daimon-plugin-opensearch

An OpenSearch k-NN backed VectorStore plugin for the Daimon AI agent framework.

This crate provides [OpenSearchVectorStore], which stores document embeddings in OpenSearch using its native k-NN plugin. It supports cosine similarity, L2 (euclidean), and inner product distance metrics with HNSW indexing via nmslib, faiss, or lucene engines.

Quick Start

use daimon_plugin_opensearch::{OpenSearchVectorStoreBuilder, SpaceType, Engine};
use daimon::retriever::SimpleKnowledgeBase;
use std::sync::Arc;

let store = OpenSearchVectorStoreBuilder::new("http://localhost:9200", 1536)
    .index("my_docs")
    .space_type(SpaceType::CosineSimilarity)
    .engine(Engine::Lucene)
    .build()
    .await?;

// Compose with an embedding model for a full RAG pipeline:
let kb = SimpleKnowledgeBase::new(embedding_model, store);

Manual Index Setup

If you prefer to manage index creation yourself, disable auto-creation and use the JSON from [index_settings]:

let store = OpenSearchVectorStoreBuilder::new(url, 1536)
    .auto_create_index(false)
    .build()
    .await?;

AWS OpenSearch Service

Enable the aws-auth feature for SigV4 authentication:

daimon-plugin-opensearch = { version = "0.15", features = ["aws-auth"] }