Oracle AI Developer Hub: A Curated Repository of Oracle AI Reference Applications

What the Hub Provides and Who Should Use It

The Oracle AI Developer Hub targets engineers and data scientists who are building AI applications on top of Oracle AI Database (referred to as Oracle AI Database 26ai in some application descriptions) and Oracle Cloud Infrastructure (OCI) services. It is not an installable framework or SDK. It is a curated collection of working code that demonstrates specific patterns: how to build a RAG system backed by Oracle vector search, how to run an MCP server that wraps Oracle AI Database tools, how to wire a LangChain agent to Oracle's row-level security policies.

The README states the repository is organised into several key areas covering apps, docs, guides, notebooks, partners, and workshops. Each area addresses a different stage in a developer's workflow: apps for end-to-end reference architecture, notebooks for exploratory learning, workshops for structured exercises.

Developers who do not use Oracle infrastructure will find that most examples assume Oracle-specific APIs. The vector search examples call Oracle AI Database vector functions, the document processing examples use DBMS_VECTOR_CHAIN, and the finance agent example treats Oracle AI Database as a unified memory core for vector, graph, spatial, and relational queries.

Repository Structure: Five Content Areas

The top-level README describes five main directories. The /apps directory holds complete, working applications with source code, deployment configurations, and documentation. The /docs directory contains documentation. The /guides directory provides topic-specific guidance. The /notebooks directory holds Jupyter notebooks for interactive exploration. The /workshops directory contains structured workshop material.

A /partners directory is also present. The top-level repository entries also include pyproject.toml and requirements-dev.txt, which configure the monorepo's development tooling: Ruff for linting and formatting (configured for Python 3.11, line length 100), and pytest with named markers for integration tests. The default pytest run excludes integration tests so that unit work stays fast; running `pytest -m integration` enables the full stack tests that need a live Oracle DB or Ollama instance.

Each application in /apps is self-contained: it includes its own source code, deployment configuration, and documentation. There is no unified installer that sets up all applications at once.

Key Applications in the Hub

The /apps directory contains a range of reference implementations that show different ways to use Oracle AI Database. The agentic_rag application is described as an intelligent RAG system with multi-agent Chain of Thought processing, handling PDF, web, and repository sources against Oracle AI Database 26ai. The team-brain application connects Slack, GitHub, and document sources into a shared Oracle AI Database table, uses in-database ONNX embeddings, applies Oracle Text plus vector hybrid retrieval fused by Reciprocal Rank Fusion, and enforces per-caller access control through a DBMS_RLS row policy. It exposes the result to Claude Code over MCP in both stdio and HTTP modes.

The supplychain-demand-planning-agent uses a LangGraph supervisor over two specialist agents, with vector knowledge, long-term memory, per-thread checkpoints, semantic LLM cache, and chat history all sharing one Oracle AI Database instance. The oraviz-mcp application is a minimal MCP server with seven read-only tools that cap returned data to compact markdown tables, 25-row previews, and a 500-row hard cap.

The idp-oracle-ai-database application stores document BLOBs, extracted text, structured JSON, and vectors together in one Oracle AI Database 26ai instance. Text extraction, summarisation, embeddings, k-NN classification, and LLM field extraction all run inside or from the database via DBMS_VECTOR_CHAIN; the README notes that AWS supplies only compute through Lambda, S3, and CloudFront.

The Development Tooling Configuration

The repository's pyproject.toml shows that the monorepo uses Ruff as its primary linter and formatter, targeting Python 3.11. The selected Ruff rule sets are pycodestyle errors and warnings, pyflakes, isort, flake8-bugbear, flake8-comprehensions, and pyupgrade. Line length is set to 100. The formatter uses double quotes for strings and space indentation.

Prettier is configured for JavaScript and TypeScript files at a print width of 100, with semicolons and double quotes. The .pre-commit-config.yaml file in the repository root coordinates both tools across file types.

Pytest declares four named markers: integration (requires Oracle DB, Ollama, or a live server), requires_oracle, requires_ollama, and requires_playwright. Declaring markers in pyproject.toml silences PytestUnknownMarkWarning across the monorepo. The addopts setting of `-m 'not integration'` means running pytest without flags skips the full-stack tests.

What the Hub Does Not Cover

The hub is specific to Oracle infrastructure. It does not provide vendor-neutral AI application patterns or examples that run entirely on open source components without Oracle services. The vector search examples rely on Oracle AI Database vector storage and retrieval APIs. The LLM orchestration examples reference Oracle-specific database tools like DBMS_VECTOR_CHAIN and Select AI.

The repository has one historical release tagged OCW24 from 2024-11-19 representing the state shown at Oracle Cloud World 2024. There are no numbered semantic versions for individual applications, so tracking changes to a specific application over time requires working from git history.

The repository does not carry a declared top-level licence. GitHub does not identify a recognised licence for the repository. Individual applications may carry their own licence files inside their directories.

How Oracle AI Developer Hub Compares to LangChain Cookbook

LangChain Cookbook is a widely used collection of AI application examples built around the LangChain framework. It covers patterns like RAG, agents, and tool use across multiple model providers without assuming a specific database vendor. That vendor-neutrality is its appeal for teams that want to switch between providers or use open source vector stores.

Oracle AI Developer Hub assumes Oracle infrastructure. Its trade-off is specificity: the team-brain application shows exactly how to enforce row-level security through DBMS_RLS on an Oracle AI Database table, which LangChain Cookbook does not cover. Engineers evaluating Oracle AI Database as a unified store for vectors, relational data, and documents will find the hub's examples more directly applicable than a generic framework cookbook.

The two collections are not substitutes. They serve different starting points: one for Oracle shops evaluating AI integration, the other for framework-first developers who want to stay provider-agnostic.

Maintenance Status

The last push to the repository was on 2026-09-11. The repository is active and not archived. The developer relations context means updates are likely tied to Oracle product releases and developer events. The OCW24 tag marks the state of the repository at Oracle Cloud World 2024.

The Python tooling configuration targets Python 3.11 with Ruff configured for recent Python upgrade rules (UP rule set). Integration tests require a reachable Oracle DB, Ollama, or Playwright environment, which means contributors need access to Oracle infrastructure to run the full test suite.

Editorial conclusion

Engineers already working with Oracle AI Database or OCI who want working reference code will find this hub directly relevant. The applications in the /apps directory are not toy examples: they show specific integration patterns like in-database ONNX embeddings, DBMS_RLS row-level security, and MCP server exposure. Developers who do not use Oracle infrastructure will find that most examples assume Oracle-specific APIs. The repository carries no declared top-level licence, so contributors and adopters should check the individual application directories or contact the maintainer before reusing code in a commercial product.

Frequently asked questions

Official sources

1. Issues
2. oracle-devrel/oracle-ai-developer-hub on GitHub
3. Project website
4. README
5. Releases