foresight-cli 0.1.2

Predict system resource exhaustion before it happens.

🔭 Foresight

Predict system resource exhaustion before it happens.

---

Traditional monitoring tools tell you what is happening. Foresight tells you what will happen — before it does.

---

📖 What is Foresight?

Foresight is a lightweight, CLI-native system resource forecaster built in Python. It collects CPU, RAM, and disk metrics locally, stores them in SQLite, and applies time-series ML models (ARIMA, Holt-Winters, Ensemble) to predict resource exhaustion before it happens.

No cloud. No dashboard. No subscription. Just your terminal.

foresight healthcheck --horizon 1h

System Health Check — next 1h via ENSEMBLE

  ✅  cpu_percent    OK          now: 18.4%   threshold: 75.0%   trend: stable
  🚨  ram_percent    CRITICAL    now: 87.4%   threshold: 80.0%   trend: rising
  ✅  disk_percent   OK          now: 17.4%   threshold: 85.0%   trend: stable

  🚨  Overall: CRITICAL

---

✨ Features

Feature	Description
Live Monitor	Real-time resource usage with color-coded health indicators
ARIMA Forecasting	AutoRegressive model for trend-based prediction
Holt-Winters Forecasting	Exponential smoothing — weights recent data more heavily
Ensemble Forecasting	Blends ARIMA + Holt-Winters for more robust predictions
Threshold Alerting	Warns when a metric is predicted to breach safe limits
Health Check	Full system overview across all metrics in one command
ASCII Charts	Visualize metric history directly in the terminal
Smart Defaults	Per-metric health thresholds based on industry standards
Horizon-based	Think in time: --horizon 30m, --horizon 1h, --horizon 2h

---

🚀 Quick Start

Install via pip

pip install foresight-cli

That's it. The foresight command is now available globally.

Collect Data

# Collect 60 snapshots every 60 seconds (1 hour of data)
foresight collect

# Quick test — 10 snapshots every 5 seconds
foresight collect --rounds 10 --interval 5

⚠️ Important: Forecasting requires historical data. For reliable 30-minute forecasts, collect at least 2-3 hours of data first. The more data, the better the predictions.

Check Your System

foresight healthcheck --horizon 30m

---

📋 All Commands

foresight collect

Collect system metrics and save to local SQLite database.

foresight collect                           # 60 rounds × 60s = 1 hour
foresight collect --rounds 30 --interval 5  # 30 snapshots every 5 seconds

foresight status

Single live snapshot with color-coded health indicators.

foresight status

foresight watch

Live refreshing monitor. Clears screen and updates every N seconds.

foresight watch                    # every 5s, 50 rounds
foresight watch --interval 2 --rounds 30

foresight show

Display recent snapshots as a color-coded formatted table.

foresight show               # last 10 snapshots
foresight show --limit 50    # last 50 snapshots

foresight chart

ASCII line chart of any metric rendered directly in the terminal.

foresight chart                               # default: cpu_percent
foresight chart --metric ram_percent
foresight chart --metric disk_percent --limit 100

Available metrics: cpu_percent, ram_percent, ram_used_mb, disk_percent, disk_used_gb

foresight forecast

Forecast future resource usage over a time horizon.

foresight forecast                                     # CPU, 30m, ensemble
foresight forecast --metric ram_percent --horizon 1h
foresight forecast --metric cpu_percent --horizon 2h --model arima
foresight forecast --metric disk_percent --model holtwinters

Models:

• ensemble (default) — Blends ARIMA + Holt-Winters. Most robust.
• arima — Better for data with clear, consistent trends
• holtwinters — Better when recent data matters more than history

foresight alert

Check if a metric is predicted to breach its health threshold.

foresight alert                                         # CPU, 30m, smart threshold
foresight alert --metric ram_percent --horizon 1h
foresight alert --metric cpu_percent --threshold 70     # custom threshold

foresight healthcheck

Full system health check across CPU, RAM, and Disk at once.

foresight healthcheck              # 30m horizon, ensemble
foresight healthcheck --horizon 1h
foresight healthcheck --model arima

---

🎯 Smart Health Thresholds

Foresight uses industry-standard thresholds by default:

Metric	Warning	Critical	Reasoning
CPU	75%	90%	Above 75% sustained → thermal throttling risk
RAM	80%	90%	Above 80% → heavy memory swapping begins
Disk	85%	95%	Above 85% → write performance degrades

Override any threshold with --threshold flag.

---

📊 How Much Data Do You Need?

Forecast Horizon	Minimum Snapshots	Recommended
30 minutes	30 snapshots	2–3 hours of data
1 hour	60 snapshots	4–6 hours of data
2 hours	120 snapshots	8–12 hours of data

Rule of thumb: Collect at least 3–5× more history than your forecast horizon for reliable predictions.

---

🧠 How Forecasting Works

Foresight uses three time-series forecasting models:

ARIMA (AutoRegressive Integrated Moving Average) Predicts future values based on patterns in past values and past prediction errors. Parameters: AR (past values), I (differencing to remove trend), MA (error correction). Best for data with a clear, consistent long-term trend.

Holt-Winters Exponential Smoothing Weights recent data more heavily than older data — exponentially decreasing influence. The model continuously updates its estimate of level and trend as new data arrives. Best for data where recent behaviour matters more than long-term history.

Ensemble (Default) Averages ARIMA and Holt-Winters predictions step by step. Two models with different assumptions tend to make different errors — averaging cancels individual mistakes out. Consistently more robust than either model alone.

---

🛠️ Tech Stack

Layer	Technology
Language	Python 3.10+
Metrics collection	psutil
Local storage	SQLite (built-in)
Forecasting	statsmodels (ARIMA + Holt-Winters)
Data handling	pandas
CLI framework	Typer
Terminal UI	Rich
ASCII charts	plotext
Packaging	setuptools + pyproject.toml

---

📁 Project Structure

foresight/
├── foresight/
│   ├── __init__.py
│   ├── collector.py    # psutil metrics collection
│   ├── storage.py      # SQLite database layer
│   ├── forecaster.py   # ARIMA, Holt-Winters, Ensemble, alerting
│   └── cli.py          # Typer CLI — all 8 commands
├── tests/
├── data/               # SQLite database (gitignored, created locally)
├── pyproject.toml
├── requirements.txt
├── CONTRIBUTING.md
├── LICENSE
└── README.md

---

🤝 Contributing

Contributions are welcome. Please read CONTRIBUTING.md first.

---

🗺️ Roadmap

• Export forecasts to CSV
• Cron job integration for automated collection
• Network metrics (bandwidth usage)
• Multi-machine support
• Weighted ensemble based on historical model accuracy
• GitHub Actions CI pipeline
• Seasonal detection when sufficient data is available

---

👨‍💻 Author

Built by Rishi Garg 2nd Year AIML Student, Delhi Technological University Member, AIMS-DTU — AI/ML Society of DTU

Built as a portfolio and learning project with a focus on time-series forecasting, system programming, and open-source contribution.

---

📄 License

MIT License — see LICENSE for details.

---

If this helped you, give it a ⭐ on GitHub