ServiceNow Otto for AI Search properties
Summarize
Summary of ServiceNow Otto for AI Search properties
This documentation outlines key system properties for configuring and modifying the behavior of ServiceNow Otto for AI Search Genius Results, specifically for Now Assist Q&A functionality. These properties allow ServiceNow customers to control AI Search’s use of caching, query eligibility, timeout settings, and content sanitization when generating answers from Knowledge base articles using the Now LLM Service.
Show less
Key Properties and Their Practical Use
- snaisassist.semanticcachemode: Controls the operational mode for caching and AI service usage in Now Assist Q&A Genius Results. It has three modes:
- off: Uses only the first-level cache and the Now LLM Service to generate answers. Dynamic Translation bypasses caches and queries the LLM directly.
- offline: Uses first- and second-level caches exclusively, without querying the LLM Service. Suitable when offline operation is needed, but no new answers are generated if not found in cache.
- online: Combines first- and second-level caches with the Now LLM Service to generate answers if caches miss. This provides the most comprehensive answer retrieval.
- snaisassist.ullmapitimeout: Defines the maximum wait time (in seconds) for the LLM Service to generate Knowledge base article answers. Default is 120 seconds. This helps manage performance expectations and system responsiveness.
- snaisassist.uquestionregex: A case-insensitive Java regular expression that restricts which search queries are eligible for Now Assist Q&A answer generation. For example, limiting queries to those starting with specific interrogative words. If unset, all supported searches are eligible. Helps refine and optimize AI answer generation by query type.
- snaisassist.usanitizehtml: Boolean option to remove HTML tags from knowledge article content before sending to the LLM for answer generation. Default is false. Enabling this can improve answer clarity by sending cleaner text to the LLM.
Practical Implications for ServiceNow Customers
- By configuring these properties, customers can tailor AI Search behavior to balance performance, accuracy, and operational constraints such as offline availability or translation needs.
- Adding or modifying these system properties requires manual entry in the System Property [sysproperties] table, as most are not present by default.
- Understanding and using the semanticcachemode property is critical for optimizing answer retrieval strategies, especially in environments with varying connectivity or translation scenarios.
- Using the uquestionregex property enables customers to limit AI answer generation to relevant queries, improving system efficiency and user experience.
- The HTML sanitization option helps ensure that the LLM receives clean input, potentially enhancing answer quality.
System properties that you can use to modify behavior for ServiceNow® Otto for AI Search Genius Results.
| Property | Description |
|---|---|
| sn_ais_assist.semantic_cache_mode | Operational mode for the Now Assist Q&A Genius Results caches. Use this mode to control how AI Search uses the caches and the Now LLM Service when looking for Now Assist Q&A Genius Result answers. For details about how AI Search caches Now Assist Q&A Genius Result answers, see Caching for Now Assist Q&A Genius Results.
|
| sn_ais_assist.u_llm_api_timeout | Maximum time in seconds to wait for Knowledge base articles Genius Result answer generation results from the LLM.
|
| sn_ais_assist.u_question_regex | Case-insensitive Java regular expression pattern that defines searches eligible for Now Assist Q&A answer generation. As an example, the pattern ^(what|where|when|why|who|whom|which|how)\b restricts eligibility to search queries that begin with one of the
specified keywords. When this system property has a value defined, only searches in supported applications that match the specified pattern are eligible for Now Assist Q&A answer generation. When no value is defined, all searches in supported applications are eligible for Now Assist Q&A answer generation.
To learn about Java regular expression pattern syntax, see the Javadoc for the java.regex.util.Pattern class. |
| sn_ais_assist.u_sanitize_html | Option to remove HTML tags from knowledge article text before sending to the LLM for Knowledge base articles Genius Result answer generation.
|