After data is uploaded, you can start a Q&A test. The OpenSearch console provides the Q&A Test page for you to run Q&A tests. After you enter a question on this page, the large language model (LLM) generates a result for the question.
This topic describes how to run a Q&A test on the Q&A Test page in the OpenSearch console. To run a Q&A test, you need to only enter your question and then click Send.
Procedure
1. Log on to the OpenSearch console and switch to OpenSearch LLM-Based Conversational Search Edition in the upper-left corner. In the left-side navigation pane, click Instance Management. On the page that appears, find the instance in which you want to run a Q&A test and click Manage in the Actions column. On the details page of the instance, click Q&A Test in the left-side navigation pane.
Q&A parameters | |||||
Parameter | Type | Required | Valid value | Default value | Description |
model | String | Yes | Qwen | The LLM used for this search. Valid values: Qwen opensearch_llama2_13b opensearch_llama2_13b-beta opensearch-falcon-70b | |
Multi-round Conversations | Boolean | No | false | false: disables the multi-round conversation feature. true: enables the multi-round conversation feature. The system returns results based on the recent n rounds of conversations. sessionId: the source of the conversation. The system returns results based on the context of the conversations with the same source. | |
2. Configure the Q&A parameters based on your search requirements, enter a question in the dialog box, and then click Send. You can also click Advanced Parameters to configure prompt parameters, document retrieval parameters, manual intervention parameters, reference image parameters, and other parameters.
Advanced parameters | |||||
Parameter | Type | Required | Valid value | Default value | Description |
attitude | String | No | normal | The tone of the conversation. Default value: normal. Valid values: normal polite patience | |
rule | String | No | simple | The level of detail in the conversation. Default value: detailed. Valid values: detailed stepbystep | |
noanswer | String | No | sorry | The information returned if the system fails to find an answer to the question. Default value: sorry. Valid values: sorry: Sorry, I cannot answer your question based on known information. uncertain: I don't know. | |
language | String | No | Chinese | The language of the answer. Default value: Chinese. Valid values: Chinese English Thai Korean | |
role | Boolean | No | true | Specifies whether to enable a custom role to answer the question. | |
role_name | String | No | AI Assistant | The custom role. Example: AI Assistant. | |
out_format | String | No | table | The format of the answer. Default value: text. Valid values: text table list markdown | |
history_max | INT | No | The maximum number of rounds of conversations based on which the system returns results. You can specify 20 rounds at most. | ||
filter | String | No | The filter that is used to retrieve documents by the category field. By default, this parameter is left empty. | ||
sf | Float | No | [0,+∞). The greater the value is, the less relevant the retrieved documents are. |
Default value for text retrieval: 1.3. Default value for intervention data retrieval: 0.3. Default value for image retrieval: 1.0. | The threshold for determining whether a retrieved document is relevant during vector-based retrieval. |
top_n | INT | No | (0, 50] | 5 | The number of documents to be retrieved. |
formula | String | No | The formula that is used to sort the retrieved documents. By default, documents are retrieved based on vector similarity. | ||
operator | String | No | AND | The operator between text tokens. Default value: AND. | |
return_hits | Boolean | No | false | Specifies whether to return the document retrieval results. | |
csi_level | String | No | none | The configurations for content moderation. none: The content is not moderated. loose: The content is moderated. If Content Moderation detects sensitive content, the content is blocked and no result is returned. strict: The content is moderated. If Content Moderation detects sensitive content or content that may be sensitive, the content is blocked and no result is returned. | |
For more information about Q&A parameters, see Q&A parameters.
You can also run a Q&A test by calling API operations or using OpenSearch SDKs.
3. The following figure shows an example of Q&A tests.
