# 🔄 Step-Level Retry Action Retries a step while preserving its full context and environment. ## 🚀 Usage ### Basic Usage Retry failed step up to 3 times (default): ```yaml steps: - uses: ultralytics/actions/retry@main with: run: python train.py ``` ### Advanced Usage Full configuration with custom retries, timeout, backoff, and jitter: ```yaml steps: - uses: ultralytics/actions/retry@main with: run: | python setup.py install pytest tests/ retries: 2 # Retry twice after initial attempt (3 total runs) timeout_minutes: 30 # Total timeout across all attempts retry_delay_seconds: 10 # Base delay between retries backoff: exponential # exponential (10s, 20s, 40s, ...) or fixed jitter: true # Randomize delay to 80-120% to avoid thundering herd shell: bash # Use python or bash shell ``` ### Python Shell Example ```yaml steps: - uses: ultralytics/actions/retry@main with: shell: python retries: 5 run: | import requests response = requests.get('https://api.example.com/data') response.raise_for_status() ``` ## 📋 Inputs | Input | Description | Required | Default | | --------------------- | ------------------------------------------------------------ | -------- | ------------- | | `run` | Command to run | Yes | - | | `retries` | Number of retry attempts after initial run | No | `3` | | `timeout_minutes` | Maximum total time in minutes, checked between attempts | No | `360` | | `retry_delay_seconds` | Base delay between retries in seconds | No | `10` | | `backoff` | Backoff strategy: `exponential` (base \* 2^n) or `fixed` | No | `exponential` | | `jitter` | Randomize delay to 80-120% of value to avoid thundering herd | No | `true` | | `shell` | Shell to use (`bash` or `python`) | No | `bash` | ## ✨ Features - Preserves environment variables and step context - Exponential backoff with ±20% jitter (best-practice defaults) - Configurable total timeout across all attempts (individual sleeps auto-capped to remaining budget) - GitHub Actions grouping for retry attempts - Supports both Bash and Python shells