Question 1

What is the difference between generic RAG and Botmartz Enterprise RAG?

Accepted Answer

Generic RAG relies on naive, fixed-size token chunking and single-stream vector similarity search, which often hallucinates or confuses specific alphanumeric part numbers. Botmartz engineers a Two-Stage Hybrid Retrieval pipeline: Lucene BM25 sparse search runs in parallel with dense vector similarity, merged via Reciprocal Rank Fusion (RRF k=60) and re-scored with a Cross-Encoder reranker. Every statement is cryptographically anchored to verbatim source spans with SHA-256 chunk hashes.

Question 2

How does Botmartz prevent unauthorized database mutations in AI agents?

Accepted Answer

We engineer agents using deterministic LangGraph cyclical state machines with Postgres WAL checkpointing. Routine analytical queries execute automatically, while high-stakes state-changing ERP or CRM actions are routed through strict Human-in-the-Loop review gates requiring 1-click human authorization.

Question 3

Is client data kept private and excluded from model training?

Accepted Answer

Yes. Botmartz systems deploy within your private enterprise perimeter (AWS, GCP, Azure VPC, or on-premise hardware). We operate under strict zero-data-retention contracts; client proprietary data is never transmitted to public model providers or used for foundation model training.

Moments, Milestones &
AI Engineering Highlights

Join the Botmartz AI Community

Moments, Milestones & AI Engineering Highlights

Join the Botmartz AI Community

Moments, Milestones &
AI Engineering Highlights