---
title: Playground for RAG
description: Use Playground to combine your Elasticsearch data with the power of large language models (LLMs) for retrieval augmented generation (RAG). The chat interface...
url: https://www.elastic.co/elastic/docs-builder/docs/3028/solutions/elasticsearch-solution-project/playground
products:
- Elastic Cloud Serverless
- Kibana
applies_to:
- Elastic Cloud Serverless: Beta
- Elastic Stack: Beta since 9.1, Preview in 9.0
---
# Playground for RAG
Use Playground to combine your Elasticsearch data with the power of large language models (LLMs) for retrieval augmented generation (RAG). The chat interface translates your natural language questions into Elasticsearch queries, retrieves the most relevant results from your Elasticsearch documents, and passes those documents to the LLM to generate tailored responses.
Once you start chatting, use the UI to view and modify the Elasticsearch queries that search your data. You can also view the underlying Python code that powers the chat interface, and download this code to integrate into your own application.
Learn how to get started on this page. Refer to the following for more advanced topics:
- [Optimize model context](https://www.elastic.co/elastic/docs-builder/docs/3028/solutions/elasticsearch-solution-project/playground-context)
- [View and modify queries](https://www.elastic.co/elastic/docs-builder/docs/3028/solutions/elasticsearch-solution-project/playground-query)
- [Troubleshooting](https://www.elastic.co/elastic/docs-builder/docs/3028/solutions/elasticsearch-solution-project/playground-troubleshooting)
Watch these video tutorials to help you get started:
- [Getting Started](https://www.youtube.com/watch?v=zTHgJ3rhe10)
- [Using Playground with local LLMs](https://www.youtube.com/watch?v=ZtxoASFvkno)
## How Playground works
Here’s a simplified overview of how Playground works:
- User **creates a connection** to LLM provider
- User **selects a model** to use for generating responses
- User **defines the model’s behavior and tone** with initial instructions
- **Example**: "*You are a friendly assistant for question-answering tasks. Keep responses as clear and concise as possible.*"
- User **selects Elasticsearch indices** to search
- User **enters a question** in the chat interface
- Playground **autogenerates an Elasticsearch query** to retrieve relevant documents
- User can **view and modify underlying Elasticsearch query** in the UI
- Playground **auto-selects relevant fields** from retrieved documents to pass to the LLM
- User can **edit fields targeted**
- Playground passes **filtered documents** to the LLM
- The LLM generates a response based on the original query, initial instructions, chat history, and Elasticsearch context
- User can **view the Python code** that powers the chat interface
- User can also **Download the code** to integrate into application
## Availability and prerequisites
For Elastic Cloud and self-managed deployments, select **Playground** from the left navigation menu.
For Elastic Serverless, Playground is available in your Elasticsearch project UI.
To use Playground, you’ll need the following:
1. An Elastic **v8.14.0+** deployment or Elasticsearch **Serverless** project. (Start a [free trial](https://cloud.elastic.co/registration)).
2. At least one **Elasticsearch index** with documents to search.
- See [ingest data](#playground-getting-started-ingest) if you’d like to ingest sample data.
3. An account with a **supported LLM provider**. Playground supports the following:
- [**Elastic Managed LLMs**](https://docs-v3-preview.elastic.dev/elastic/docs-builder/docs/3028/reference/kibana/connectors-kibana/elastic-managed-llm)
- **Amazon Bedrock**
- Anthropic: Claude 3.5 Sonnet
- Anthropic: Claude 3 Haiku
- **OpenAI**
- GPT-3 turbo
- GPT-4 turbo
- GPT-4 omni
- **Azure OpenAI** (note: Buffers responses in large chunks)
- GPT-3 turbo
- GPT-4 turbo
- **Google**
- Google Gemini 2.5 Pro
You can also use locally hosted LLMs that are compatible with the OpenAI SDK. Once you’ve set up your LLM, you can connect to it using the OpenAI connector. Refer to the following for examples:
- [Using LM Studio](https://www.elastic.co/elastic/docs-builder/docs/3028/explore-analyze/ai-features/llm-guides/connect-to-lmstudio-observability)
- [LocalAI with `docker-compose`](https://www.elastic.co/search-labs/blog/localai-for-text-embeddings)
## Getting started
### Connect to LLM provider
To get started with Playground, you need to create a [connector](https://www.elastic.co/elastic/docs-builder/docs/3028/deploy-manage/manage-connectors) for your LLM provider. You can use Elastic Managed LLMs by default. You can also connect to [locally hosted LLMs](#playground-local-llms) which are compatible with the OpenAI API, by using the OpenAI connector.
To connect to an LLM provider, use the following steps on the Playground landing page:
1. Select **New Playground**. Select the `wrench` button in the **Large Language Model (LLM)** tile to connect an LLM.
2. Select your **LLM provider**.
3. **Name** your connector.
