As We May Think

An implementation of Vannevar Bush's Memex. Simple optional scrolling RSS feed reader (no CSS or real parsing!), note taking and retrieval, and a local LLM assistant that can answer questions based on your notes.

Structure

The frontend is located in lib, the backend is located in backend. Place LLM GGUFs in backend/models. Backend is a FastAPI project running SQLite as a local database, frontend is Flutter. Descriptions of architecture and features are located in descriptions.

Getting Started

Prerequisites

• Python 3.10+
• Flutter SDK (latest stable)

1. Run the Backend

1. Navigate to the backend directory:

   cd backend

2. Create and activate a virtual environment with uv:

   uv venv
   source .venv/bin/activate

3. Install dependencies and sync environment:

   uv sync

4. Start the backend server:

   uv run uvicorn main:app --reload

   The backend will run on http://localhost:8000 by default.

2. Run the Frontend

1. Return to the project root:

   cd ..

2. Get Flutter packages:

   flutter pub get

3. Run the app (choose your platform):
   • Mobile (Android/iOS):

     flutter run

   • Web:

     flutter run -d chrome

   • Desktop (Linux/Mac/Windows):

     flutter run -d linux
     # or -d macos / -d windows

Vibe Coding

What's more interesting than the software itself is its development process. I pick a feature from the handwritten "roadmap" in descriptions/main.md and specify its functional requirements to an LLM. To meet non-functional requirements, I specify certain implementation details (such as names of API endpoints and their functions) and pitfalls to avoid. I ask the LLM to write a detailed document for the implementation, and I review it. The LLM writes a first draft of these functions and their tests, fully covering the desired behavior. I then iterate through a test-driven development loop until the tests pass, and finally manually QA myself.