This topic describes how to enable the intelligent dialogue feature for an industry algorithm edition instance.
Enable intelligent dialogue
The industry algorithm edition now supports the intelligent dialogue feature. This feature calls a large model in OpenSearch to generate intelligent answers to user queries.
From the homepage, navigate to Retrieval-Augmented Generation > Intelligent Dialogue. Select an instance and click Add intelligent dialogue to start the configuration.
Configure parameters
When you add dialogue parameters, you must specify a dialogue parameter name and configure large model parameters, context parameters, and more.
-
Dialogue parameter name: A unique name between 1 and 30 characters long. It must start with a letter and can contain letters, digits, and underscores.
-
Large model parameters: Includes Model name, Prompt template, and Model parameters.
-
Model name: Select a large model provided by OpenSearch.
The following four models are currently supported. For more information, see Introduction to Qwen large language models.
Model
Description
OpenSearch-Qwen-Turbo (ops-qwen-turbo)
Uses the Qwen-Turbo large model as the base model. It is fine-tuned with supervised learning to enhance retrieval augmentation and reduce harmful content.
Qwen-Turbo (qwen-turbo)
A large-scale language model that supports Chinese, English, and other languages.
Qwen-Plus (qwen-plus)
An enhanced version of the Qwen large-scale language model that supports Chinese, English, and other languages.
Qwen-Max (qwen-max)
A 100-billion-parameter large-scale language model that supports Chinese, English, and other languages.
-
Prompt template: Select an existing prompt template.
-
Model parameters: Includes parameter name and parameter value fields. You can add, modify, or delete model parameters. The system configures optimal parameters by default, based on best practices.
-
-
Context parameters: Includes Number of input documents and Input fields.
-
Number of input documents: The number of documents to provide as context to the large model. The maximum value is x.
-
Input fields: Includes Field name and Field description. You can add, modify, or delete input fields. You must select up to 20 field names from the existing fields in the instance.
-
-
More parameters: Click More Parameters to display the Field Summary Parameters section, which includes reranking model, chunking model, number of retained chunks, and maximum chunk length.
-
Reranking model: Select a reranking model provided by OpenSearch.
-
Chunking model: Select a chunking model provided by OpenSearch.
-
Number of retained chunks: The maximum number of chunks to retain after a long document is split.
-
Maximum chunk length: The maximum length of each chunk for chunk-based retrieval. Value range: [100, 3000]. Default: 300.
-
View results
After completing the configuration, click Search test. Enter a value for the required raw_query parameter, enable the intelligent dialogue toggle, and then click Search. You can view the output in the Intelligent dialogue results tab.
For more information about the search test, see Search test.