Migrating from V1 to V2

Welcome to dIKta.me V2! If you are a developer, power user, or contributor who spent time in the diktate V1 repository, this guide will help you understand the architectural shifts and what happened to your data.

1. Architectural Paradigm Shift

The biggest change is the underlying technology stack.

V1 Structure: The application was inherently split. It used an Electron (Node.js/React) frontend for the UI and a background Python FastAPI server to handle hardware audio capture and AI model execution.
V2 Structure: The application is now a single, native, compiled C# (.NET 8) process utilizing WinUI 3 for presentation.

This means you no longer have to manage mismatched Python dependencies, pip environments, or debug WebSocket IPC failures between the frontend and the backend. The core business logic (DiktaMe.Core) and UI (DiktaMe.App) run in the exact same memory space.

2. Terminology Changes

Many internal concepts have been rebranded for clarity in V2:

V1 (Legacy) Concept	V2 Equivalent	Notes
`dIKtate`	dIKta.me	The branding has officially dropped the "ate" for ".me".
"Continuous Dictation"	Batch Processing	Audio is now captured into memory and processed in chunks, regardless of length.
"Live Dictation"	Streaming Pipeline	WebSockets are used natively via `IStreamingSTTProvider` (e.g. Deepgram) to return text while you speak.
`config.json`	`settings.json`	Application state and configurations.
`prompts.json`	Dictation Modes (stored in `settings.json`)	V1 system prompts and trigger modes have been unified into customizable CRU objects containing their own LLM/STT selections.

3. Where is My Data?

In V1, configurations were heavily scattered in the installation folder or local AppData depending on the build.

In V2, absolutely everything related to your user state is strictly isolated in a single directory: %APPDATA%\DiktaMe\

This folder contains:

settings.json: This replaces both the old config.json and prompts.json. It stores your hotkeys, API selections, and all 16 custom Dictation Modes.
history.db: A robust SQLite database replacing messy filesystem logs. It tracks your transcription history, metrics, and wallet balances securely.
secrets.dat: Your API keys are no longer stored in plaintext. They are encrypted at rest using Windows DPAPI algorithms.

4. Local Execution (Whisper & Ollama)

If you relied on whisper.cpp or local HuggingFace transformers in V1, the pipeline is significantly more stable now.

Transcription: We utilize Whisper.net configured to automatically use the Vulkan runtime. It achieves near 7x parity with pure CUDA deployment without requiring you to install gigabytes of NVIDIA C++ runtime dependencies manually.
Formatting: We now officially support Ollama natively. The V2 app handles keep-alives, model pulls, and version parsing directly via their local HTTP API. You no longer need to manage a separate Python environment just to run local language models.

For detailed API documentation on adding new providers to V2, see the api/stt-providers.md and api/llm-providers.md developer guides.