during processing
• Lose progress after refreshes/crashes
• Lack async validation pipelines
• Become unusable with large datasets
• Are tightly coupled to specific frameworks
• Cannot recover interrupted workflows

ETL CoreStream solves this with a reactive stream-first architecture designed for modern large-scale data workflows.

Users can:

• Import massive files while keeping the UI responsive
• Edit rows during processing
• Resume interrupted imports
• Persist datasets locally
• Revalidate incrementally
• Connect validators/transforms directly to backend services
• Export results to files, streams, APIs, or cloud providers
• Replace internal modules with custom implementations

🚀 Installation

npm install @etl-corestream/core

⚡ Quick Browser Implementation

ETL CoreStream ships with a browser-oriented preset architecture that provides a ready-to-use ETL pipeline using:

• PapaParse importer
• IndexedDB persistence
• Recovery engine
• Validation engines
• Export system
• Reactive viewer
• Stream processing pipeline

Quick Example

import { ETLBrowserOrchestrator } from "@etl-corestream/core/examples";
import { LayoutExample } from "@etl-corestream/core/examples";

// Minimal quickstart
const orchestrator = ETLBrowserOrchestrator();

orchestrator.selectLayout(LayoutExample);

await orchestrator.selectFile(file);

await orchestrator.export("Export Just Name and Email", "File");

This quick example shows the minimal flow to get started. See "Advanced Browser Configuration" for tuning.

Advanced Browser Configuration

import { ETLBrowserOrchestrator } from "@etl-corestream/core/examples";
import { LayoutExample } from "@etl-corestream/core/examples";

// Create orchestrator with advanced options
const orchestrator = ETLBrowserOrchestrator({
  importer: {
    worker: true,
    importerChunkSize: 1024 * 1024 * 5,
  },

  persistence: {
    chunkSizeQtd: 50,
  },

  recover: {
    checkRecoveryPoint: true,
  },
});

// Select layout
orchestrator.selectLayout(LayoutExample);

// Observe reactive state
orchestrator.state$.subscribe(console.log);

orchestrator.progress$.subscribe(console.log);

// Start processing
await orchestrator.selectFile(file);

// Edit rows while processing continues
orchestrator.editRow(rowId, "email", "[email protected]");

// Export anytime
await orchestrator.export("Export Just Name and Email", "File");

The entire pipeline remains reactive and non-blocking during processing.

🧠 Layout-Driven Architecture

ETL CoreStream uses layouts to define how files should be interpreted and processed.

Layouts define:

• Headers
• Header aliases
• Header remapping
• Required fields
• Local validators
• Local transforms
• Global validators
• Global transforms
• Export rules
• Entity creation logic

export const LayoutExample: LayoutBase = {
  id: "contact-management-layout-v1",

  name: "Contact Management Layout",

  description: "Example layout for processing contact information",

  allowUndefinedColumns: false,

  headers: [
    {
      key: "name",
      label: "Full Name",
      alternativeKeys: ["fullname", "nombre"],
      required: true,
    },

    {
      key: "email",
      label: "Email Address",
      alternativeKeys: ["correo", "contact_email"],
      required: true,
    },
  ],

  localSteps: [
    {
      id: "email-processing",

      name: "Email Processing",

      order: ["transforms", "validators"],

      transforms: [trim("email"), toLowerCase("email")],

      validators: [required("email"), email("email")],
    },
  ],

  globalSteps: [
    {
      name: "Global Validation",

      order: ["validators"],

      validators: [AsyncValidateDataExample()],
    },
  ],

  exports: [
    {
      name: "Export Just Name and Email",

      fn: (row) => ({
        name: row?.value?.name,
        email: row?.value?.email,
      }),
    },
  ],
};

This keeps ETL workflows declarative, reusable, and independent from parsing logic.

🎯 Use Cases

Common use cases where ETL CoreStream excels:

• Massive CSV imports
• CRM imports
• ERP migration pipelines
• Browser-side ETL apps
• Data cleaning interfaces
• AI enrichment pipelines
• Streaming exports
• Incremental validation systems
• Recoverable import workflows

🔄 Local Transform Pipelines

Local transforms run at row level and are ideal for normalization and lightweight transformations.

export const toLowerCase = (headerKey: string): LocalStepTransform => ({
  headerKey,

  name: "toLowerCase",

  fn: (value: string) => value.toLowerCase(),
});

export const trim = (headerKey: string): LocalStepTransform => ({
  headerKey,

  name: "trim",

  fn: (value: string) => value.trim(),
});

Perfect for:

• String normalization
• Cleanup
• Parsing
• Date formatting
• Entity preparation

✅ Local Validation Pipelines

Local validators run synchronously for immediate row-level feedback.

export const minValue = (headerKey: string, min: number): LocalStepValidator => ({
  headerKey,

  name: "Min Value",

  args: [min],

  fn: (value: string, row: any, minVal: number) => {
    const numValue = parseFloat(value);

    const isValid = numValue >= minVal;

    return {
      isValid,

      validationCode: "MIN_VALUE",

      message: !isValid ? `Value must be at least ${minVal}` : undefined,

      value,

      step: "local",
    };
  },
});

