US 12,443,515 B1
Context engine test generation
Rajkumar Janakiraman, Sammamish, WA (US)
Assigned to Dropbox, Inc., San Francisco, CA (US)
Filed by Dropbox, Inc., San Francisco, CA (US)
Filed on Jun. 21, 2024, as Appl. No. 18/749,893.
Claims priority of provisional application 63/632,833, filed on Apr. 11, 2024.
Int. Cl. G06F 11/3668 (2025.01); G06F 8/30 (2018.01)
CPC G06F 11/3684 (2013.01) [G06F 8/30 (2013.01); G06F 11/3688 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
decomposing an intent query into multiple query subcomponents defining discrete portions of function code that, when executed, accomplish corresponding portions of the intent query;
generating, using a context engine machine learning model integrated with a language model, a series of code generation prompts corresponding to the multiple query subcomponents which provide respective instructions to the language model to generate respective test codes corresponding to the discrete portions of function code based on structured programmatic data indicated within the multiple query subcomponents of the intent query;
generating, utilizing the language model integrated with the context engine machine learning model, function code whose execution satisfies the corresponding portions of the intent query;
generating, utilizing the language model, the respective test codes defining function tests for determining that the function code satisfies the multiple query subcomponents of the intent query; and
executing the respective test codes to determine that the function code satisfies the multiple query subcomponents of the intent query.