Memories

Memories are the fundamental unit of storage in Memorer. Each memory represents a piece of knowledge about a user or topic.

Memory types

Type	Description	Example
Direct	Explicitly stored via `remember()`	”User prefers dark mode”
Derived	Extracted from conversations	”User mentioned they live in Seattle”
Inferred	Generated through graph reasoning	”User likely works in tech (based on tools mentioned)“

Storing memories


user = client.for_user("user-123")
 
# Simple text
result = user.remember("Alice prefers Python over JavaScript")
 
# The result contains extraction metadata
print(result.entities_created)       # Number of entities extracted
print(result.relationships_created)  # Number of relationships created
print(result.episodes_created)       # Number of episodes created
print(result.processing_time_ms)     # Processing time

You can also pass multiple items, dicts, or Document objects:


user.remember(["First fact", "Second fact"])

Recalling memories


# Semantic search
results = user.recall("What programming language does Alice like?")
 
for result in results:
    print(result.content)          # The memory text
    print(result.relevance_score)  # 0.0 to 1.0
 
# Access the combined context string
print(results.context)

Recall options


results = user.recall(
    "What does Alice do?",
    top_k=10,                      # Number of results (default: 10)
    use_graph_reasoning=True,      # Enable multi-hop graph traversal
    use_emotional_ranking=True,    # Apply emotional scoring (default: True)
)

Forgetting memories


# Delete a specific memory by ID
user.forget("memory-uuid-here")

Memory consolidation

Over time, memories can become redundant or contradictory. Memorer’s consolidation process automatically:

Merges duplicate or near-duplicate memories
Updates memories when newer information contradicts older ones
Removes stale or low-relevance memories
Strengthens frequently-accessed memories