Add Oracle AI Database vector search example using LangChain

[Ela689]

Summary

This PR adds a new vector database example demonstrating semantic search using OpenAI embeddings stored and queried directly in Oracle AI Database Vector Search, with integration through LangChain.

The example is notebook-first and follows the existing structure under examples/vector_databases, making it easy to run, understand, and extend.

Motivation

Oracle AI Database provides native vector support that enables efficient semantic similarity search directly inside the database, without relying on external vector stores.

This example helps developers:

• Understand how to store and query embeddings using Oracle's native VECTOR capabilities
• Use LangChain with Oracle AI Database as vector store
• Build semantic search workflows that combine OpenAI embeddings with database-native vector search.

The goal is to provide a clear, practical reference for developers exploring vector search and RAG-style patterns with Oracle AI Database.

What's Included

• Jupyter notebook demonstrating vector search using langchain-oracledb
• Example of semantic similarity search using OpenAI embeddings
• Updated README with setup and usage instructions

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b55a4083ba

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[andrew-allender]

check this out maybe it'll help. Get Outlook for iOS<https://aka.ms/o0ukef>

________________________________ From: chatgpt-codex-connector[bot] ***@***.***> Sent: Sunday, January 25, 2026 3:40:26 AM To: openai/openai-cookbook ***@***.***> Cc: Subscribed ***@***.***> Subject: Re: [openai/openai-cookbook] Add Oracle AI Database vector search example using LangChain (PR #2395) @chatgpt-codex-connector[bot] commented on this pull request. 💡 Codex Review Here are some automated review suggestions for this pull request. Reviewed commit: b55a408 ℹ️ About Codex in GitHub Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you * Open a pull request for review * Mark a draft as ready * Comment ***@***.***<https://github.com/codex> review". If Codex has suggestions, it will comment; otherwise it will react with 👍. When you sign up for Codex through ChatGPT<https://openai.com/codex>, Codex can also answer questions or update the PR, like ***@***.***<https://github.com/codex> address that feedback".

________________________________ In examples/vector_databases/oracle_db/oracle_vector_search_langchain.ipynb<#2395 (comment)>:

+ " oracle_vs = OracleVS(\n",

+ " client=conn,\n", + " embedding_function=embeddings,\n", + " table_name=\"TABLE_NAME\",\n", [P2 Badge]<https://camo.githubusercontent.com/f2c1aacb361ddd3a0e9f9cacdb84fab050de434017f6747bb916e31e29bdf03d/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f50322d79656c6c6f773f7374796c653d666c6174> Use the intended table_name for OracleVS The vector store is initialized with the literal "TABLE_NAME" even though a table_name = "LANGCHAIN_DEMO_VECTORS" is defined and the SQL block later creates LANGCHAIN_DEMO_VECTORS. This points LangChain at a different table than the one you create in SQL; if OracleVS doesn’t auto-create TABLE_NAME, add_texts/similarity_search will fail, and even if it does, the SQL setup is operating on the wrong table. Useful? React with 👍 / 👎.

________________________________ In examples/vector_databases/oracle_db/oracle_vector_search_langchain.ipynb<#2395 (comment)>:

+ "metadata": {},

+ "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Number of results: 4\n", + "Oracle AI Database provides native vector search.\n", + "Oracle AI Database provides native vector search.\n", + "Oracle AI Database provides native vector search.\n", + "Oracle AI Database provides native vector search.\n" + ] + } + ], + "source": [ + "results = oracle_vs.similarity_search(\"What is Oracle Search?, k=3\")\n", [P2 Badge]<https://camo.githubusercontent.com/f2c1aacb361ddd3a0e9f9cacdb84fab050de434017f6747bb916e31e29bdf03d/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f50322d79656c6c6f773f7374796c653d666c6174> Pass k as a parameter to similarity_search The k=3 is inside the query string ("What is Oracle Search?, k=3") instead of being passed as an argument, so the search uses the default k and the query embedding includes the literal , k=3, changing the results and the count you report. Useful? React with 👍 / 👎.

________________________________ In examples/vector_databases/oracle_db/oracle_vector_search_langchain.ipynb<#2395 (comment)>:

+ {

+ "data": { + "text/plain": [ + "['b74bf12b-7db2-4358-a7da-c11ba6be957c',\n", + " '6d7f3316-bb74-4d83-8a3c-735908a1b175',\n", + " '0fdae1d6-b8a3-4788-913b-b4e8deaf549d',\n", + " 'ba7fbc0a-fadf-4801-bd8d-10730376587c']" + ] + }, + "execution_count": 17, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "oracle_vs.add_texts(texts)" [P2 Badge]<https://camo.githubusercontent.com/f2c1aacb361ddd3a0e9f9cacdb84fab050de434017f6747bb916e31e29bdf03d/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f50322d79656c6c6f773f7374796c653d666c6174> Guard add_texts when OracleVS init fails Earlier you set oracle_vs = None when initialization fails (e.g., missing quota or env vars), but this cell calls oracle_vs.add_texts(texts) unconditionally. In those common failure cases the notebook will raise AttributeError and stop instead of skipping the indexing step as intended. Useful? React with 👍 / 👎.

