Products & Services
Live Demo
Core Product
HadoopForge — Enterprise Hadoop Cluster Intelligent O&M Platform
Filling the Apache Ambari official Stack version gap — doing what the open-source community cannot.
HadoopForge is an enterprise-grade Hadoop cluster full lifecycle management platform independently developed by Shenzhen Qixuan Software Co., Ltd. Built on Apache Ambari 3.0.0 + Bigtop 3.5.0 ecosystem, it is the first commercial product in China to complete this adaptation, solving core pain points in enterprise big data infrastructure deployment, operations, and domestic substitution compliance.
Core Problems Solved
| Industry Pain Point | HadoopForge Solution |
|---|---|
| Ambari official Stack stuck on old versions, unable to manage Bigtop 3.5.0 components | Complete rewrite of BIGTOP/3.5.0 Stack with 15+ service metainfo.xml |
| Hive 4.0.1 fully incompatible with Ambari 3.x | Rewrote hive_check.py, Beeline replaces deprecated HiveCLI, fixed auth case sensitivity |
| Cloudera discontinued free edition, enterprises in dilemma | Open-source community edition free, enterprise edition 60%+ cost reduction, no vendor lock-in |
| 2027 domestic substitution 100% replacement deadline | Fully based on Apache open-source ecosystem, supports domestic chips and OS |
| Traditional deployment takes 2–3 weeks, 70% success rate | Visual drag-and-drop orchestration, 4–6 hours deployment, 98% success rate |
Core Product Features
- Complete BIGTOP/3.5.0 Stack Definition — 15+ service metainfo.xml rewritten, ${stack_version} dynamic mapping, Service Advisor logic rewrite, compatible with Hive 4.0.1 / Spark 3.5.0 / Kafka 3.7.0
- Visual Cluster Orchestration Engine — Drag-and-drop cluster topology design, auto-generated deployment plans, optimized resource allocation
- One-Click Automated Deployment — Dual-engine (Ambari + Bigtop) compatibility, minutes to go live, 98% success rate
- AI Intelligent O&M Brain — LSTM + Isolation Forest algorithm, fault prediction accuracy 92%, configuration drift detection and auto-repair
- Enterprise Security Integration — Kerberos + Ranger + OpenLDAP complete security stack
- Hive 4.0.1 Deep Adaptation — Industry’s first complete solution for Hive 4.0.1 compatibility, a problem the community left unsolved for 2 years
Supported Ecosystem Components (Bigtop 3.5.0)
- Storage Engines: HDFS, HBase, Kudu
- Computing Engines: Spark 3.5.0, Hive 4.0.1, Flink, Storm
- Message Queues: Kafka 3.7.0, Pulsar
- Data Warehouse / OLAP: Doris, ClickHouse, Hudi
- Security & Governance: Ambari, Ranger, Atlas, Kerberos
- Monitoring & Observability: Grafana, Prometheus
Customer Validation
- City Commercial Bank (80 nodes) — Deployment time reduced from 3 weeks to 5 hours, zero-downtime migration
- Telecom Operator (200 nodes) — 1.2PB daily processing, Hive 4.0.1 stable in production, AI alert accuracy 92%
- Manufacturing Enterprise (120 nodes) — Zero configuration drift incidents, O&M headcount reduced 40%, saving ¥1.2M per year
500+ nodes · 12 months · 0 production incidents · Based on Ambari 3.0.0 + Bigtop 3.5.0
Product Editions
Community Edition (Free)
- Complete BIGTOP/3.5.0 Stack definition, experience the latest features immediately
- Apache open-source ecosystem, rich upstream and downstream community projects
- Open source on GitHub, developer community support
- Completely free
Enterprise Edition
- Professional commercial support team (phone, IM, email, on-site)
- Highest priority 7×24 fault response
- Expert support: planning / implementation / proactive inspection / fault investigation / knowledge transfer / critical period coverage
- AI O&M Brain advanced features (fault prediction + auto-repair)
- Domestic substitution compliance specialist support
Cloud Hosted Edition
- SaaS-style rapid deployment, easily manage cloud Hadoop clusters
- Supports major object storage (AWS S3, Alibaba Cloud OSS, MinIO, etc.)
