​
Auto-Validation

By default, Memobase validates all new profile information before saving it. This process serves two main purposes:
  1. Ensures Meaningful Data: It filters out low-quality or irrelevant information that the LLM might generate, such as “User has a job” or “User did not state their name.”
  2. Maintains Schema Adherence: It verifies that the extracted information aligns with the descriptions you have defined for your profile slots.
However, if you find that too much information is being filtered out, you can disable this feature by setting profile_validate_mode to false in your config.yaml:
config.yaml
profile_validate_mode: false
Disabling validation will result in more data being saved, but it may also lead to less accurate or less relevant profile information.

​
Strict Mode

By default, Memobase operates in a flexible mode, allowing the AI to extend your defined profile schema with new, relevant sub-topics it discovers during conversations. For example, if your configuration is:
config.yaml
overwrite_user_profiles:
    - topic: "work"
      sub_topics:
        - "company"
Memobase might generate a more detailed profile like this:
Possible Profile Output
{
    "work": {
        "company": "Google",
        "position": "Software Engineer",
        "department": "Engineering"
    }
}
This is often useful, as it’s difficult to anticipate all the valuable information your users might provide. However, if you require that only the profile slots you have explicitly defined are saved, you can enable strict mode:
config.yaml
profile_strict_mode: true
In strict mode, Memobase will adhere rigidly to the schema in your config.yaml.