Central HUB – AI-Driven Transportation Data Intelligence Platform

Tech Stack: Oracle ADB (AI Vector Search) • SQL/PLSQL • APEX • Python (optional embeddings)

Use Cases: Incident similarity search • Duplicate detection • Root-cause triage • Knowledge lookup

Author: Tarek I. Adam — Oracle Certified Cloud Consultant (Top 10 in Race to Certification 2025)

This repository demonstrates an end-to-end pattern for using Oracle Autonomous Database (ADB) with AI Vector Search to power semantic search over transportation operations data (shipments & incident reports). It mirrors a real enterprise scenario (“Central HUB”) integrating ADB, Oracle Integration Cloud (OIC), Oracle APEX, and Oracle Analytics Cloud (OAC).

---

Table of Contents

• What this project shows
• Architecture
• Quick start
• Files
• Example Semantic Query
• Author

---

What this project shows

• Create ADB tables for shipments & incidents with a VECTOR column for embeddings
• Load sample transport data
• Build a vector index
• Run semantic similarity queries using VECTOR_DISTANCE(...)
• (Optional) Generate embeddings with Python and push them back into ADB
• Wire an APEX page to search semantically

💡 This repo is self-contained: you can run the SQL scripts directly in SQL Developer Web on ADB. The Python notebook is optional for generating embeddings if you prefer that route.

---

Architecture

Data sources (OIC) → ADB (tables + VECTOR) → Vector Index → SQL queries
                                           ↘ APEX UI (search & results)

---

Quick start

Prerequisites: Oracle ADB 23ai (VECTOR/JSON enabled) and SQL Developer Web access.
(Optional) Python 3.10+ with oracledb, numpy, and your embedding library if generating embeddings.

1. Create an ADB instance (or use an existing one) and open SQL Developer Web.
2. Run schema.sql to create tables and the vector index.
3. Run sample_data.sql to insert sample shipments & incidents.
4. If you have embeddings ready, run vector_update.sql to update the embedding column (vector) for each incident.
   • Alternatively, open embeddings.ipynb to compute embeddings and update ADB via Python.
5. Try the semantic search in vector_query.sql:
   • Replace :query_embedding with a vector literal or a bind variable produced by your embedding function.

---

Disclaimer:
For demo and educational purposes only. Example scripts use portable defaults — review and harden before production deployment.

Files

• schema.sql — Create tables (shipments, incidents) and the vector index
• sample_data.sql — Seed data for shipments & incident reports
• vector_update.sql — Example UPDATE statements to attach embeddings
• vector_query.sql — Example semantic search queries with VECTOR_DISTANCE(...)
• embeddings.ipynb — Python notebook to generate embeddings and update ADB
• apex/ — (optional) screenshots or export of a simple APEX search page

---

Example Semantic Query

-- Find incidents most similar to a given query vector
SELECT
  incident_id,
  description,
  VECTOR_DISTANCE(embedding, :query_embedding) AS similarity
FROM   shipment_incidents
ORDER  BY similarity ASC
FETCH FIRST 5 ROWS ONLY;

Oracle syntax for vector index/parameters may differ slightly by version; consult your ADB 23ai docs if needed. Scripts here use portable defaults.

⬆ Back to top

---

Author

Built by Tarek I. Adam – Oracle Cloud Certified Professional
as a public demonstration of practical Oracle analytics on Oracle Cloud.
Stack: Oracle ADB (Vector Search), PL/SQL, APEX, OAC, OIC
LinkedIn | GitHub

⬆ Back to top