Ideal for:

• Required validations
• Numeric ranges
• Regex validation
• Formatting checks
• Immediate UI feedback

🌐 Async Global Validation Pipelines

Global validators can run asynchronously and integrate directly with APIs or backend services.

export const AsyncValidateDataExample = (): GlobalStepValidator => ({
  name: "AsyncValidateDataExample",

  fn: async (rows: RowObject[]) => {
    const validationResults = await validateDataExample(
      rows.map((row) => ({
        id: row.__rowId,
        value: row.value["headerKey"],
      }))
    );

    return {
      validationErrors: validationResults
        .filter((result) => !result.isValid)
        .map((result) => ({
          __rowId: result.id,

          headerKey: "headerKey",

          validationCode: result.validationCode,

          message: result.message,

          step: "AsyncValidateDataExample",
        })),

      removedValidationErrors: [],
    };
  },
});

This enables:

• Backend-aware validation
• Batch validation APIs
• Database checks
• Cross-row validation
• External business rule engines

🔁 Async Global Transform Pipelines

Global transforms allow asynchronous dataset-wide transformations.

export const AsyncTransformDataExample = (): GlobalStepTransform => ({
  name: "AsyncTransformDataExample",

  fn: async (rows: RowObject[]) => {
    const transformedItems = await transformDataExample(
      rows.map((row) => ({
        id: row.__rowId,
        value: row.value["headerKey"],
      }))
    );

    const rowMap = new Map(rows.map((r) => [r.__rowId, r]));

    transformedItems.forEach((item) => {
      const row = rowMap.get(item.id);

      if (row) {
        row.value["headerKey"] = item.value;
      }
    });
  },
});

Perfect for:

• Backend enrichment
• AI processing
• Data normalization
• Entity synchronization
• Cross-dataset transformations

💾 Persistent & Resumable Workflows

ETL CoreStream supports persistent ETL sessions.

Imports can continue after:

• Browser refreshes
• Crashes
• Tab closures
• Interrupted processing

Users can edit imported rows while processing continues in the background.

Persistence Capabilities

• Resume interrupted imports
• Restore previous sessions
• Persist partially processed datasets
• Incremental reprocessing
• Editable persisted data
• Recoverable ETL workflows

const orchestrator = ETLBrowserOrchestrator({
  recover: {
    checkRecoveryPoint: true,
  },

  persistence: {
    chunkSizeQtd: 50,
  },
});

Long-running imports become safe, recoverable, and interactive.

📤 Export System

ETL CoreStream exporters can export directly to:

• Files
• Streams
• APIs
• Cloud providers
• Custom destinations

The exporter system can also expose a ReadableStream directly, allowing you to consume transformed data without implementing a custom exporter.

await orchestrator.export("Export Just Name and Email", "Stream");

This makes it possible to:

• Pipe exports to APIs
• Upload directly to cloud storage
• Send data through WebSockets
• Stream data into another ETL pipeline
• Build custom exporters externally

🔁 Processing Pipeline

Typical processing flow inside ETL CoreStream:

File ↓ Importer ↓ Mapper ↓ Local Steps Engine (row-level transforms & validators) ↓ Persistence (chunked storage / indexedDB) ↓ Global Steps Engine (dataset-level transforms & validators) ↓ Exporter / Viewer

This diagram highlights the runtime flow from raw file input to export/view output and clarifies where modules can be swapped.

🏗️ Architecture Overview

                          ┌────────────────────┐
                          │      Provider      │
                          │ Dependency Injector│
                          └─────────┬──────────┘
                                    │
                    Injects replaceable modules
                                    │

┌───────────────────────────────────────────────────────────────┐
│                                                               │
│                        Orchestrator                           │
│                    (replaceable module)                      │
│                                                               │
└───────────────┬───────────────────────────────────────────────┘
                │

     ┌──────────┼──────────┬──────────┬──────────┬──────────┐
     ↓          ↓          ↓          ↓          ↓          ↓

┌────────┐ ┌────────┐ ┌────────┐ ┌────────┐ ┌────────┐ ┌────────┐
│Importer│ │ Mapper │ │Persist │ │Recover │ │ Viewer │ │ Logger │
│replace │ │replace │ │replace │ │replace │ │replace │ │replace │
└────┬───┘ └────┬───┘ └────┬───┘ └────┬───┘ └────┬───┘ └────────┘
     │           │          │          │          │
     └───────────┼──────────┼──────────┼──────────┘
                 │
                 ↓

    ┌───────────────────────────────────────────────┐
    │           Local Steps Engine                  │
    │    Validators + Transforms (row-level)       │
    └─────────────────┬─────────────────────────────┘
                      ↓

    ┌───────────────────────────────────────────────┐
    │          Global Steps Engine                  │
    │ Validators + Transforms (dataset-level)      │
    └─────────────────┬─────────────────────────────┘
                      ↓

               ┌──────────────┐
               │   Exporter   │
               │   replace    │
               └──────┬───────┘
                      ↓

           File / Stream / API / Cloud

🧩 Fully Modular Architecture

Every internal module can be replaced.

