Hero AI

hero ai is swimlane’s collection of artificial intelligence features that help orchestrators and analysts in their day to day usage of turbine the features of hero ai largely take advantage of the swimlane private large language model (llm) the swimlane llm is a proprietary llm hosted by swimlane and tuned to our users’ needs the swimlane llm is hosted in swimlane cloud and offers a level of security and privacy not available with public models your data never leaves the swimlane cloud it is not sent to a third party, it is not stored in shared log files, and it is not used to train or fine tune our model to enable hero ai in your environment, contact swimlane support artificial intelligence data privacy and security swimlane does not collect or store any sensitive customer data in centralized storage locations only metadata about model usage and performance is retained centrally customer data processed by ai or machine learning (ml) models is exclusively stored in the customer’s own dedicated database instance which is logically separated from other customer instances all prompts and features involving customer data are processed using a private swimlane llm swimlane does not use customer data to train or fine tune the swimlane llm customers must “opt in” to use hero ai in their instance of turbine individual risk assessments are conducted on all ai/llm projects before deployment multi model error handling and fallback logic hero ai within turbine may fall back to use a different model in certain circumstances, such as hitting a rate limit, temporary connectivity issues or temporary high load hero ai uses three anthropic claude models, configured as primary, secondary and tiertiary the system is designed gracefully to handle model specific errors with the following logic priority model id description primary claude sonnet 4 5 v1 0 preferred for optimal response quality secondary claude sonnet 4 v1 0 backup model for reliability tiertiary claude sonnet 3 5 v2 0 fallback model for degraded mode error handling and fallback logic the system is designed to handle model specific errors gracefully exception type handling strategy next step timeout retry with the same model retry up to maxattempts = 3 rate limit / token exceed switch to next available model arn retry with fallback model (if available) other exceptions log and return the error no retry hero ai features hero ai includes a number of features meant to help analysts and orchestrators get the most out of turbine and better secure your companies' infrastructure hero ai features include docid e8xqth18xylh6cl6hkbh docid\ k3hu izfx6ogs2lv0fi3t docid\ bbcgtaqyctwozro7p0x5m https //docs swimlane com/solutions/crafted ai prompts docid\ xcprtqgxdugg13ytveies