Oracle 发布 oracle-ai-developer-hub：面向 Oracle AI Database 与 OCI 的 AI 开发资源库

This repository contains technical resources to help AI Developers and Engineers build AI applications, agents, and systems using Oracle AI Database and OCI services alongside other key components of the AI/Agent stack.

What You'll Find

This repository is organized into several key areas:

📱 Apps (/apps)

Applications and reference implementations demonstrating how to build AI-powered solutions with Oracle technologies. These complete, working examples showcase end-to-end implementations of AI applications, agents, and systems that leverage Oracle AI Database and OCI services. Each application includes source code, deployment configurations, and documentation to help developers understand architectural patterns, integration approaches, and best practices for building production-grade AI solutions.

Name	Description	Link
FitTracker	Gamified fitness platform built with Oracle 26ai JSON Duality Views (FastAPI + Redis), created live during a webinar.
agentic_rag	Intelligent RAG system with multi-agent Chain of Thought (CoT), PDF/Web/Repo processing, and Oracle AI Database 26ai integration
finance-ai-agent-demo	Financial services AI agent with Oracle AI Database as a unified memory core for vector, graph, spatial, and relational queries
generative-ui-data-chat	Next.js and OpenUI data chatbot that pairs Oracle AI Database 26ai SQL, vector, and hybrid search with typed generative UI components such as charts, KPI cards, tables, and source cards.
oci-generative-ai-jet-ui	Full-stack AI application with Oracle JET UI, OCI Generative AI integration, Kubernetes deployment, and Terraform infrastructure
tanstack-shoe-store	AI chat app using TanStack Start and Oracle 26ai Select AI to query a shoe store database with natural language
team-brain	Team knowledge base (the what to a personal second brain's who): Slack, GitHub and docs connectors into one shared table, in-database ONNX embeddings, Oracle Text + vector hybrid retrieval fused by RRF through langchain-oracledb, and per-caller access control enforced by a DBMS_RLS row policy on the table itself, exposed to Claude Code over MCP (stdio or HTTP) and to a LangChain create_agent CLI.
oracle-data-migration-harness	AI agent harness that migrates a RAG corpus from MongoDB into Oracle AI Database 26ai while preserving vector search and unlocking SQL/JSON Duality queries
supplychain-demand-planning-agent	Multi-agent demand-planning assistant with a LangGraph supervisor over two specialists; vector knowledge, long-term memory, per-thread checkpoints, semantic LLM cache, and chat history all share one Oracle AI Database
idp-oracle-ai-database	Intelligent Document Processor that stores the BLOB, extracted text, structured JSON, and vector for each document in one Oracle AI Database 26ai — text extraction, summarization, embeddings, k-NN classification, and LLM field extraction all run inside or from the database via DBMS_VECTOR_CHAIN; AWS supplies only compute (Lambda + S3 + CloudFront)
vector-development	AI Database vector development sample apps for semantic search, RAG, product discovery, code search, and geospatial search. For Python development, see the Oracle VecDB Python SDK.
oraviz-mcp	Minimal, visualization-first MCP server for Oracle AI Database 26ai -- seven read-only tools that cap what returns to the model's context (compact markdown tables, 25-row previews, 500-row hard cap) and render results as PNG charts, including PCA projections of VECTOR columns.

📓 Notebooks (/notebooks)

Jupyter notebooks and interactive tutorials covering:

• AI/ML model development and experimentation
• Oracle Database AI features and capabilities
• OCI AI services integration patterns
• Data preparation and analysis workflows
• Agent development and orchestration examples

