This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend
What I Built
I built Journal Buddy, a privacy-first AI journal analyzer for my friend Alex. Alex loves journaling daily to manage stress but often struggles to review past entries to identify emotional trends or get a quick recap of the week. However, Alex is extremely privacy-conscious and outright refuses to use cloud-based AI tools (like ChatGPT or Claude) to analyze their deeply personal thoughts.
Journal Buddy solves this by processing all journal entries completely offline on their local machine. It provides a quick summary of their week and an emotional sentiment breakdown, allowing them to reflect without ever compromising their privacy.
Demo
While the tool I gave my friend runs 100% offline via CLI for maximum privacy, I also deployed a web version using Gradio on Hugging Face Spaces so you can try it out!
👉 Try the Live Web App Here!
(Because the web app uses Hugging Face's ZeroGPU, it might take a couple of seconds to boot the first time you run it!)
Code
You can find the entire code repository below. It is incredibly simple to set up and run locally on your own machine!
| title | Journal Buddy |
|---|---|
| emoji | 🦀 |
| colorFrom | blue |
| colorTo | indigo |
| sdk | gradio |
| sdk_version | 4.43.0 |
| app_file | app.py |
| pinned | false |
Journal Buddy 📔
A completely offline, privacy-first journal analyzer powered by Open-Source AI.
Built for the Hacktoberfest Weekend Challenge: Build for a Friend.
Why this exists
Many people love journaling for mental health but want a quick summary and emotional check-in at the end of the week. However, sending deeply personal journal entries to cloud APIs (like ChatGPT or Claude) is a massive privacy risk.
Journal Buddy solves this by using Hugging Face open-weight models (transformers) to process your journal entries 100% locally on your own machine. Your data never leaves your hard drive.
Features
- 100% Local Inference: Runs entirely on your CPU/GPU without internet access (after initial model download).
- Automated Summarization: Condenses long journal entries into a short recap.
- Sentiment Analysis: Detects the overall emotional vibe…
How I Built It
I built this using Python and the incredible open-source Hugging Face transformers library. The core project runs inference entirely locally using open-weight models, meaning no cloud APIs are ever pinged during text analysis.
-
Summarization: I used a distilled version of BART (
sshleifer/distilbart-cnn-12-6), which is small enough to run on a standard laptop CPU without significant delays. -
Sentiment Analysis: I used the robust
distilbert-base-uncased-finetuned-sst-2-englishmodel. -
UI: I built a gorgeous CLI using
Richfor local use, and a web interface usingGradiofor the live demo.
Why Does Open Innovation Matter?
Open innovation is the only reason this project could exist for Alex. A closed API would require sending highly personal, vulnerable journal entries to a corporate server for analysis. This was a complete dealbreaker for my friend.
By utilizing open-weight models and running local inference, open-source AI gave us the power of advanced Natural Language Processing without the privacy trade-offs. It runs entirely on their own laptop, costs absolutely $0 per query, and guarantees that Alex's data never leaves their hard drive. It empowered me to build a personalized, AI-driven tool completely tailored to their strict privacy requirements.
Prize Categories
- Hugging Face (Building with open-source AI models/frameworks)
- Pinata (Local inference / keeping things offline)
Top comments (0)