RAG Engine & Model Router Architecture
💬
User Query
Natural language question
from any namespace
➔
🧠
Query Classifier
Factual / Navigation /
Analytical / Exploratory
➔
⚙️
Engine Selection
Zero-token path or
RAG + LLM synthesis
➔
✔️
Governed Response
Grounded, cited,
ADR-audited
⚡
Zero-Token Engine
Deterministic pattern matching for compliance detection, PII scanning, and factual lookups. No LLM call needed.
$0 API Cost
🔍
RAG + LLM Synthesis
4-tier knowledge base search, chunk retrieval, context window construction, LLM grounded response generation.
API Call
📈
Semantic Cache
Embedding-based query matching. If a similar question was answered recently, return cached response instantly.
Cached
Model Router
➔
Claude Sonnet 4
Primary • Enterprise
➔
Claude Haiku
Fallback • Budget
➔
Local Ollama
Budget Gate • $0
➔
Azure OpenAI
Data Residency
AIXaaS Intelligence Platform
RAG Engine & Model Router