← Back to OverviewSecPhi Deep Architecturev1.0
SecPhi Home →

System Architecture

Deep technical breakdown of SecPhi's multi-agent AI vulnerability analysis platform

1

High-Level Architecture

👥
Users
Security Teams
→
🖥️
Frontend
Next.js 14
→
⚡
API Layer
Edge Functions
→
🧠
AI Engine
Multi-Agent
→
🗄️
Data Sources
NVD, CISA, OSV
2

Multi-Model LLM Orchestration

📥
CVE Analysis Request
User triggers analysis
↓
🔀
Intelligent Router
Routes to best available model with fallback
↓↓↓
🔷
Gemini Flash
Primary • Fast
🟢
GPT-4o Mini
Fallback 1
🟠
Claude Haiku
Fallback 2
↓↓↓
🤝
Multi-Agent Consensus
Scanner, Scorer, Exploit, Patch agents reach agreement
↓
✅
Validated Analysis
Severity rating with confidence scores
3

Data Processing Pipeline

1
📥
Input
User enters CVE ID or tech stack
CVE IDPackage List
→
2
🔍
Fetch
Query NVD API for CVE data
NVD APIREST
→
3
🏛️
Enrich
Check CISA KEV status
CISA KEVJSON
→
4
📊
Score
Calculate confidence from sources
Weighted AvgCVSS
→
5
🧠
Analyze
Multi-agent AI analysis
GeminiGPTClaude
→
6
📋
Output
Render results with visualizations
ReactCharts
4

Screen Components & Connections

01CVEWatch
Real-time vulnerability scanning and analysis
NVDAIKEV
Click for detailed architecture →
02Agent Debate
Multi-agent AI security assessment
ClaudeGPTGemini
Click for detailed architecture →
03Attack Path
Exploitation chain visualization
AIMITRE
Click for detailed architecture →
04Confidence
Weighted scoring from multiple sources
NVDCISAOSV
Click for detailed architecture →
05LLM Orchestration
Multi-provider AI with failover
GeminiGPTClaude
Click for detailed architecture →
06Data Pipeline
Real-time vulnerability data aggregation
NVDAPIKEV
Click for detailed architecture →
07StackScan
Dependency vulnerability detection
OSVGitHubAI
Click for detailed architecture →
08Reports
PDF report generation and export
PDFAPI
Click for detailed architecture →
5

External API Sources

🤖
AI Models
🔷
Google Gemini
Primary LLM
🟢
OpenAI GPT
Fallback LLM
🟠
Anthropic Claude
Fallback LLM
🛡️
Security Data
🏛️
NVD Database
CVE data source
⚠️
CISA KEV
Exploit catalog
📦
OSV Database
Package vulns
💻
GitHub Advisory
Code vulns
☁️
Infrastructure
▲
Vercel
Hosting & Edge
⚡
Next.js 14
App Framework
🎨
Tailwind CSS
Styling
6

Multi-Agent Analysis System

🔍
Scanner Agent
Detection
Identifies vulnerabilities in code, configs, and dependencies
📊
Scorer Agent
Risk Assessment
Calculates risk scores using CVSS, EPSS, and context
⚡
Exploit Agent
Threat Intel
Analyzes exploitation likelihood and attack surface
🔧
Patch Agent
Remediation
Recommends fixes, patches, and mitigation strategies

SecPhi Architecture Documentation • Version 1.0