[Advanced] Recall Mechanisms

The memory system uses two distinct recall methods, enhancing proactive and context-aware responsiveness.

Recall mechanisms are fundamental to the dynamic interplay of the Memory-Knowledge-Reasoning (M-K-R) system. They are the processes by which Memory is actively brought into the current cognitive cycle to inform Knowledge application and shape Reasoning. Amigo employs two distinct recall methods:

Implicit Recall

Triggered by: Information gap detection during the M-K-R cycle, specifically when current Reasoning or Knowledge application identifies a need for Memory not immediately present in the active context (e.g., L2 User Model).

Process:

• Agent autonomously identifies information gaps by evaluating current conversation (active M-K-R state) against the user model (L2 Memory).

• Drills down to L0 (perfect context Memory) when required information isn't in the user model.

• Applies contextual reasoning (Reasoning influenced by current Knowledge) to synthesize relevant insights from raw conversation data (L0 Memory).

• Seamlessly integrates retrieved information (new Memory) into the M-K-R cycle, potentially recontextualizing existing Knowledge or altering Reasoning paths, while maintaining conversation flow.

The power of implicit recall comes from the user model's comprehensive coverage—approximately 90% of important information is already available with proper contextualization and precision at the start of each session. This makes information gaps relatively rare but immediately obvious when they occur during the M-K-R process.

When the agent detects an information gap (by analyzing the context graph, interaction logs, and user model against the current conversation), it automatically initiates a targeted retrieval process. This ensures the M-K-R system maintains perfect contextual understanding without disrupting conversation flow.

This recall process leverages dimensional importance hierarchies to prioritize information retrieval, ensuring that critical data (marked with higher precision requirements) is retrieved first and with complete contextual preservation. The system employs proximity-based search methods that focus on contextually relevant information—as determined by the current M-K-R state—rather than simple keyword matching.

Explicit Recall

Triggered by: Predetermined recall points defined in the context graph structure, representing anticipated needs within the M-K-R flow for specific Memory to inform critical Knowledge or Reasoning steps.

Process:

• Activates at specific decision points predefined in the context graph (nodes in the M-K-R flow).

• Generates precisely targeted queries based on functional requirements (Reasoning needing specific Memory for Knowledge application).

• Recontextualizes requested information through user model integration (Memory brought into current M-K-R context).

• Guarantees consistent information access for critical decision points in the M-K-R cycle.

Unlike implicit recall, explicit recall is deliberately engineered into the conversation flow through the context graph. These recall points ensure critical Memory is always available at key decision moments, rather than relying on gap detection by the active M-K-R process.

Explicit recall requirements can be designed in two ways:

• Hard-coded requirements that specify exactly what information must be retrieved at particular points

• Dynamic requirements that define the scope or category of information needed, with the specific retrieval targets determined at runtime

This flexibility allows for both rigid precision when certain exact information must be available (such as medication allergies before prescribing) and adaptable scope when the category of information is known but specific details may vary (such as retrieving relevant exercise history before creating a new workout plan).

The integration with the user model transforms raw recall queries into contextually-rich retrievals. Each explicit recall query is recontextualized against the user model dimensions (current Memory state), ensuring the retrieved information (newly active Memory) maintains perfect context preservation and functional relevance for ongoing Knowledge application and Reasoning.

This dual-mechanism approach ensures both responsive, dynamic information retrieval (implicit recall driven by the M-K-R cycle's needs) and guaranteed access to critical information at predetermined points (explicit recall supporting planned M-K-R junctures), creating a robust memory system optimized for enterprise reliability within the unified M-K-R framework.