Technology

Designed at the protocol level, accelerated on the GPU, and extended with AI

Core technology areas where TSoft goes deep

TSoft builds its own AI infrastructure instead of relying on external AI SaaS

Four Technology Domains
01

Remote Desktop Technology

Designed at the protocol level, accelerated on the GPU

DXGINVENCAMFQSVCustom Protocol

Direct design and implementation at the protocol level

High-performance, low-latency transmission technology

02

AI / RAG

Search, understand, and respond

Hybrid SearchVector DBLLM PipelineEmbeddings

Self-built RAG pipeline in operation

03

Automation

See the screen, make decisions, and execute

Image MatchingBrowser AutomationRPAPlaywrightOCR

ai-crawler, T-Cafe agent runtime

04

Systems Programming

From kernel to app, built in-house

KMDF DriverWin32FlutterCross-platform

Virtual HID driver built in-house

Technical Capabilities in Detail

Remote Desktop Protocol

Direct design and implementation at the protocol level, achieving high-performance low-latency communication

Kernel Driver

Virtual HID driver built in-house at the KMDF kernel-mode level

Hardware Encoding

Native integration per GPU vendor with direct NVENC/AMF/QSV wrapper implementation

AI/RAG

Self-built RAG pipeline combining hybrid search and LLM pipeline in operation

Browser Automation

Deterministic execution based on Playwright powering ai-crawler and T-Cafe agent runtime

Cross-Platform

Flutter multi-platform supporting iOS/Android/Windows/macOS

In-House AI Infrastructure

TSoft builds its own AI infrastructure instead of relying on external AI SaaS

01

RAG (Retrieval-Augmented Generation)

rag-service

AI-powered Q&A service

02

AI Crawling

ai-crawler

Automated game news collection

03

Vector Search

Vector DB-based search

Semantic search

04

Hybrid Search

Keyword + semantic integrated search

High-accuracy information retrieval

05

LLM Pipeline

LLM-based response generation

Natural language Q&A

What Sets TSoft Apart

Protocol-Level Depth

Not just using APIs, but understanding and implementing at the protocol level

Full-Stack Hardware Integration

Direct development of the entire pipeline from GPU encoding to kernel drivers

AI Internalization

Self-built RAG, crawling, and vector search instead of depending on external SaaS

Product Ecosystem Design

Building an interconnected product ecosystem, not isolated products