- Elastic scaling, pay-as-you-go
- Unified permission security enhancement
Professional Services
Consulting Services
- Big data infrastructure planning and architecture design
- Ambari / Bigtop migration assessment and solution design
- Domestic substitution compliance consulting (2027 replacement target)
- Digital transformation strategy consulting
Implementation Services
- HadoopForge cluster deployment
- Migration from legacy Ambari / Cloudera environments
- System integration and custom development
Operations Services
- 7×24 technical support hotline
- Cluster health monitoring and alerting
- AI-driven performance optimization
- Emergency fault response (Critical: within 1 hour)
Training Services
- Hadoop / Ambari principles and hands-on training
- HadoopForge administrator certification program
- On-site + online dual-mode delivery
- Enterprise customized training programs
Technical Support Response Levels
| Level | Response Time | Scenario |
|---|---|---|
| Critical P0 | ≤ 1 hour | Production cluster completely unavailable |
| High P1 | ≤ 4 hours | Core functionality impaired |
| Medium P2 | ≤ 8 hours | Partial feature anomaly |
| Low P3 | ≤ 24 hours | Inquiry and optimization advice |
Competitive Advantages
- Fills the Version Gap — First in China to complete Ambari 3.0.0 + Bigtop 3.5.0 full adaptation
- Hive 4.0.1 First-Mover — Industry’s only complete solution, fixing what the community left broken for 2 years
- 60%+ Cost Reduction — vs. Cloudera, with no vendor lock-in
- Domestic Substitution Native — Fully Apache open-source, supports domestic chips and operating systems
- AI-Driven O&M — LSTM + Isolation Forest, 92% fault prediction accuracy
Edge AI Inference Acceleration Product Line
Hongyuan Edge AI Inference Accelerator
The world’s first accelerator to deploy 轻量化低功耗推理 quantization on FPGA hardware. 轻量化低功耗推理 lightweight low-power inference constrains weights to 极低精度, dramatically reducing inference computation — enabling every ¥200-class domestic FPGA to run AI Transformer inference.
2025 Shenzhen RISC-V & HarmonyOS Innovation Competition · Architecture Innovation Award 🏆
Key Metrics
| Metric | Value |
|---|---|
| Storage Compression | 16x |
| Sparse Weight Auto-skip | 33% |
| Matrix Computation Error | 0 (256/256 exact match) |
| Hardware BOM Cost | <¥200 |
| Domestic Component Rate | 100% (Anlogic FPGA + RISC-V + HarmonyOS LiteOS) |
| Full-Chain Validation | 6/6 All PASS, real hardware (no simulation) |
| Total Codebase | 16,487 lines (RTL + OS + inference engine) |
Technical Architecture (4-Layer Full-Stack)
- Application: Transformer inference engine (QKV / Multi-Head Attention / FFN full pipeline)
- OS: HarmonyOS LiteOS (task scheduling / IPC / memory management)
- Accelerator: 轻量化低功耗推理 16×16 Tile array (lightweight inference matrix multiply → add/subtract, 33% sparse skip)
- Hardware: Anlogic EG4X20 FPGA (20K LUT) · PicoRV32 RISC-V @50MHz · 128MB SDRAM
Product Roadmap & Pricing
| Phase | Product | Target Customer | Price |
|---|---|---|---|
| Phase 1 (Current) | AI Inference Dev Board | Universities / Makers / R&D | Contact us |
| Phase 2 (6–12 months) | Vision Inference Module (OV2640 + FPGA + SDK) | Industrial inspection / Security / Agriculture | Contact us |
| Phase 3 (12–18 months) | Wi-Fi AI Edge Node | Smart factory / Agricultural IoT | Contact us |
| Phase 4 (18–24 months) | Inference Acceleration IP Core License (RTL IP) | FPGA / SoC vendors | Contact us |
Target Applications
- University AI hardware labs · FPGA AI course experimental platforms
- Industrial vision inspection · <100ms latency, fully offline
- Smart agriculture IoT · Low power, works in field without internet
- Military / government research institutes · 100% domestic, no export control risk
Smart Terminal AI Middleware
Edge AI tiered inference platform for IoT / industrial embedded scenarios. Fills the gap between “can’t run models on-device” and “pure cloud is unreliable”, letting every MCU device answer device status queries in natural language.
Key Metrics
| Metric | Value |
|---|---|
| MCU Clock | 240MHz RISC-V |
| Peak SRAM Usage | 128KB |
| Local Inference Latency | <100ms (Tier 1 keyword) |
| Codebase | 3,500+ lines (MVP all modules) |
3-Tier Progressive Inference Architecture
- Tier 1 Keyword Match: <100ms · Stored in Flash · Zero dependency · Offline
- Tier 2 Rule NLP Engine: <500ms · 19 intent types · Local execution
- Tier 3 Cloud LLM: <3s · Calls cloud LLM when WiFi available
The inference scheduler automatically selects the optimal path based on query complexity + network state, with automatic fallback from Tier 3 to Tier 2.
Four Core Components
- AI Inference Engine Library — Portable to any RISC-V/ARM MCU + RTOS, unified
ai_engine_query()API - Smart CLI Framework — Mixed natural language and traditional command input, TAB completion, zero dependency
- Device Monitoring SDK — Standardized CPU / memory / temperature / network / sensor collection
- FPGA Acceleration Module (optional) — Compresses inference latency to under 200ms, theoretical 50x speedup
Business Model
- SDK licensing: tiered by device volume — contact us
- Cloud LLM inference service: per-call billing — contact us
- Custom development: contact us
AI Education Product Line
Otowyue Learn AI (悦学智伴)
China’s first AI education platform integrating FSRS × VRM digital human deeply. Fully offline deployment with data never leaving the device. Emotion sensing at the core, FSRS scientific memory as the engine, VRM digital humans as the interface — an AI teacher with soul.
