Debug and profile agentic app experiences with Instruments
AI & Machine Learningiosipadosmacos
Key Points
Covers the enhanced FoundationModels instrument in Xcode to inspect behavior and optimize the performance of agentic flows.
Explains how to inspect prompts, analyze latency, and trace control flow in advanced use cases that leverage multiple LanguageModelSessions and profiles.
The session moves through LLM app development mindset, Inspect and diagnose an agentic experience, Recording a trace with Instruments, Navigating the Instruments UI, Performance metrics.
Key concepts include Instruments, FoundationModels, LanguageModelSessions, Foundation Models, Private Cloud Compute, GenerateCraftIdeaTool, SwitchToTutorialModeTool, LLMs.
Platform coverage: ios, ipados, macos.
Condensed Flow
01
Inspect and diagnose an agentic experience:
Focuses on Private Cloud Compute.
02
Recording a trace with Instruments:
Focuses on Instruments.
03
Navigating the Instruments UI:
Focuses on Instruments, Foundation Models.
More Detail
Additional details
Detailed flow: LLM app development mindset -> Inspect and diagnose an agentic experience -> Recording a trace with Instruments -> Navigating the Instruments UI -> Performance metrics.
APIs and concepts to recognize: Instruments, FoundationModels, LanguageModelSessions, Foundation Models, Private Cloud Compute, GenerateCraftIdeaTool, SwitchToTutorialModeTool, LLMs.
Implementation focus: GenerateCraftIdeaTool, SwitchToTutorialModeTool, Foundation Models, LLMs.