4. Select a **URL endpoint** (or use the default).
5. Enter **access credentials** for your LLM provider. (If you’re running a locally hosted LLM using the OpenAI connector, you must input a value in the API key form, but the specific value doesn’t matter.)
If you need to update a connector, or add a new one, click the 🔧 **Manage** button beside **Model settings**.
### Ingest data (optional)
*You can skip this step if you already have data in one or more Elasticsearch indices.*
There are many options for ingesting data into Elasticsearch, including:
- The [Elastic crawler](https://www.elastic.co/guide/en/enterprise-search/current/crawler.html) for web content (**NOTE**: Not yet available in *Serverless*)
- [Elastic connectors](https://docs-v3-preview.elastic.dev/elastic/docs-builder/docs/3028/reference/search-connectors) for data synced from third-party sources
- The Elasticsearch [Bulk API](https://www.elastic.co/docs/api/doc/elasticsearch/operation/operation-bulk) for JSON documents
To add a few documents to an index called `books` run the following in Dev Tools Console:
```json
{ "index" : { "_index" : "books" } }
{"name": "Snow Crash", "author": "Neal Stephenson", "release_date": "1992-06-01", "page_count": 470}
{ "index" : { "_index" : "books" } }
{"name": "Revelation Space", "author": "Alastair Reynolds", "release_date": "2000-03-15", "page_count": 585}
{ "index" : { "_index" : "books" } }
{"name": "1984", "author": "George Orwell", "release_date": "1985-06-01", "page_count": 328}
{ "index" : { "_index" : "books" } }
{"name": "Fahrenheit 451", "author": "Ray Bradbury", "release_date": "1953-10-15", "page_count": 227}
{ "index" : { "_index" : "books" } }
{"name": "Brave New World", "author": "Aldous Huxley", "release_date": "1932-06-01", "page_count": 268}
{ "index" : { "_index" : "books" } }
{"name": "The Handmaids Tale", "author": "Margaret Atwood", "release_date": "1985-06-01", "page_count": 311}
```
We've also provided some Jupyter notebooks to ingest sample data into Elasticsearch. Find these in the [elasticsearch-labs](https://github.com/elastic/elasticsearch-labs/blob/main/notebooks/ingestion-and-chunking) repository. These notebooks use the official Elasticsearch Python client.
### Select Elasticsearch indices
Playground does not support [data stream](https://www.elastic.co/elastic/docs-builder/docs/3028/manage-data/data-store/data-streams) indices.
Once you’ve connected to your LLM provider, it’s time to choose the data you want to search.
1. Select **Add data sources**.
2. Select one or more Elasticsearch indices.
3. Click **Save and continue** to launch the chat interface.
You can always add or remove indices later by selecting the **Data** button from the main Playground UI.
### Chat and query modes
The main Playground UI has two modes:
- **Chat mode**: The default mode, where you can chat with your data via the LLM.
- **Query mode**: View and modify the Elasticsearch query generated by the chat interface.
The **chat mode** is selected when you first set up your Playground instance.
To switch to **query mode**, select **Query** from the main UI.
Learn more about the underlying Elasticsearch queries used to search your data in [View and modify queries](https://www.elastic.co/elastic/docs-builder/docs/3028/solutions/elasticsearch-solution-project/playground-query)
### Set up the chat interface
You can start chatting with your data immediately, but you might want to tweak some defaults first.
You can adjust the following under **LLM settings**:
- **AI Connector**. The model used for generating responses.
- **Instructions**. Also known as the *system prompt*, these initial instructions and guidelines define the behavior of the model throughout the conversation. Be **clear and specific** for best results.
- **Include citations**. A toggle to include citations from the relevant Elasticsearch documents in responses.
Playground also uses another LLM under the hood, to encode all previous questions and responses, and make them available to the main model. This ensures the model has "conversational memory".
Under **Indices**, you can edit which Elasticsearch indices are searched. This affects the underlying Elasticsearch query.
Click **✨ Regenerate** to resend the last query to the model for a fresh response.Click **⟳ Clear chat** to clear chat history and start a new conversation.
### View and download Python code
Select `export` to view the Python code that powers the chat interface. You can integrate it into your own application, modifying as needed. We currently support two implementation options:
- Elasticsearch Python Client + LLM provider
- LangChain + LLM provider
### Next steps
Once you’ve got Playground up and running, and you’ve tested out the chat interface, you might want to explore some more advanced topics:
- [Optimize model context](https://www.elastic.co/elastic/docs-builder/docs/3028/solutions/elasticsearch-solution-project/playground-context)
- [View and modify queries](https://www.elastic.co/elastic/docs-builder/docs/3028/solutions/elasticsearch-solution-project/playground-query)
- [Troubleshooting](https://www.elastic.co/elastic/docs-builder/docs/3028/solutions/elasticsearch-solution-project/playground-troubleshooting)