Veta — Agentic Mineral-Tracking Decision Support
Published:
Private product. This is proprietary work; the live deployment is private. This page describes the architecture and intent without exposing internal data or logic.
Veta is a decision-support system for mineral-processing operations. Ask it a question — typed or spoken — and a multi-stage agent pipeline interprets the intent, assembles the relevant plant state, routes to the right solver tier (cheap heuristics first, heavier optimization only when warranted), and answers with the supporting evidence attached via retrieval over the tracked sites’ data (RAG).
Why it exists
In a plant, the knowledge needed to answer “what should we do about this feed and these constraints?” is spread across telemetry, historical runs, and expert intuition. Pulling it together is slow and hard to audit — and an answer with no traceable evidence is hard to trust on the floor. Veta compresses the loop from question to defensible recommendation and keeps the reasoning auditable rather than a black box.
What it demonstrates
- A single conversational front door (text + voice) to plant decisions.
- A staged agent pipeline with solver tiers that spend compute only when justified.
- RAG-grounded answers — every recommendation carries its evidence.
- Reproducibility by construction: an offline pipeline builds per-site, content-addressed (sha256-manifested) artifacts.
- Deployed live (private), dual-target across a VPS and Azure Container Apps.
A working pattern for agentic decision support in heavy industry — fast, defensible, reproducible.