Intelligent Memory System

BB's memory system enables your LLM to remember important information across collaborations and projects. With automatic memory management and user control, BB builds knowledge that grows smarter with every interaction.

Why Memory Matters

Traditional LLM interactions start fresh every time. BB's memory system changes this by allowing the LLM to remember and recall information across collaborations, building a persistent knowledge base that improves over time.

Without Memory

  • Repeat project context in every collaboration
  • Re-explain preferences and conventions
  • Lose insights when collaborations end
  • No knowledge sharing across projects

With BB Memory

  • LLM remembers your project structure
  • Recalls your coding style and preferences
  • Builds knowledge from past collaborations
  • Shares insights across your team

Multi-Level Memory Architecture

BB organizes memories into three levels, each serving a different purpose. Memories are always read from all levels, creating a rich context. You control where new memories are written.

🌍 Global Memories

Scope: Available across all your projects

Best for: Personal preferences, coding style, conventions you use everywhere, general knowledge that applies to all your work

Example Use Cases:

  • "I prefer TypeScript with strict mode enabled"
  • "Always use functional components in React"
  • "Follow the Airbnb style guide for JavaScript"
  • "I work in the healthcare domain with HIPAA requirements"

📁 Project Memories

Scope: Available to all collaborations within a specific project

Best for: Project-specific architecture, design decisions, coding patterns, team conventions, domain knowledge

Example Use Cases:

  • "This project uses Deno Fresh framework"
  • "Authentication is handled by the /auth/middleware module"
  • "We use a custom logger in utils/logger.ts"
  • "The database schema is in the /db/schema directory"

💬 Collaboration Memories

Scope: Specific to one collaboration only

Best for: Task-specific context, temporary notes, exploratory work, one-off experiments

Example Use Cases:

  • "User reported bug in the login flow on mobile devices"
  • "Exploring performance optimization for the search feature"
  • "Testing new approach to data caching"
  • "Working on prototype for client presentation"

How Memory Works

Automatic Memory Management

The LLM automatically decides what information is worth remembering and saves it to your configured location. This happens in the background without interrupting your workflow.

What Gets Remembered:

  • Important project decisions and rationale
  • Coding patterns and conventions you establish
  • Key architecture and design choices
  • Domain knowledge and business logic
  • User preferences and working style

User Control

While the LLM manages memories automatically, you maintain full control over the system:

What You Can Do:

  • Choose where memories are saved (global/project/collaboration)
  • Manually add important information to memory
  • Review and modify existing memories
  • Remove memories that are no longer relevant
  • Search memories to find specific information

📖 How Memories Are Read

When you start a collaboration, BB loads memories from all three levels, creating a comprehensive context:

  1. Global memories load first (your personal preferences and conventions)
  2. Project memories add project-specific context
  3. Collaboration memories provide collaboration-specific details
  4. The LLM uses this combined knowledge to provide better, more contextual responses

Configuring Memory Settings

Control where new memories are saved through BB's cascading settings system. You can configure defaults globally, override them per project, and adjust them for individual collaborations.

Memory Location Settings
Choose where the LLM saves new memories: Global, Project (default), or Collaboration

Global Settings

Set your default memory location in Global Settings. This applies to all projects unless overridden.

Recommended: Set to "Project" for most users. This keeps memories organized by project while remaining accessible to your whole team.

Project Settings

Override the global default for specific projects in Project Settings. Useful when a project needs different memory behavior.

Example: Use "Global" for experimental projects where you want to capture learnings that apply everywhere.

Collaboration Settings

Change memory location for individual collaborations in Collaboration Options. Perfect for temporary work or exploration.

Example: Use "Collaboration" when testing ideas you don't want to permanently save to project memory.

Best Practices

✅ Effective Memory Usage

  • Use Project as default: Keeps memories organized and accessible to all project collaborations
  • Save personal preferences globally: Coding style, conventions, and preferences you use everywhere
  • Document important decisions: Let the LLM remember why you made key architectural choices
  • Use collaboration memories for experiments: Try things without permanently affecting project knowledge
  • Review memories periodically: Remove outdated information to keep the knowledge base accurate

⚠️ What to Avoid

  • Don't save everything: Let the LLM decide what's important—trust its automatic management
  • Avoid duplicate information: If something is already in global memory, don't repeat it in project memory
  • Don't store sensitive data: Memories are stored as files—follow your security practices
  • Avoid overly specific details: Save patterns and principles, not every implementation detail

Advanced: Direct Memory Access

For advanced users: Memories are stored as Markdown files in your project's data directory. You can directly view and edit these files if needed.

Location: [config-root]/projects/[project-id]/memories/

⚠️ Caution:

Direct file editing is for advanced users only. Improperly formatted memory files may cause issues. It's generally better to use BB's memory management features.

Coming Soon

Future Memory Features

  • Memory Management UI: Search, browse, edit, and organize memories through a dedicated interface
  • Memory Export/Import: Share memory files between projects or team members
  • Team Memory Sharing: Collaborate with built-in team features for shared knowledge bases
  • Memory Analytics: Understand what the LLM is learning and how memories are being used

Next Steps

Context Management

Learn how BB automatically manages collaboration context to handle unlimited collaboration lengths without hitting token limits.

Global Settings

Configure your default memory location and other system-wide preferences.

Last updated: December 7, 2025