Brain & Memory

The Brain is the identity and memory core of Gumm. It defines who your assistant is, what it remembers about you, and what it knows about the world.

Brain identity

The Brain’s identity determines how Gumm behaves in every conversation. It is injected as a system prompt before each message.

Edit identity

Go to Brain → Identity section → click Edit.

Changes take effect immediately on the next conversation.

You can also update individual fields via the CLI:

gumm brain config-set identity.name "Atlas"
gumm brain config-set identity.personality "Concise, direct, developer-focused assistant"

Memory

Gumm has a persistent memory system that remembers facts about you across conversations.

How auto-memory works

After each conversation, Gumm runs a lightweight analysis in the background to extract personal facts — things like:

• Your name and preferences
• Projects you are working on
• Tools and technologies you use
• Recurring habits or goals

These facts are stored with a strength score. A fact mentioned multiple times becomes stronger and is more likely to be included in future context windows.

Viewing memory

Go to Brain → Memory tab to see all stored facts.

You can filter by:

• Namespace — brain for global memory, or a module ID for module-specific memory
• Type — fact, preference, context, event

Adding memory manually

You can tell Gumm to remember something directly in chat:

“Remember that I prefer dark mode and use Neovim.”

Or save facts explicitly via the CLI:

gumm brain memory list

Deleting a memory

Click the trash icon next to any memory entry in the Brain dashboard to permanently delete it.

Disabling auto-memory

If you prefer that Gumm does not automatically extract facts, you can disable auto-memory:

gumm brain config-set brain.autoMemory false

Or from the Brain dashboard, toggle the Auto Memory switch in Settings.

Knowledge base

Beyond personal facts, Gumm maintains an evolving knowledge base — a collection of structured notes the brain can write, update, and delete during conversations.

What gets stored in the knowledge base

The brain itself decides when to save something. Examples:

• Step-by-step procedures you’ve taught it
• Context about your ongoing projects
• Corrections to things it got wrong
• Notes from research it performed

Knowledge files are stored in .data/brain-knowledge/ (inside the Docker volume) and are automatically included in the system prompt.

View the knowledge base

Go to Brain → Knowledge tab.

Constraints

Knowledge files never contain secrets or executable code — the brain is instructed to reject any such content.

Brain configuration keys

Advanced users can set any brain configuration key directly:

gumm brain config-set <key> <value>

Brain statistics

The Brain dashboard shows a live summary:

Via the CLI:

gumm brain stats