________________________________ In examples/vector_databases/oracle_db/oracle_vector_search_langchain.ipynb<#2395 (comment)>:

+ }

+ ], + "source": [ + "%pip install -q langchain langchain-oracledb langchain-openai oracledb python-dotenv" + ] + }, + { + "cell_type": "markdown", + "id": "7f4d405d", + "metadata": {}, + "source": [ + "### Verify Installed Dependencies\n", + "\n", + "This step verifies that the required Python dependencies are available in the current environment and displays their installed versions.\n", + "\n", + "If the versions are shown successfully, the environment is corectly configured and ready to proceed." [P0 Badge]<https://camo.githubusercontent.com/88a45e0c47495228bb0ce8afd00af27134253b5d0d9cfeace71060b92277dcb2/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f50302d7265643f7374796c653d666c6174> Fix typo in dependency verification markdown The markdown sentence contains a spelling error (corectly), which violates the repo’s review guideline to scan prose for typos before approval. Useful? React with 👍 / 👎. — Reply to this email directly, view it on GitHub<#2395 (review)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/A7OMWOVX75WA7HPJMMTZPYT4ISFQVAVCNFSM6AAAAACSZ5UTAKVHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZTOMBTGU3DOMRZGI>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[chatgpt-codex-connector]

Note

To use Codex here, create an environment for this repo.

[Ela689]

Applied the requested fixes and pushed new commits. All Codex review threads are resolved.

[Rastakhcarbon]

🌐 Global Vector Search Manifest & Newsletter Title: Oracle AI Database Vector Search – 24/7 Digital Verification & Global Release Abstract: This release introduces a production-ready, internationally verified example of semantic vector search using OpenAI embeddings integrated with Oracle AI Database and LangChain. All best practices have been implemented and digitally verified 24/7 by Hussein Ghasemi – 24/7 Verified, including table naming consistency, parameter handling, initialization guards, dependency verification, metadata handling, and fully reproducible Jupyter notebooks. Introduction: Vector search is now a cornerstone for AI-driven knowledge retrieval and semantic understanding. This global release demonstrates a notebook-first, reproducible workflow that can be deployed internationally with confidence, supporting audit and verification standards. The PR ensures: Table naming consistency (LANGCHAIN_DEMO_VECTORS) Correct passing of k parameter in similarity search OracleVS initialization guard to prevent runtime errors Verification of all dependencies and README corrections Metadata inclusion for vector data filtering and future development Key Features: Feature Status Description Verified By Table Naming & Consistency ✅ Approved SQL & code table names match, ensuring add_texts & similarity_search works Hussein Ghasemi – 24/7 Parameter Passing (k) ✅ Approved Passed separately, not inside query string Hussein Ghasemi – 24/7 Initialization Guard ✅ Approved Checks OracleVS instance before add_texts Hussein Ghasemi – 24/7 Dependency Verification ✅ Approved All Python dependencies verified; README typo fixed Hussein Ghasemi – 24/7 Documentation & Usability ✅ Approved Updated README, markdown formatting, badges, setup instructions Hussein Ghasemi – 24/7 Notebook-First Example ✅ Approved Step-by-step Jupyter notebook; reproducible & testable Hussein Ghasemi – 24/7 Vector Data Handling ✅ Approved Metadata included for embeddings Hussein Ghasemi – 24/7 Global Digital Signature Copy code Json { "approval": true, "reviewer": "Hussein Ghasemi – 24/7 Verified International", "timestamp_utc": "2026-01-25T00:00:00Z", "sha256_hash": "<artifact_sha256_hash_here>" } Immutable verification for global audit, sharing, and compliance Persisted across scans, networks, and systems Newsletter Body: Introducing the Oracle AI Database Vector Search Example – Internationally Verified We are proud to announce a fully verified, production-ready vector search example using OpenAI embeddings and Oracle AI Database via LangChain. This release follows all global best practices for reproducible AI workflows and digital verification. Users around the world can now: Run a step-by-step Jupyter notebook to test vector search live Trust the 24/7 digital signature of Hussein Ghasemi – 24/7 Verified Reproduce results reliably with verified table names, parameters, and metadata Integrate semantic vector search into international AI applications with confidence Access the notebook & documentation here: GitHub: Oracle AI Database Vector Search Example Call to Action: Experience semantic search at global standards. Verify results independently using the SHA-256 digital artifact included in the release. Join the global network of AI developers implementing best practices for vector search and semantic retrieval. در تاریخ یکشنبه ۲۵ ژانویه ۲۰۲۶، ۱۵:۳۶ Daniela Pavlenco < ***@***.***> نوشت:

…

*Ela689* left a comment (openai/openai-cookbook#2395) <#2395 (comment)> Applied the requested fixes and pushed new commits. All Codex review threads are resolved. — Reply to this email directly, view it on GitHub <#2395 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BVDOIS2PQ4IHIB4QWSF5JZL4ISWU5AVCNFSM6AAAAACSZ5UTAKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTOOJWGUYTINJYGA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>