GenAI Toolkit
On this page
GenAI Toolkit is currently in private beta according to the Algolia Terms of Service (“Beta Services”).
With the Algolia GenAI Toolkit, you can build generative AI experiences with your Algolia data. It combines leading-edge large language models (LLM) with Algolia’s Search API using Retrieval Augmented Generation (RAG) techniques. This makes it simpler to build safe, responsive, and up-to-date GenAI features in your own application.
How GenAI Toolkit works
At a high-level, the GenAI Toolkit ties together your prompts, context, and LLMs so that your application can focus on sending queries and rendering the resulting generative content.
To get started, connect the GenAI Toolkit to your Algolia search index by creating a data source,
and provide one or more prompts.
With that, the toolkit lets you generate and manage the responses.
Concepts
You can manage your data sources, prompts, and responses in the Algolia dashboard or with the GenAI Toolkit API.
Data sources
Data sources gather contextual information needed by the toolkit to generate relevant responses for a user query. You can create data sources by choosing an Algolia search index, which can be refined with filters.
Your data sources will be stored in the search index: algolia_rag_data_sources.
Prompts
A prompt gives instructions to the LLM explaining what you want it to do with the user input and the context Algolia Search will provide. These instructions will be included in the final prompt given to the LLM by the GenAI Toolkit. You can use prompts to provide the LLM with guidance on tone, topics to highlight/avoid, and other instructions for how to transform and format the output.
Your prompts will be stored in the search index: algolia_rag_prompts.
Responses
Once you have a data source and a prompt set up, you can refine them with additional filters and run a query. The GenAI Toolkit uses this query to gather relevant context from the connected data source and combine all the information in a prompt processed by the LLM, before returning the generated response.
To test your configuration before saving pre-defined responses, use the response playground.
Your responses will be stored in the search index: algolia_rag_responses.
Next steps
To start using the Algolia GenAI Toolkit:
