Skip to main content

Aurix Project: Task Checklist

This document outlines the tasks required to build the Aurix application, based on the voice-first architecture.

Phase 0: Core Setup & Environment​

  • Initialize Electron Forge project with Vite + TypeScript + React.
  • Establish secure IPC bridge between main and renderer processes using contextBridge.
  • Set up basic UI shell with Tailwind CSS and initialize shadcn/ui.
  • Integrate electron-store for local encrypted storage (instead of SQLite).
  • Set up a basic logging service for debugging.

Phase 1: Hybrid Voice Capture & Transcription​

  • Implement the core HybridTranscriptionService in the main process to manage transcription sources.
  • Implement connectivity monitoring to switch between online/offline modes automatically.
  • Renderer Process: Implement the WebSpeechHandler to manage the Web Speech API.
  • Main Process: Implement the Vosk model manager to download and load the local model (stubbed due to native module issues).
  • Main Process: Integrate the Vosk engine for offline transcription (stubbed due to native module issues).
  • Create the IPC channels for the renderer to start/stop transcription and for the main process to receive results from both sources.
  • Implement the UI for recording and displaying the live transcript from the hybrid service.
  • Add a "Privacy Mode" toggle in the UI that allows the user to force offline transcription.

Phase 2: AI Workflow (LangGraph)​

  • Define the WorkflowState object that will be passed through the graph.
  • Build the transcription_node to handle the incoming data from the HybridTranscriptionService.
  • Build the analysis_node using a local LLM (via Ollama) to classify content and calculate a complexity score.
  • Build the document_generation_node to generate structured Markdown from the transcript.
  • Build the diagram_generation_node to detect diagram descriptions and generate Mermaid syntax.
  • Build the assembly_node to combine all parts into a final document.
  • Build the cognitive_load_index_node to calculate the final θ score.
  • Wire all nodes together in a LangGraph workflow with appropriate conditional logic for error handling.

Phase 3: Core UI Implementation​

  • Build the main application layout (e.g., sidebar for sessions, main view for content).
  • Implement the real-time UI for an active recording session, showing the live transcript and the progressively generated document.
  • Create the dashboard view for browsing and managing past sessions/documents.
  • Implement the UI for displaying the Cognitive Load Index and its historical trends.
  • Build the user feedback mechanism (e.g., a slider) to capture subjective workload.

Phase 4: Finalization & Stretch Goals​

  • Implement the optional authentication flow using Auth0 for cloud features.
  • Implement the data synchronization service using PouchDB/CouchDB for authenticated users.
  • Add support for additional recording modes (PTT, VAD).
  • Implement advanced voice commands.
  • Build out the settings page (e.g., select microphone, choose AI models, configure hotkeys).
  • Test and finalize the production build and auto-update process.