Skip to main content
Memobase serves as a user profile backend for LLM applications, enabling them to track and update specific user attributes over time.

A sample user profile in Memobase, showing structured data slots.

By default, Memobase includes a set of built-in profile slots for common use cases, but it also offers full customization to control the specific memories your application collects.

​
Locating the config.yaml File

Memobase uses a config.yaml file for backend configuration. You can find this file at src/server/api/config.yaml in your self-hosted instance. A typical configuration looks like this: 
max_chat_blob_buffer_token_size: 1024
buffer_flush_interval: 3600
llm_api_key: sk-...
best_llm_model: gpt-4o-mini
# ... other settings

​
Understanding Profile Slots

Memobase comes with a default schema of profile slots, such as: 
- basic_info
    - name
    - gender
- education
    - school
    - major
You can extend this schema by adding custom slots under the additional_user_profiles field in config.yaml: 
additional_user_profiles:
    - topic: "Gaming"
      description: "Tracks the user's gaming preferences and achievements."
      sub_topics:
        - name: "FPS"
          description: "First-person shooter games like CSGO, Valorant, etc."
        - name: "LOL"
    - topic: "Professional"
      sub_topics:
        - name: "Industry"
        - name: "Role"
Memobase will then track these additional slots and update the user profile accordingly. If you need to define a completely custom schema, use the overwrite_user_profiles field instead. For detailed instructions on formatting profile slots, see Profile Slot Configuration.