OracleSummary integration

Oracle AI Database supports AI workloads where you query data by meaning (semantics), not just keywords. It combines semantic search over unstructured content with relational filtering over business data in a single system—so you can build retrieval workflows (like RAG) without introducing a separate vector database and fragmenting data across multiple platforms.

This guide demonstrates how to generate document summaries using OracleSummary from the Oracle AI Vector Search LangChain integration.

Why summarization here? Summaries are a practical way to compress long documents into retrieval-friendly content (previews, metadata, or condensed context) while keeping governance and operational guarantees close to the data.

If you are just starting with Oracle Database, consider exploring the free Oracle 26 AI. For background on user administration, refer to the official Oracle guide.

Overview

Integration details

Setup

To access OracleSummary, install the @oracle/langchain-oracledb helpers (with @langchain/core) and make sure the Oracle Database driver prerequisites are satisfied for your environment.

Credentials

Set credentials (or use a secrets manager) for the Oracle user that owns your summarization configuration:

export ORACLE_USER=testuser
export ORACLE_PASSWORD=testuser
export ORACLE_DSN="localhost:1521/free"

If you plan to call third-party providers such as OCI Generative AI or Hugging Face, create credentials inside Oracle Database first (for example OCI_CRED, HF_CRED) using the PL/SQL helpers documented in the Oracle AI Vector Search guide.

Installation

npm install @oracle/langchain-oracledb @langchain/core

yarn add @oracle/langchain-oracledb @langchain/core

pnpm add @oracle/langchain-oracledb @langchain/core

Instantiate the tool

import oracledb from "oracledb";
import { OracleSummary } from "@oracle/langchain-oracledb";

const connection = await oracledb.getConnection({
  user: process.env.ORACLE_USER,
  password: process.env.ORACLE_PASSWORD,
  connectionString: process.env.ORACLE_DSN,
});

const summary = new OracleSummary(
  connection,
  {
    provider: "database",
    glevel: "S",
    numParagraphs: 1,
    language: "english",
  },
  process.env.HTTP_PROXY, // optional
);

The third constructor argument is an optional proxy string. Supply it only when outbound requests must traverse an HTTP proxy (for example, when invoking Hugging Face endpoints).

Summarize text with in-database models

Run ONNX summarization models directly in Oracle Database to keep data on the same host as your transactional workloads.

const result = await summary.getSummary(
  "Oracle Database combines relational and vector workloads so you can build secure RAG pipelines.",
);

console.log(result); // => concise summary text

Use managed providers

Switch the provider parameter to route summarization requests through OCI Generative AI or Hugging Face. Provide the credential name you registered in Oracle Database and, if required, a proxy string.

const ociSummary = new OracleSummary(
  connection,
  {
    provider: "ocigenai",
    credential_name: "OCI_CRED",
    url: "https://inference.generativeai.us-chicago-1.oci.oraclecloud.com/20231130/actions/summarizeText",
    model: "cohere.command",
  },
  process.env.HTTP_PROXY,
);

const hfSummary = new OracleSummary(connection, {
  provider: "huggingface",
  credential_name: "HF_CRED",
  url: "https://api-inference.huggingface.co/models/",
  model: "facebook/bart-large-cnn",
  wait_for_model: "true",
});

Chain with LangChain tools

You can wrap OracleSummary in a custom LangChain tool or runnable to integrate it with agent tool-calling.

import { tool } from "@langchain/core/tools";
import { z } from "zod";

const summarizeTool = tool(
  async ({ text }: { text: string }) => summary.getSummary(text),
  {
    name: "oracle_summary",
    description: "Summarize Oracle-sourced documents.",
    schema: z.object({
      text: z.string().describe("Full text to summarize"),
    }),
  },
);

const toolSummary = await summarizeTool.invoke({ text });

Next steps

• Load content with OracleDocLoader
• Chunk and normalize with OracleTextSplitter
• Generate embeddings using OracleEmbeddings and store them in OracleVS

API reference

For detailed documentation of all OracleSummary parameters and return types, see the Oracle LangChain Oracle DB repository.