Site-specific connection info and project settings for the target compute environment.
_remote_info/
├── flow/ # e.g. Supercomputer "Flow"
│ ├── user_info.md # SSH connection, remote working directory
│ ├── command_list.md # Site-specific commands (qsub, pjsub, etc.)
│ ├── sample_bash.sh # Job script sample
│ ├── load_custom_module.md # Module loading instructions
│ └── node_resource_groups.md # Resource group specs (required)
│
└── miyabi/ # Another site
└── (same structure)
Markdown table with resource group details for budget tracking:
- Group name (cx-small, fx-large, etc.)
- Min/max nodes, CPU/GPU cores, memory
- Max walltime (default and limit)
- Rate (points/sec)
- Copy from the site's official documentation
- Convert to markdown table
- Include the billing formula (e.g.
rate × GPU_count × seconds)
- PM: Resource allocation strategy
- PG: Job submission (selecting the right resource group)
- Budget tracking: Cost estimation from ChangeLog.md
- **SSH**: username@supercomputer.example.jp
- **Working directory**: /data/username/VibeCodeHPC/project_name/- File permissions:
chmod 600 - Private keys managed via ssh-agent (never stored here)