LifeLog / README.md
arunsa's picture
Fix Gradio sdk_version to 5.29.0
1caf5dc
|
Raw
History Blame Contribute Delete
2.92 kB
---
title: LifeLog
emoji: πŸ”§
colorFrom: green
colorTo: gray
sdk: gradio
sdk_version: "5.29.0"
app_file: app.py
pinned: false
license: apache-2.0
tags:
- backyard-ai
- thousand-token-wood
- tiny-titan
- off-brand
- best-demo
short_description: A debugger for your life decisions
---
# πŸ”§ LifeLog β€” A Debugger for Your Life Decisions
> `$ git blame my-life-choices`
Your life is code. Decisions are commits. Consequences are bugs or features. **LifeLog is your debugger.**
We make hundreds of decisions but never track them. We repeat the same mistakes, forget what worked, and can't see patterns in our own choices. There's no version control for life β€” until now.
## What It Does
Log a decision β€” *"I turned down a job offer to stay at my current company."*
The AI asks follow-up questions: Why? What were you afraid of? What's the best case? It categorizes the decision (career, financial, health, relationship), predicts potential consequences, and generates a visual **moment card** β€” an illustrated watercolor snapshot of that crossroads in your life.
Weeks later, log the outcome. LifeLog connects it to the original decision. Over time, it spots patterns:
- *"Your career decisions have 80% positive outcomes, but you avoid health decisions until they become emergencies."*
- *"You make impulsive financial decisions on weekends."*
## Features
- **Three input modes**: Type it, speak it, or upload a photo (resignation letter, medical report, apartment listing)
- **AI follow-up conversation**: 3 probing questions that help you think through your decision
- **Smart categorization**: AI-assigned categories, severity scores, and consequence predictions
- **Moment cards**: FLUX.2-generated watercolor illustrations for each decision
- **Git Log timeline**: Visual timeline styled like `git log --oneline --graph`
- **Debug Dashboard**: Category breakdowns, prediction accuracy, behavioral patterns
- **Outcome tracking**: Resolve decisions and compare predictions vs reality
## Tech Stack (All Models ≀ 4B β€” Tiny Titan Eligible)
| Role | Model | Params |
|------|-------|--------|
| Text reasoning | MiniCPM5-1B | 1B |
| Vision (image reading) | MiniCPM-V 2.6 | ~1.3B |
| Voice transcription | Whisper Small | 244M |
| Moment card art | FLUX.1 Schnell | 4B |
**Total: ~6.5B parameters across 4 models.**
## The Debugger Metaphor
| Life | Code |
|------|------|
| Making a decision | `git commit` |
| Decision timeline | `git log` |
| Categories | Git branches |
| Analyzing patterns | Debugging |
| Logging outcomes | Bug/feature resolution |
| Good outcomes | ✨ Features shipped |
| Bad outcomes | πŸ› Bugs found |
## Who It's For
> "I built this for myself β€” a grad student making big life decisions about careers, cities, and relationships with no framework for learning from past choices."
## Built With
Gradio Β· HuggingFace Spaces Β· Zero GPU Β· Python