Including:

• Orchestrator
• Importer
• Mapper
• Persistence
• Recover module
• Viewer
• Logger
• Exporter
• Local Steps Engine
• Global Steps Engine The Provider is the composition root of the system. It is responsible for dependency injection, module registration, and orchestration wiring.

As long as interfaces are respected, any module can be replaced with a custom implementation.

const provider = new ProviderModule({
  importer: {
    module: CustomCSVImporter,
  },

  persistence: {
    module: PostgresPersistence,
  },

  exporter: {
    module: S3Exporter,
  },
});

const orchestrator = new CustomOrchestratorModule();

orchestrator.initialize(provider);

This allows developers to build entirely custom ETL ecosystems while preserving compatibility with the CoreStream pipeline architecture.

⚛️ Reactive by Design

ETL CoreStream exposes reactive state through:

• RxJS Observables
• Preact Signals

Compatible with:

• React
• Vue
• Angular
• Svelte
• SolidJS
• Vanilla JS

orchestrator.state$.subscribe(console.log);

orchestrator.progress$.subscribe(console.log);

orchestrator.metrics$.subscribe(console.log);

const state = orchestrator.state;

const metrics = orchestrator.metrics;

📊 Performance Characteristics

ETL CoreStream is designed for massive datasets and constrained environments.

Memory

• Constant memory footprint
• Chunk-based processing
• Stream pipelines
• Lazy pagination
• Minimal duplication

CPU

• Non-blocking processing
• UI-safe architecture
• Incremental updates
• Background pipelines

Storage

• Indexed persistence
• Recoverable sessions
• Incremental synchronization

Network

• Async backend integration
• Chunked transfers
• Cancellable operations

📈 The 1GB Challenge

While many browser ETL tools crash or freeze processing large datasets, ETL CoreStream keeps processing incrementally while maintaining responsive interaction.

Users can:

• Navigate pages
• Edit rows
• Trigger exports
• Apply filters
• Continue working

Even while background processing is still running.

Benchmark (example)

| Rows | File Size | Memory Usage | UI Freeze | | --------- | --------- | ------------ | --------- | | 1,000,000 | 1GB | ~constant | No |

📦 Capability Comparison

| Capability | ETL CoreStream | Traditional Browser ETL | | ------------------------ | -------------- | ----------------------- | | Stream processing | ✅ | ❌ | | Constant memory usage | ✅ | ❌ | | Resumable imports | ✅ | ❌ | | Editable datasets | ✅ | ❌ | | Async backend validation | ✅ | ⚠️ | | Reactive state | ✅ | ❌ | | Modular architecture | ✅ | ⚠️ | | Replaceable orchestrator | ✅ | ❌ | | Real-time revalidation | ✅ | ❌ | | Persistent sessions | ✅ | ❌ | | Stream exports | ✅ | ❌ |

🌍 Deployment Flexibility

ETL CoreStream is environment-agnostic.

Build adapters for:

• Browsers
• Node.js
• Deno
• Docker
• Kubernetes
• Edge Workers
• Microservices

One orchestration engine, multiple environments.

📚 Documentation & Examples

Detailed guides and examples are available in /docs.

Topics include:

• Creating layouts
• Header mapping
• Local validators
• Local transforms
• Global validators
• Global transforms
• Exporters
• Persistence engines
• Browser implementations
• Recovery systems
• Async pipelines
• Editing workflows
• Custom orchestrators
• Custom modules

Additional adapters and ecosystem integrations are maintained in separate repositories.

You can also find the full set of "how-to" guides in the repository docs folder on GitHub: ETLCoreStream-Core/docs

🔌 Official Adapters

The first official adapter is the React adapter:

• @etl-corestream/react — React integration (viewer components and helpers)

  • npm: https://www.npmjs.com/package/@etl-corestream/react
  • package: @etl-corestream/react
  • GitHub: https://github.com/cristianm-developer/ETLCoreStream-React

  Install:

  npm install @etl-corestream/react

🤝 Open Source

ETL CoreStream is fully open source.

You are free to:

• Fork
• Extend
• Create adapters
• Contribute
• Open issues
• Improve documentation
• Build integrations

Community contributions are welcome.

🧭 Design Philosophy

ETL CoreStream follows a small set of guiding principles:

• Streams over in-memory datasets
• Recoverability over ephemeral processing
• Composition over monolithic pipelines
• Reactive state over polling
• Incremental processing over blocking operations

These principles shape API decisions and architecture trade-offs across the project.

🛠️ Current Status

ETL CoreStream is evolving rapidly toward a stable v1 release.

Current priorities include:

• API stabilization
• Adapter ecosystem expansion
• Persistence optimization
• Additional exporters
• Stream optimizations
• Documentation expansion
• More recovery strategies
• Additional environment presets