Upsoma's existing curriculum vitae, GitHub, learner-answer, skill-identity, skill-extraction and goal-categorisation decisions now run through clearly named native agents. Their prompts live beside the agents by responsibility, while onboarding and telemetry retain their own state and persistence rules. This establishes the execution structure for upcoming job-posting evidence and personalised learning-path work without changing current learner behaviour.
Platform infra
Model-powered decisions now share one native agent structure
Upsoma's existing model-powered decisions now use clearly named native agents and prompt modules, giving onboarding and skill evidence one consistent execution structure.
Version v0.1.65