Name	Description	Stack	Link
agentic_rag_langchain_oracledb_demo	Multi-agent RAG with langchain-oracledb: OracleVS, OracleEmbeddings, OracleTextSplitter, and CoT agents	Oracle AI Database, langchain-oracledb, Ollama
fs_vs_dbs	Compare filesystem vs database agent memory architectures.	LangChain, Oracle AI Database, OpenAI
memory_context_engineering_agents	Build AI agents with 6 types of persistent memory.	LangChain, Oracle AI Database, OpenAI, Tavily
oracle_rag_agents_zero_to_hero	Learn to build RAG agents from scratch using Oracle AI Database.	Oracle AI Database, OpenAI, OpenAI Agents SDK
oracle_rag_with_evals	Build RAG systems with comprehensive evaluation metrics	Oracle AI Database, OpenAI, BEIR, Galileo
oracle_data_migration_harness_walkthrough	Walk through a MongoDB-to-Oracle AI Database migration harness with vector parity, verification, and JSON Relational Duality	Oracle AI Database 26ai, MongoDB, FastAPI, React, sentence-transformers
agent_reasoning_demo	Interactive demo of 11 cognitive architectures (CoT, ToT, ReAct, Self-Reflection, and more) for agent reasoning	Ollama, agent-reasoning
oracle_agentic_rag_hybrid_search	Agentic RAG with vector, keyword, and hybrid search in a single SQL query using LangGraph ReAct agent	Oracle AI Database, langchain-oracledb, LangGraph, OpenAI
f1_miami_strategy_oracle_26ai	F1 Miami GP strategy intelligence for 2026 — SQL, hybrid vector+keyword search, JSON documents, and property graph in one Oracle 26ai database using real FastF1 data	Oracle AI Database, FastF1, sentence-transformers, Plotly
multicloud/	AWS, Azure, Google Cloud, and MongoDB API samples running Oracle AI Database outside OCI	Oracle AI Database + AWS / Azure / Google / MongoDB
vector-development	AI Database vector development notebook collection covering auto-embedding, BYOV, search diagnostics, indexing, OCI embeddings, Gemini RAG, parks search, and bulk loading.	AI Database vector APIs, Python, Pandas

📚 Guides (/guides)

Comprehensive documentation, reference materials, and conference presentations covering AI agent architecture, reasoning strategies, and memory systems.

Name	Description	Link
Building the Brain and Backbone of Enterprise AI Agents	Advanced reasoning and infrastructure strategies for enterprise AI agents. Covers the 2026 agent stack (layered architecture), reasoning patterns (Chain of Thought, Tree of Thoughts, Self-Reflection, Least-to-Most, Decomposed Prompting), and context/belief updates. Presented at DevWeek SF 2026 by Nacho Martinez.
Memory Engineering: The Discipline Behind Memory Augmented Agents	Deep dive into memory engineering as a discipline for AI agents — the science of helping agents remember, reason, and act. Covers the memory ecosystem, form factors, and key disciplines shaping memory-augmented agents. Presented at DevWeek SF 2026 (Keynote) by Richmond Alake.
Agent Memory with Oracle AI Database	Agent memory architectures and Oracle AI Database as the memory core for AI agents. Presented at the AI Developer Conference hosted by DeepLearning.AI in April 2026 by Eli Schilling.
Memory Engineering with Deep Agents	How memory becomes permanent infrastructure for agents rather than a workaround for a smaller model or context window. Walks through the four memory layers, the Deep Agents middleware architecture (AGENTS.md, skills, checkpointing), procedural and episodic memory, and the four ways memory breaks. Presented by Colin Francis (LangChain JavaScript OSS team, Deep Agents).
Oracle AI Database + Agent Memory with LangChain	Technical session on building agent memory on Oracle AI Database with LangChain and LangGraph. Covers agent form factors and build modes, then a deep dive into a multi-agent stack on a single database — AsyncOracleSaver checkpointer for short-term memory, long-term memory, OracleSemanticCache, and vector + lexical retrieval — tied to the supply-chain demand agent workshop. Presented by Richmond Alake (Oracle) and Colin Francis (LangChain).
How Oracle AI Database Meets the AI Agent Era	Overview deck mapping Oracle AI Database to the AI agent landscape: the four AI application form factors, four build modes from no-code to custom code, and the matching Oracle products (Private Agent Factory, Select AI / APEX AI Assistant, frameworks like LangChain/LlamaIndex/OAMP/Haystack). Frames the agent as model + harness and memory as an engineering discipline. Presented by Richmond Alake (Oracle) and Colin Francis (LangChain).

🧠 Agent Memory (/notebooks/agent_memory)

Notebooks focused on the Oracle AI Agent Memory package (oracleagentmemory) — the AI-Agent Memory Package built on top of Oracle AI Database. These notebooks demonstrate how to use Oracle AI Database as the unified memory core for AI agents, serving conversation history, durable facts, and entity state from a single converged engine instead of stitching together a vector DB, key-value store, and relational store.

The collection covers the package's developer guide, benchmarks against naive memory, and three end-to-end framework examples (OpenAI Agents SDK, Claude Agent SDK, LangGraph).