Key Metrics
| Metric | Value |
|---|---|
| Emotion Recognition Accuracy | 92% |
| FSRS Memory Retention Rate | 93%+ |
| Learning Efficiency Improvement | 3x vs. traditional methods |
| End-to-End Voice Response | <800ms |
| Compatible LLM Models | 11+ |
| Production Code Volume | 60,000+ lines |
Five Core Innovations
- Emotion Sensing Engine — Text / voice / expression 3-channel multimodal fusion, 92% accuracy, 8 emotion types
- FSRS × Digital Human (China First) — MIT open-source FSRS algorithm + VRM digital human deep integration, 42% retention improvement
- Knowledge Graph Navigation — Covers all K-12 subjects, prerequisite/dependency mapping, automatic bottleneck detection
- Four-Role AI Collaboration — AI Teacher / AI Assistant / AI Advisor / AI Counselor working in concert
- VRM Digital Human Interaction — Humanized digital teacher, MFA phoneme-level pronunciation evaluation, 11+ LLM compatible
Fully Offline Architecture
Tech stack: Ollama + Sherpa-ONNX + MeloTTS + VRM + FSRS
- Video stays on device, only emotion labels transmitted — meets school intranet security requirements
- Works without internet — applicable to rural and remote schools
- Apache 2.0 open source license, freely customizable, no vendor lock-in
- 11+ LLM compatible, LLM-agnostic design
Product Editions & Pricing
| Edition | Target Customer | Pricing |
|---|---|---|
| ToB Institutional | K-12 schools / Training institutions / Corporate | Contact us |
| ToC Family | Students / Individual learners | Contact us |
| ToG Government | Education bureaus / Rural revitalization | Contact us |
Social Impact
- Covering 200+ rural schools, benefiting 100,000+ students — advancing educational equity
- 92% emotion issue detection rate, early mental health warning system
- Reduces 50% of repetitive teacher workload, 60% teaching efficiency improvement
Digital Transformation Program
SME All-in-One ERP Platform
Targeting the core pain points of SME digital transformation, Qixuan Software delivers a ready-to-use integrated ERP solution. Combining inventory management, financial management, HR, CRM, and procurement into one platform — built on industry best practices to help businesses shift from “experience-driven” to “data-driven” operations.
Five Core Modules
Inventory Management
Full-chain integration of purchasing, sales, and warehouse operations. Real-time inventory visibility, automatic replenishment suggestions, 30%+ improvement in inventory turnover.
- Multi-warehouse / multi-batch management
- Barcode / scan operations
- Stock alerts and auto-replenishment
- Complete in/out traceability
Financial Management
Integrated general ledger, AR/AP, cost accounting, and profit analysis. Real-time sync with business data, automatic financial reports, 60%+ improvement in month-end closing efficiency.
- General ledger management
- Automated AR/AP reconciliation
- Multi-dimensional cost analysis
- Automatic financial statement generation
Human Resources
Digital management of employee records, attendance, payroll, and performance. 50% improvement in HR processing efficiency, employee self-service reduces repetitive work.
- Full employee lifecycle management
- Smart scheduling and attendance
- Automated payroll calculation
- Performance appraisal and OKR
CRM
Full-cycle management of leads, follow-ups, contracts, and after-sales. Industry best-practice funnel model, average 25%+ improvement in sales conversion rates.
- Lead → Opportunity → Contract funnel
- Customer profiling and segmentation
- Sales forecasting and analytics
- After-sales ticket tracking
Procurement & Supply Chain
Fully digital supplier management, RFQ, purchase orders, and receiving. Supplier rating system, 10-15% procurement cost reduction, improved on-time delivery.
- Supplier profiles and ratings
- Online RFQ management
- PO full-cycle tracking
- Receiving inspection and QC
Industry Best Practice Results
| Area | Traditional Approach | Qixuan ERP | Improvement |
|---|---|---|---|
| Inventory | Manual Excel, delayed info | Real-time visibility, auto alerts | 30%+ turnover improvement |
| Month-end Close | Manual 1-2 week process | Real-time sync, close anytime | 60% faster closing |
| Procurement | Offline approvals, opaque | Online approvals, full audit trail | 10-15% cost reduction |
| Sales | Scattered data, hard to track | Visual funnel, real-time dashboard | 25%+ conversion lift |
| HR | Manual attendance/payroll | Automated calculation, self-service | 50% efficiency gain |
Editions & Pricing
| Edition | Company Size | Core Features | Price |
|---|---|---|---|
| Startup | 1-20 people | Inventory + Basic Finance | Contact us |
| Growth | 20-100 people | All modules + CRM | Contact us |
| Professional | 100-500 people | All modules + Custom + Private | Contact us |
Try ERP Platform →{target="_blank"} Contact Us for a Solution →