@objective-inc/langchain-retriever

A lightweight wrapper around LangChain's retrieval functionality, designed to work seamlessly with Objective's search API.

Features

• Seamless integration with Objective's search API
• Configurable search parameters
• Compatible with LangChain's retrieval interfaces
• Type-safe implementation in TypeScript

Installation

# npm
npm install @objective-inc/langchain-retriever

# yarn
yarn add @objective-inc/langchain-retriever

# pnpm
pnpm add @objective-inc/langchain-retriever

Configuration Options

The ObjectiveRetriever accepts the following configuration options:

interface ObjectiveRetrieverInput {
  indexId: string; // Required: The ID of your Objective search index
  apiKey: string; // Required: Your Objective API key
  limit?: number; // Optional: Maximum number of results (default: 1)
  objectFields?: string; // Optional: Fields to return in results (default: "")
  rankingExpr?: string; // Optional: Custom ranking expression
}

Usage

import { MultiQueryRetriever } from "langchain/retrievers/multi_query"
import { ChatOpenAI } from "@langchain/openai"
import { ObjectiveRetriever } from "@objective-inc/langchain-retriever";

// Initialize the retriever
const retriever = new ObjectiveRetriever({
  indexId,
  apiKey,
});

// Retrieve relevant documents
const documents = await retriever.getRelevantDocuments("your query here");

Advanced Usage

const retriever = new ObjectiveRetriever({
  indexId,
  apiKey,
  objectFields: "title,content,metadata", // Specify fields to return
  limit: 5 // Return up to 5 results
});

Integration with LangChain

The retriever can be used with other LangChain components:

import { MultiQueryRetriever } from "langchain/retrievers/multi_query";
import { ChatOpenAI } from "@langchain/openai";

// Create a base retriever
const baseRetriever = new ObjectiveRetriever({
  indexId,
  apiKey,
});

// Use with MultiQueryRetriever for enhanced retrieval
const model = new ChatOpenAI({
  modelName: "gpt-4o",
  temperature: 0,
  openAIApiKey,
  streaming: true,
})

const multiQueryRetriever = MultiQueryRetriever.fromLLM({
  llm: model,
  retriever: objectiveRetriever,
  verbose: false,
})

const documents = await multiQueryRetriever.getRelevantDocuments("your query");

Error Handling

The retriever includes built-in error handling for:

• Invalid queries
• API connection issues
• Authentication errors

TypeScript Support

This package is written in TypeScript and includes full type definitions for all features.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.