Name	Description	Stack	Link
OAMP Developer Guide	Step-by-step guide to the oracleagentmemory API: connection, the three core primitives (users/agents, memories, threads), automatic extraction, and vector retrieval.	OAMP, LiteLLM
OAMP Benchmarks	Quantify token cost, latency, and response quality of OAMP vs. naive flat-history memory across 80 scripted turns with three agent variants.	OAMP, LiteLLM, OpenAI
Deep Research Agent	Build a deep research agent for human genome exploration that uses Tavily for live web search and Oracle AI Agent Memory for durable findings across sessions.	OpenAI Agents SDK, Tavily, OAMP
Supply Chain Assistant	A supply chain assistant that tracks shipment cargo via in-process tools and an MCP server, with shipment records and operational notes persisted in OAMP.	Claude Agent SDK, MCP, OAMP
Mortgage Approval Workflow	A deterministic mortgage approval workflow modeled as a LangGraph StateGraph where OAMP persists applicant data and audit trails so failed runs can resume.	LangGraph, OAMP

See the Agent Memory README for a recommended reading order, prerequisites, and Open-in-Colab links.

🔗 LangChain Ecosystem (/notebooks/langchain_ecosystem)

Notebooks that build on the LangChain ecosystem — LangChain, LangGraph, Deep Agents, and Oracle's first-party integrations (langchain-oracledb, langgraph-oracledb, langchain-oci) — using Oracle AI Database as the single backend for vectors, agent memory, checkpoints, the LLM cache, and chat history. They build up from a starter RAG app to a multi-agent supervisor.

Name	Description	Stack	Link
oracle_langchain_example	The starter: build a minimal RAG application on Oracle AI Vector Search with OracleVS and LangChain.	Oracle AI Database, langchain-oracledb, HuggingFace
langchain_oracle_semantic_cache_chat_history	Two langchain-oracledb primitives: OracleSemanticCache (reuse answers for meaning-equivalent prompts) and OracleChatMessageHistory (durable chat transcripts).	langchain-oracledb, OpenAI (embeddings), Anthropic
research_agent_with_deepagents_oracle	A durable Deep Agents research agent via langchain_oci.create_deepagents_agent; Oracle backs the hybrid-search corpus, checkpoints, long-term notes, and the virtual filesystem.	langchain-oci (deepagents), Anthropic, HuggingFace, langgraph-oracledb, Oracle AI Database
ai_oncall_triage_langchain	A LangGraph supervisor over two specialists that triages incoming issues; vector KB, long-term memory, checkpoints, semantic cache, and chat history share one Oracle AI Database.	langchain, langgraph, langgraph-supervisor, langchain-oracledb, langgraph-oracledb, Anthropic (Claude Opus 4.8), OpenAI

See the LangChain Ecosystem README for a recommended reading order, prerequisites, and Open-in-Colab links.

🎓 Workshops (/workshops)

Hands-on workshops and guided learning experiences that take developers from fundamentals to production patterns with Oracle AI Database. Each workshop is self-contained with a student notebook (TODO gaps to fill in), a complete reference notebook, step-by-step part guides, and a ready-to-run Codespaces / devcontainer environment with Oracle AI Database pre-configured. Workshops progress from information retrieval and RAG, through agentic systems and orchestration, to memory-augmented agents — together they cover the full stack for building AI applications on Oracle.

Pull a single workshop without cloning the whole hub — each workshop README includes git sparse-checkout instructions so you can fetch only the folder you need.

Name	Description	Stack	Link
Information Retrieval to RAG	Build a Research Paper Assistant over 200 ArXiv papers by implementing five retrieval strategies (keyword, vector, hybrid, graph) and a full RAG pipeline wired to OCI GenAI.	Oracle AI Database, sentence-transformers, oracledb, OCI GenAI (xAI Grok 3 Fast)
From RAG to Agents	Extend the RAG pipeline into a multi-agent system — wrap retrieval as agent tools, compose orchestration, and add persistent session memory backed by Oracle.	Oracle AI Database, sentence-transformers, oracledb, OpenAI API (GPT-5), openai-agents
Agent Memory	Build memory-aware agents: implement a MemoryManager with six memory types in Oracle, apply context-engineering techniques, and compare agent runs with and without memory.	Oracle AI Database, langchain-oracledb, sentence-transformers, OCI GenAI, Tavily
Enterprise Data Agent Harness	Build a memory-aware enterprise data agent on Oracle AI Database 26ai with OAMP, hybrid vector + Oracle Text retrieval, JSON Relational Duality Views, Deep Data Security, and DBMS_SCHEDULER — then see it running in a live Flask + React chat UI.	Oracle AI Database 26ai, OAMP, langchain-oracledb, OCI GenAI, Flask, React
Soccer Analytics Agent	FIFA World Cup analytics agent with Oracle-backed working/episodic/semantic memory, LangChain OracleVS hybrid retrieval, LangGraph OracleDB observability, and a 92-feature XGBoost prediction pipeline exposed through FastAPI + React.	Oracle AI Database, langchain-oracledb, langgraph-oracledb, XGBoost, FastAPI, React, OCI GenAI (Grok 4)
Supply-Chain Demand Planning Agent	Multi-agent demand-planning assistant with a LangGraph supervisor over two specialists; vector knowledge, long-term cross-thread memory, per-thread checkpoints, semantic LLM cache, and chat history all share one Oracle AI Database. Ships both the workshop notebooks and a live FastAPI + React chat app with per-agent context and an animated topology view.	Oracle AI Database, langchain-oracledb, langgraph-oracledb, langgraph-supervisor, FastAPI, React, OCI GenAI (xAI Grok 4.1 Fast Reasoning)

