Software

Core Role

Unifies the data integration structure between heterogeneous systems and databases into a standard REST API based on an in-memory hub (Redis) and handles real-time synchronization.

Key Features & Specifications

• Supports integration with 7 DBMSs: Oracle, Tibero, MSSQL, MySQL, MariaDB, PostgreSQL, and CUBRID.
• Provides 3 synchronization modes: Poller (DB → Redis), Reverse Poller (Redis → DB), and Table Replicator (DB → DB).
• Rust Native Acceleration: High-speed processing of data validation and template rendering using the Rust native core (redgw_core).
• Enterprise-grade Security: Granular ACL permissions, TLS encryption, Rate Limiting, and 3-level audit logs.

Core Role

Safely delegates large-scale AI/GPU inference tasks to independent external inference servers after storing them in an asynchronous queue, offloading the main web/app servers and preventing OOM failures.

Key Features & Specifications

• Brokers 5 major AI operations: Text Embedding (TEI), Text Generation (vLLM), Translation (NLLB), Speech Recognition (STT, Whisper), and OCR (PaddleOCR).
• Secure Isolation Model: Ensures system stability by having workers communicate with external inference containers via FFI/HTTP instead of loading AI models directly.
• Highly Reliable Pipeline: A 3-stage structure of [Queue Registration] → [Batch Processing] → [Result Outbox Storage & Caching].
• Server Protection & Optimization: Supports Continuous Batching and queue threshold backpressure.

Core Role

Effectively resolves the core bottlenecks that arise during large-scale data migration and indexing by partitioning tables into logical units and distributing the workload across multiple processes up to the limits of the hardware.

Key Features & Specifications

• Verified Performance: Migrates 100 million rows in approx. 16 minutes using 10 parallel processes (approx. 80,000 rows/sec).
• 4 Logical Partitioning Methods: Evenly divides a single table into N parts for parallel processing using RANGE, MOD, ROWID, and HASH techniques.
• 6 Data Store Adapters: Target loading to relational DBs (PostgreSQL, Oracle, MySQL, MSSQL), Graph DBs (Memgraph), and Search Engines (MeiliSearch).
• High-Reliability Modules: Memory saving via Server-Side Cursors, Dual Locks, and automatic error row tracking (ErrorDebugger).

Core Role

Delivers real-time one-way alert messages reliably across large-scale user systems, using the Server-Sent Events (SSE) protocol and a high-performance message broker.

Key Features & Specifications

• Large-scale Scalability: A high-availability architecture designed to handle 1,000,000 concurrent connections and 2,000,000 users.
• Message Broker Architecture: Fully integrates the stateless SSE layer and the REST API publishing layer through a NATS JetStream message bus.
• Loss Prevention Mechanism: Supports Last-Event-ID based reconnection recovery to hold and retransmit undelivered messages during momentary network disconnections.
• Multicasting Topics: Supports individual user alerts, department-level alerts, full broadcasts, and remote administrator command channels.

Core Role

Integrates Korean public standard (KCMVP) and international standard cryptographic algorithms into a single high-performance Rust engine, providing consistent API bindings for various programming languages.

Key Features & Specifications

• 8 Standard Algorithms: KCMVP standards (ARIA-256, SEED-128, LEA-128) and international standards (AEGIS-128L/256, AES-256-GCM/SIV, ASCON).
• Diverse Consumption Channels: A CLI binary plus high-performance native FFI bindings for Python, Node.js, and Java (JNI).
• Hardened Memory Security: Built-in Zeroize automatically and physically wipes memory holding secrets once they are no longer in use.
• Functional Integrity: Supports Additional Authenticated Data (AAD), batch and parallel batch operations, file streaming, and HMAC integrity verification.

Core Role

Recursively explores nested compressed files like zip/tar/7z to extract text and structured data for AI and search engines from over 100 different document formats.

Key Features & Specifications

• Highly Versatile Format Support: Office (docx, pptx, xlsx), Hancom (hwp, hwpx), PDF, over 100 types of configuration/source code files, and OCR fallback.
• Recursive Compression Exploration: Builds a Virtual File System (VFS) to recursively decompress up to 16 levels of nested compression while defending against zip bombs.
• Diverse Loading Models: Supports 14-column standard JSONL/CSV output and direct loading into SQLite, PostgreSQL, MySQL, MariaDB, and Oracle.
• Non-disruptive Fault Isolation: Safely isolates the Extractor when exceptions/panics occur during file analysis, preventing complete pipeline interruption.