Memory Recall
This article focuses on functional explanation. For detailed interface fields and limits, please click the link above.
1. Capability Overview
Memory recall is responsible for quickly retrieving the most relevant memory fragments when the user initiates a new request.
- Role: Ensures that the model does not start “from scratch” when generating responses, but instead integrates the user's history, preferences, and context.
- Returned Results: The recalled content is presented as plaintext facts.
- Traceable: Each memory is accompanied by its source, timestamp, and confidence level.
- Highly controllable: Developers have full control over which memories enter downstream logic.
- Features:
- Seamless recall: Users don’t need to repeat their previous choices or preferences.
- Structured output: Separates factual and preference memories, making it easier for developers to control whether to inject them.
2. Advanced: Deep Customization
In MemOS, recall and completion are not achieved through a single path, but through combinations of multiple strategies and components. Different scenarios may require different configurations. This section lists the main steps and customizable points for you to flexibly choose according to business needs.
| Layer | Customizable Points | Example |
|---|---|---|
| Memory Recall | Adjust recall strategy | Raise similarity threshold to only return memories with confidence ≥0.9 |
| Set filters | Only retrieve the last 30 days of conversations; only preference memories, not factual ones | |
| Output Governance & Audit | Compliance fallback | Automatically prepend “Answer must comply with regulations” before completion |
| Logging & traceability | Record used memories and few-shot selection each call | |
| A/B testing | Run two concatenation templates simultaneously, compare user satisfaction differences |
3. Next Steps
Learn more about MemOS core capabilities:
Memory Schedule
Memory scheduling acts as the neural center of the memory system. It dynamically allocates and reclaims cognitive resources in the background by organizing and coordinating operations such as adding, updating, transforming, and retrieving memories, thereby continuously optimizing the entire memory system.
Memory Lifecycle
In MemOS, memories are not stored statically; they continuously evolve over time and through usage.