Memory Bank

The Memory Bank is a file-based context management system designed to give AI agents (like Roo) persistent, mode-specific memory.

Structure

.
├── .roo/
│   └── rules/                  # Mode-specific context loading rules
│       ├── architect-mode.md
│       └── code-mode.md
├── memory/
│   ├── tasks/                  # Active and completed tasks
│   │   └── working-gh1-setup.md
│   ├── projectBrief.md         # The Vision
│   ├── productContext.md       # The Requirements
│   ├── activeContext.md        # The Current Focus
│   ├── systemPatterns.md       # The Standards
│   ├── techContext.md          # The Stack
│   └── architecture_decisions_record.md # The Decisions

How to Use

1. Start a Session: The agent will automatically load the rules from .roo/rules/ corresponding to its active mode.
2. Context Loading: The agent reads the required files as defined in the rule file.
3. Task Management:
   • Create a new task file in memory/tasks/ using the format [status]-gh[issue]-[description].md.
   • Update activeContext.md to reflect the new focus.
4. Completion:
   • Mark task items as done.
   • Update activeContext.md and move the task to closed status (rename file).

Core Files

• projectBrief.md: High-level vision and "vibe".
• activeContext.md: The single source of truth for "what are we doing right now?".
• systemPatterns.md: Coding standards and architectural patterns.