feat(langchain-oracle): Add Oracle Document Loader and Oracle Vector Search

[hackerdave]

Description

This PR integrates Oracle Document Loader and Oracle Vector Search with LangChain.js. The integration enables the use of Oracle's advanced vector search capabilities within the LangChain.js framework.

Key Features

• Adds support for Oracle Document Loader in LangChain.js.
  • loading documents either from the file system or a table
  • generating embeddings either via an ONNX model loaded into the database or a third-party REST call
  • summarizing documents either via Oracle Text or a third-party REST call
• Adds support for Oracle Vector Search in LangChain.js.
  • includes filter support
• Source is under libs/langchain-oracle
• Includes doc with example usage and tests for the integration.

Example usage

Text Splitter

import {OracleTextSplitter} from "@langchain/oracle";

const splitter = new OracleTextSplitter(conn, splitter_params);
let chunks = await splitter.splitText(doc.pageContent);

Embeddings

import {OracleEmbeddings} from "@langchain/oracle";

const embedder = new OracleEmbeddings(conn, embedder_params, proxy);
const embed = await embedder.embedQuery(chunk);

Testing

• Verified integration with unit tests.
• Added relevant test cases to ensure proper functionality.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Preview	Comments	Updated (UTC)
langchainjs-docs	Error			Aug 21, 2025 7:22pm

1 Skipped Deployment

Project	Deployment	Preview	Comments	Updated (UTC)
langchainjs-api-refs	Ignored			Aug 21, 2025 7:22pm

[christian-bromann]

@hackerdave thanks for the contribution 🙏 our team has been working on the upcoming v1 release of LangChain and unfortunately didn't had the resources yet to look at this PR. We will provide an update soon.

[cjbj]

Split this into two examples. Make the second have a method e.g dbPoolConnection() to get a connection from the pool.

Consider adding links to the node-oracledb connection & pool creation doc sections:

• https://node-oracledb.readthedocs.io/en/latest/user_guide/connection_handling.html#standalone-connections
• https://node-oracledb.readthedocs.io/en/latest/user_guide/connection_handling.html#creating-a-connection-pool

Examples later in the doc need connecting handling to be consistent. E.g. the "Testing the Integration" example calls both dbPool() and dbConnect(). The latter call should be changed to invoke the new dbPoolConnection() method.

[hntrl]

@hackerdave @cjbj what's your thoughts on splitting this into its own repo similar to how the python integration is? We're trying to scope down on the amount of maintenance footprint this repo has.

[cjbj]

@hntrl I'll defer to @hackerdave on this.

[hackerdave]

Yes we are looking into splitting the repo like the python integration so langchain-oracle would be in the oracle repo. It will take a month to get approvals. Should the documentation (i.e., cookbook and docs) remain in the langchain-ai repo?

[hntrl]

@hackerdave Going forward, it's probably best to add docs to the docs repo