Enterprise RAG Platform
AI That Finds Answers From Enterprise Documents
Find information quickly and accurately from enterprise documents with Dense Vector + BM25 + TEXT_MATCH 3-Way hybrid search. Integrated with Milvus vector DB, multimodal OCR, and Korean-optimized NLP, AI understands context to provide precise answers.
Core Values
3-Way Hybrid Search
Maximize search accuracy by combining Dense Vector (semantic search), BM25 (keyword matching), and TEXT_MATCH (exact match). Leverage the strengths of each approach for optimal results without missing documents.
Milvus Vector Database
Built on Milvus optimized for large-scale vector search, enabling millisecond searches across millions of documents. Ensures scalability and high availability.
Multimodal OCR Pipeline
Extract text from PDFs, images, and scanned documents, converting tables and charts to structured data. Can recognize handwriting and stamps.
Korean-Optimized NLP (Kiwi)
Leverage Kiwi morphological analyzer to accurately understand Korean documents. Handles compound words, neologisms, and technical terms precisely.
Key Features
Multiple Search Strategies
Select or combine optimal search strategies for your use case.
- Dense Vector: Semantic similarity document search
- BM25: Keyword-focused precise search
- TEXT_MATCH: Exact phrase matching search
- Hybrid: Weight-based combined search
Document Processing Pipeline
Automatically process and index documents in various formats.
- Formats: PDF, DOCX, PPTX, XLSX, TXT, HTML, Markdown
- OCR: Tesseract + deep learning table extraction
- Chunking: Semantic Chunking preserves context
- Metadata: Auto-extraction and custom tags
Multi-LLM Support
Integrate various LLMs for answer generation.
- OpenAI: GPT-4, GPT-4o, GPT-3.5 Turbo
- Anthropic: Claude 3.5 Sonnet, Claude 3 Opus
- Google: Gemini Pro, Gemini Ultra
- Korean LLMs: EXAONE, HyperCLOVA X
Async Processing (Celery)
Perform large document processing asynchronously to maintain system responsiveness.
- Batch upload: Process thousands of documents at once
- Progress: Real-time processing status monitoring
- Retry: Automatic reprocessing of failed jobs
- Priority: Priority processing for urgent documents
Conversational Q&A
Ask questions in natural language and receive accurate document-based answers.
- Context retention: Follow-up conversations based on previous questions
- Source citation: Display source documents and pages
- Confidence score: Display answer confidence level
- Multi-language: Questions and answers in Korean/English/Japanese/Chinese
Enterprise Security
Provides enterprise-grade security and access control.
- Per-document access permissions
- Role-based access control (RBAC)
- Audit logs: Record all access and queries
- Data encryption: Encryption at rest and in transit
Use Cases
Internal Knowledge Search
Search internal policies, manuals, and guidelines in natural language. Instantly answer questions about onboarding, policies, and procedures.
Customer Support Automation
Auto-generate answers to customer questions from product manuals, FAQs, and technical documents. Assist agents to reduce response time.
Legal/Contract Document Analysis
Quickly find specific clauses, obligations, and risk factors in contracts and legal documents. Dramatically reduce legal team document review time.
R&D Document Management
Search for related research in reports, patents, and technical documents. Prevent duplicate research and effectively leverage existing results.
Technical Specifications
Unlock the Potential of Your Enterprise Documents
Find information from enterprise documents quickly and accurately with XRag.