🤝 Partners (/partners)

Notebooks and apps contributed by partners in the AI ecosystem. AI Developers can use these resources to understand how to use Oracle AI Database and OCI alongside tools such as LangChain, Galileo, LlamaIndex, and other popular AI/ML frameworks and platforms.

Name	Description	Stack	Link
CopilotKit	An agentic travel app on Oracle AI Database — a CopilotKit chat over AG-UI backed by a portable Oracle Agent Spec agent on LangGraph, with Oracle Agent Memory (durable recall + supersession) and human-in-the-loop booking.	CopilotKit, Agent Spec, LangGraph, AG-UI, oracleagentmemory
spraay_x402_payment_agents_oracle	Build payment-enabled AI agents with Oracle AI Database and Spraay x402 Gateway — cost-aware agents that pay for premium inference via USDC micropayments	Oracle AI Database, langchain-oracledb, LangChain, Spraay x402

Getting Started

1. Explore Applications: Start with the applications in /apps to see complete, working examples
2. Follow Workshops: Check /workshops for guided learning paths
3. Experiment with Notebooks: Use /notebooks for hands-on experimentation
4. Build Memory-Augmented Agents: Dive into /notebooks/agent_memory for the Oracle AI Agent Memory package
5. Reference Guides: Consult /guides for detailed documentation
6. Check Partner Resources: Explore /partners for integrations with popular AI tools and frameworks

Contributing

This project is open source. Please submit your contributions by forking this repository and submitting a pull request! Oracle appreciates any contributions that are made by the open-source community.

Development Setup

Before contributing, please set up pre-commit hooks to ensure code is automatically formatted:

1. Install pre-commit:

   pip install pre-commit

2. Install additional dependencies (optional, includes pre-commit and ruff):

   pip install -r requirements-dev.txt

3. Install pre-commit hooks:

   pre-commit install

4. Optional: Format existing code:

   pre-commit run --all-files

The pre-commit hooks will automatically format your code using:

• Ruff for Python files (formatting and linting)
• Prettier for JavaScript, TypeScript, JSON, YAML, and Markdown files

For more detailed information, see SETUP_PRE_COMMIT.md.

License

Copyright (c) 2024 Oracle and/or its affiliates.

Licensed under the Universal Permissive License (UPL), Version 1.0.

See LICENSE for more details.

ORACLE AND ITS AFFILIATES DO NOT PROVIDE ANY WARRANTY WHATSOEVER, EXPRESS OR IMPLIED, FOR ANY SOFTWARE, MATERIAL OR CONTENT OF ANY KIND CONTAINED OR PRODUCED WITHIN THIS REPOSITORY, AND IN PARTICULAR SPECIFICALLY DISCLAIM ANY AND ALL IMPLIED WARRANTIES OF TITLE, NON-INFRINGEMENT, MERCHANTABILITY, AND FITNESS FOR A PARTICULAR PURPOSE. FURTHERMORE, ORACLE AND ITS AFFILIATES DO NOT REPRESENT THAT ANY CUSTOMARY SECURITY REVIEW HAS BEEN PERFORMED WITH RESPECT TO ANY SOFTWARE, MATERIAL OR CONTENT CONTAINED OR PRODUCED WITHIN THIS REPOSITORY. IN ADDITION, AND WITHOUT LIMITING THE FOREGOING, THIRD PARTIES MAY HAVE POSTED SOFTWARE, MATERIAL OR CONTENT TO THIS REPOSITORY WITHOUT ANY REVIEW. USE AT YOUR OWN RISK.

Note: This repository is actively maintained and updated with new resources, examples, and best practices for Oracle AI development.