System Architecture

Understand how VoxCore protects your most valuable asset: real-time SQL queries.

Data Flow
1
User Request
Natural language query from VoxCore web application. Example: "Show revenue by region"
↓
2
AI Query Engine
Claude/GPT-4 processes the request. Analyzes intent, identifies required tables/columns, generates SQL with context awareness.
↓
3
SQL Firewall
Validates generated SQL before execution. Detects injection attacks, policy violations, risky operations. Assigns risk score (0.0-1.0).
↓
4
Database Execution
Safe SQL executes on your database (Snowflake, SQL Server, PostgreSQL). Results returned immediately with audit log.
↓
5
Results & Visualization
Charts, tables, and natural language summary displayed in UI. All queries logged for compliance and audit.

Core Components

🎨
VoxCore UI
React 18 frontend with real-time query interface, governance dashboard, and audit logs.
🤖
AI Query Engine
LLM integration (Claude 3 or GPT-4) with prompt engineering for safe SQL generation.
🔥
SQL Firewall
Proprietary validation layer. Blocks injection attacks, policy violations, and risky patterns.
💾
Database Connectors
Native support for Snowflake, SQL Server, PostgreSQL. Works with your existing infrastructure.
📊
Query History
Complete audit trail of executed queries, risk scores, and user attribution for compliance.
🔐
Governance Layer
Policy enforcement engine. Define rules for allowed tables, columns, operations.

Security Layers

🛡️

SQL Injection Detection

AST parsing and pattern matching to detect malicious SQL fragments before execution.

👤

Query Fingerprinting

Unique fingerprint for each query type. Detects anomalies and repeated attack patterns.

📋

Policy Enforcement

User-defined rules restrict access to sensitive tables, columns, and operations.

📝

Audit Logging

All queries logged with user, timestamp, risk score, and results for compliance.

Data Flow Through Firewall

User Input
"Show revenue by region for Q4 2025"
↓
LLM Prompt
Database schema, safety guidelines, user permissions → Claude/GPT-4 → SQL output
↓
Firewall Checks
✓ SQL syntax ✓ No injection ✓ Policy compliance ✓ PII detection ✓ Risk scoring
↓
Decision
Risk Score: 0.08 (Low) → ✅ APPROVE & EXECUTE
↓
Results
Chart + Table + Natural Language Summary + Audit Log Entry

Technology Stack

Frontend
React 18 TypeScript Vite CSS Grid
Backend
Python 3.11 FastAPI Uvicorn
AI/LLM
LangChain Claude 3 GPT-4
Databases
Snowflake SQL Server PostgreSQL
Deployment
Render (Cloud) GitHub Actions Docker

🚀 Production Deployment

Frontend voxcore.onrender.com
React + Vite auto-build
Backend API voxcore.onrender.com/api
FastAPI + Uvicorn auto-deploy
Database Your databases (Snowflake/SQL Server)
VoxCore never stores your data