macOS · SwiftUI
Suggst combines reference context, meeting-specific context and transcribed speech to prepare an answer after an explicit user action.
Overview
The problemA question in a meeting may require combining reference information with discussion-specific details while the conversation continues.
The solutionSuggst pairs two saved contexts with authorized audio capture. On a manual trigger, it transcribes speech, extracts the request and prepares a contextual answer.
Save reusable reference context and information specific to the current conversation. Long-lived reference information and discussion-specific details are stored separately.
1 / 4: Prepare the contexts. Previous step / Next step
A SwiftUI-driven pipeline processes audio locally, extracts the request and calls OpenAI with selected context.
SwiftUI
Interface and explicit user actionAVFoundation, ScreenCaptureKit
Microphone, system audio and identified clipsSwift, WhisperKit
Final transcription and request extractionOpenAI Responses API
Answer grounded in the request and contextsAsynchronous operations may finish out of order or deliver duplicate events.
Per-question identities, sealed results and idempotent ingestion that rejects inconsistent events.
Microphone and system audio rely on separate providers and permissions.
Two coordinated native providers, with an explicit state when system capture is unavailable.
Saved contexts and conversation content may contain sensitive information.
Local document storage, an API secret in Keychain and targeted context sent to OpenAI.
Main technical stack
Context editing, listening, preparation and answer views brought together in a SwiftUI interface.
Audio capture, WhisperKit transcription, local extraction and OpenAI generation tied to a manual trigger.
Developer cockpit, structured traces, test corpora and build scripts included in the repository.