Build an Elasticsearch AI Assistant with DeepSeek‑R1 in 1 Minute

Alibaba Cloud AI Search Open Platform provides componentized AI search services for enterprises and developers. This article focuses on using the built‑in DeepSeek‑R1 series model to build an Elasticsearch AI Assistant that brings generative AI capabilities to intelligent operations.

Key Benefits

Deep Integration: Directly call Elasticsearch APIs for real‑time cluster health diagnostics and dynamic visual dashboards.

Intelligent Interaction: Natural‑language commands automatically generate precise query DSL, enabling end‑to‑end query construction, execution, and optimization.

Scenario Optimization: Provides smart suggestions for troubleshooting, threat hunting, and business analysis, lowering technical barriers.

Custom Knowledge: Configure a private knowledge base and leverage DeepSeek models for assisted Q&A.

Step‑by‑Step Setup

1. Create an Elasticsearch Instance

Follow the official documentation to create an Elasticsearch 8.15+ instance and configure Kibana:

https://help.aliyun.com/zh/es/user-guide/ai-assistant-usage-guidelines

2. Create a DeepSeek Connector

In the connector creation page, select the model "deepseek‑r1" from the default model options and save.

3. Use the DeepSeek Connector for Conversation

Select the newly created DeepSeek connector when starting a chat with the AI Assistant.

Scenario Demonstrations

Scenario 1 – Cluster Operations & Index Management

Create an index with a high number_of_replicas to trigger a yellow cluster state:

PUT test/
{
    "settings": {
        "number_of_replicas": 10
    }
}

Ask the AI Assistant to analyze the issue; it can suggest and, with authorization, automatically adjust the replica setting to restore a green state.

List all non‑hidden indices.

Increase shard relocation speed to 100 MB/s.

Create a new index test_new with three shards, mirroring test.

Explain why the cluster load is high.

Scenario 2 – Visual Analytics

Import sample data via Kibana, then use the AI Assistant to generate visual charts, such as listing the top‑5 product categories by sales from the kibana_sample_data_ecommerce index.

Scenario 3 – DSL Generation & Query Assistance

Ask the AI Assistant to write DSL queries, explain errors, or suggest cluster and index settings adjustments. Example query: "How many product categories exist in kibana_sample_data_ecommerce, list them, and provide the DSL."

Recommended Models and Pricing

DeepSeek‑r1: Input 0.004 CNY/1k tokens, Output 0.016 CNY/1k tokens.

DeepSeek‑v3: Input 0.002 CNY/1k tokens, Output 0.008 CNY/1k tokens.

For more details, visit:

Alibaba Cloud Elasticsearch: https://www.aliyun.com/product/bigdata/elasticsearch

AI Search Open Platform: https://www.aliyun.com/activity/bigdata/opensearch/platform