Doze: A Local AI Thought Inbox That Lets My To-Do List Rest
This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady What I Built I built Doze , a calm local thought inbox for scattered notes, reminders, bookmarks, and half-finished ideas. You paste a messy list into Doze and it creates a smaller brief with four lanes: Now
Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady Submission 🦥
This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady
I built Doze, a calm local thought inbox for scattered notes, reminders, bookmarks, and half-finished ideas.
You paste a messy list into Doze and it creates a smaller brief with four lanes:
It saves the repeated micro-decision of reopening the same note and asking, “What should I do with this?”
Doze runs locally in a browser at http://127.0.0.1:8765.
I tested it with four loose notes: sending a draft, reading an article, waiting for an invoice, and planning a dinner. Doze sorted them into Now, Reference, Waiting, and Later.
Doze uses a small Python server and a single HTML page.
Its local AI path is designed for Ollama with open-weight models such as llama3.2:1b or qwen2.5:1.5b. When Ollama is available, Doze sends the private inbox only to the local Ollama server at 127.0.0.1:11434 and asks the selected model for structured categories.
I also built a local rule-based fallback. If Ollama is unavailable, Doze still organizes notes without sending them to a cloud service.
A thought inbox can include private work reminders, personal relationships, health notes, money errands, and unfinished ideas. That information should not have to leave someone’s laptop just to be sorted into a few calm categories.
Open-weight models make Doze local-first. A user can choose a model that fits their computer, swap models later, run without an account, and use the tool without an internet connection after setup.
A closed API would make this personal automation depend on a third party, recurring costs, and an internet connection. With local inference, the useful intelligence remains under the user’s control.
Doze does not try to create more work. It reduces the friction of deciding which thoughts deserve attention right now.
For my sample inbox, it turned four unstructured notes into one immediate action, one waiting item, one reference, and one task that could safely wait.