Pocket Agent

Memory

Configure how the agent remembers and searches information.

Settings

EmbeddingsProvider for vector embeddings:
  • OpenAI — Requires OpenAI API key
  • None — Keyword search only
Vector WeightHow much to weight semantic similarity (0-1).
Default: 0.7
Keyword WeightHow much to weight keyword matching (0-1).
Default: 0.3
Min ScoreMinimum relevance threshold for results (0-1).
Default: 0.35
Max ResultsMaximum memories to retrieve per search (1-20).
Default: 6

Tips

  • Higher Vector Weight — Better at finding conceptually similar content
  • Higher Keyword Weight — Better at exact matches and specific terms
  • Lower Min Score — More results, but may be less relevant
  • Higher Max Results — More context, but slower responses