oracle-ai-developer-hub

From RAG to Agents Workshop

Build a complete RAG pipeline and agentic system with Oracle AI Database, OpenAI, and the Agents SDK

What You Will Build

Starting from raw data, you will construct a Research Paper Assistant — a multi-agent system that retrieves, reasons over, and discusses 200 ArXiv papers stored in Oracle AI Database. Along the way you’ll implement five retrieval strategies, build an end-to-end RAG pipeline, wrap retrieval as agent tools, compose a multi-agent orchestration system, and finish with persistent session memory backed by Oracle.

Workshop Parts

Part	Topic	Guide
1	Oracle AI Database setup and connection	Part 1 Guide
2	Data loading and embedding generation	Part 2 Guide
3	Database table setup and data ingestion	Part 3 Guide
4	Retrieval mechanisms (keyword, vector, hybrid, graph)	Part 4 Guide
5	Building a RAG pipeline	Part 5 Guide
6	AI agents — basics and tools	Part 6 Guide
7	Agent orchestration and chat system	Part 7 Guide
8	Session memory with Oracle AI Database	Part 8 Guide

TODO Checklist — all 7 tasks at a glance with links to their guide sections.

Getting Started

This workshop lives inside the oracle-ai-developer-hub repository. Use git sparse-checkout to pull just this workshop without cloning the rest of the hub:

# Clone the hub with no files and no blobs
git clone --filter=blob:none --no-checkout https://github.com/oracle-devrel/oracle-ai-developer-hub.git
cd oracle-ai-developer-hub

# Enable sparse-checkout and select only this workshop
git sparse-checkout init --cone
git sparse-checkout set workshops/from_rag_to_agents_workshop

# Materialise the files and move into the workshop
git checkout main
cd workshops/from_rag_to_agents_workshop

# Start Oracle AI Database
docker compose -f .devcontainer/docker-compose.yml up -d oracle

# Install dependencies
pip install -r requirements.txt

# Launch Jupyter
jupyter lab workshop/notebook_student.ipynb

Wait approximately 2 minutes for Oracle to initialise before running notebook cells.

Updating later: git pull from inside oracle-ai-developer-hub refreshes only the paths you’ve selected with sparse-checkout.

Workshop Files

from-rag-to-agents-workshop/
├── .devcontainer/
│   ├── devcontainer.json       Codespaces configuration
│   ├── docker-compose.yml      Oracle AI Database container
│   ├── setup_build.sh          Dependency installation and kernel registration
│   ├── setup_runtime.sh        Oracle startup and vector memory configuration
│   ├── start_oracle.sh         Oracle health check on Codespace restart
│   └── oracle-init/
│       └── 01_vector_memory.sql  Vector memory pool initialisation
├── workshop/
│   ├── notebook_student.ipynb    Your working notebook (contains TODO gaps)
│   └── notebook_complete.ipynb   Complete reference (do not open until done)
├── docs/
│   ├── part-1-oracle-setup.md
│   ├── part-2-data-loading.md
│   ├── part-3-table-setup.md
│   ├── part-4-retrieval.md
│   ├── part-5-rag-pipeline.md
│   ├── part-6-agents-basics.md
│   ├── part-7-orchestration.md
│   ├── part-8-session-memory.md
│   └── TODO-checklist.md
├── images/
├── requirements.txt
└── README.md

Stack

Oracle AI Database via gvenzl/oracle-free:23-full
sentence-transformers — local embedding model (nomic-embed-text-v1.5, 768-dim), no API key needed
oracledb — Python Oracle driver
OpenAI API — LLM generation (GPT-5)
openai-agents — Agent SDK (Agent, Runner, function_tool, orchestration)

Where to Next?

Oracle AI Developer Hub — More technical assets, samples, and projects with Oracle AI
Oracle Developer Resource — Documentation, tools, and community for Oracle developers
OpenAI Agents SDK Documentation — Full reference for the Agent SDK used in Parts 6-8

Built for the Oracle AI Developer Experience team.