Part 5/13:
Consolidating these notes reduces storage requirements significantly (roughly a five-to-one compression ratio) and improves searchability. These summarized memories are stored in dedicated "notes" folders, distinct from raw logs, enabling efficient lookups through semantic search.
The system considers vectorizing entire conversation chunks, allowing Raven to retrieve relevant past interactions swiftly during conversations. This creates a semantic database that acts as an internal knowledge base (KB), effectively mimicking human-like gist memory.