Your life already generates telemetry. LifeOS turns that telemetry into reflection.
A fully local AI cognition layer for behavioral tracking, reflective analysis, and context-aware daily intelligence powered by Ollama, FastAPI, and local memory persistence.
LifeOS is not a cloud productivity app. It is a private, self-hosted behavioral interface designed to observe patterns across routines, synthesize context over time, and generate adaptive insights entirely on your machine.
No subscriptions. No cloud storage. No surveillance.
Just local intelligence.
Most trackers stop at: counting visualizing reminding
LifeOS attempts something different:
contextual reasoning behavioral interpretation adaptive reflection temporal memory daily prioritization
Instead of simply storing metrics, the system builds continuity across time using rolling contextual memory windows generated from historical logs.
The result is closer to:
a personal cognitive interface
than a traditional productivity dashboard.
Powered entirely through Ollama using locally hosted LLMs.
Your: health data financial tracking workouts behavioral logs reflections
never leave your machine.
LifeOS automatically compiles rolling context windows from previous daily logs.
This allows the system to: remember trends maintain continuity generate context-aware insights avoid isolated one-shot analysis
Lightweight and resilient Python backend using: FastAPI Pydantic validation asynchronous routing structured response handling
LLMs hallucinate formatting.
LifeOS compensates using: deterministic RegEx extraction fallback handling validation layers crash resilience
to ensure stable structured outputs.
A dependency-light dark mode interface designed to feel tactile and responsive.
Features: animated transitions modal workflows responsive mobile access dashboard-style telemetry input lightweight architecture
Daily outputs are automatically archived into:
logs/daily.mdThis acts simultaneously as: human-readable journaling exportable behavioral history AI memory substrate
| Layer | Technology |
|---|---|
| Frontend | Vanilla HTML/CSS/JS |
| Backend | FastAPI |
| Validation | Pydantic |
| AI Runtime | Ollama |
| LLM | qwen3:8b |
| Storage | Markdown Logs |
| Serving | Uvicorn |
lifeos_webui/
│
├── index.html
├── lifeos.html
├── main.py
├── logs/
│ └── daily.md
├── README.md
├── LifeOS-main.zip
└── mistral-7b-instruct-v0.2.Q4_K_M.gguf┌──────────────────────┐
│ Frontend UI │
│ HTML • CSS • JS │
└──────────┬───────────┘
│
▼
┌──────────────────────┐
│ FastAPI Backend │
│ Validation & Routing │
└──────────┬───────────┘
│
▼
┌──────────────────────┐
│ Context Builder │
│ Rolling Memory Logs │
└──────────┬───────────┘
│
▼
┌──────────────────────┐
│ Ollama API │
│ qwen3:8b │
└──────────┬───────────┘
│
▼
┌──────────────────────┐
│ Structured AI Output │
│ Reflection + Actions │
└──────────────────────┘git clone https://github.com/swappy-ops/lifeos_webui.git
cd lifeos_webuiDownload and install Ollama:
Verify installation:
ollama --versionollama pull qwen3:8bMake sure the Ollama server is running in the background.
pip install fastapi uvicorn pydantic requestsOptional but recommended:
python -m venv venv
source venv/bin/activateWindows:
venv\Scripts\activateuvicorn main:app --host 0.0.0.0 --port 8000 --reloadOpen:
http://localhost:8000Because the server binds to:
0.0.0.0you can access the dashboard from devices on the same network.
Example:
http://192.168.1.50:8000-
Enter: sleep calories steps spending workouts notes
-
Add gym exercises dynamically
-
Click:
Analyze Day- LifeOS: builds memory context retrieves historical patterns queries the local LLM validates outputs generates structured analysis archives the result
LifeOS is intentionally local-first.
No: cloud APIs analytics tracking telemetry uploads external databases
Your behavioral data remains fully under your control.
LifeOS is currently an experimental cognitive systems project exploring: local AI interfaces behavioral reflection systems temporal memory quantified self workflows AI-assisted self-regulation
The architecture is intentionally lightweight and rapidly iterable.
SQLite persistence structured memory indexing visualization graphs semantic tagging streaming responses modular prompt templates
embeddings-based retrieval longitudinal pattern analysis emotional trend classification anomaly detection adaptive recommendations session summarization
multimodal input voice journaling calendar integration financial categorization autonomous planning agents local vector memory predictive behavioral modeling
Add screenshots and gifs here.
Suggested additions: dashboard overview mobile interface AI analysis output telemetry workflow
Modern productivity systems optimize for: engagement streaks dopamine loops
LifeOS explores a different direction:
software as reflection.
Not performance theater. Not hustle dashboards.
A quieter system that helps interpret continuity across behavior, thought, and time.
Contributions, experiments, and architectural discussions are welcome.
Areas especially relevant: local AI systems memory architectures HCI and UX quantified self tooling behavioral analysis agentic workflows
MIT License
Built by swappy-ops
LifeOS began as a daily tracker.
It is slowly evolving into something closer to:
a personal cognition substrate.
And that distinction changes everything.