Memory API
For the current authenticated HTTP contract, use the Versioned Memory API and OpenAPI. The SDK examples below describe the legacy client surface; they have not been migrated to the platform routes and do not establish tenant isolation.
API for agent memory operations.
Agent Memory#
Initialize Memory#
from qilbeedb import QilbeeDB
db = QilbeeDB("http://localhost:7474")
memory = db.agent_memory('my_agent')
Episode Storage#
Store Episode#
from qilbeedb.memory import Episode
# Conversation
conversation = Episode.conversation(
agent_id='my_agent',
user_input='What is 2+2?',
agent_response='The answer is 4'
)
memory.store_episode(conversation)
# Observation
observation = Episode.observation(
agent_id='my_agent',
content='User seems frustrated'
)
memory.store_episode(observation)
# Action
action = Episode.action(
agent_id='my_agent',
action='Sent email',
result='Email delivered successfully'
)
memory.store_episode(action)
Memory Retrieval#
Recall Recent#
# Most recent episodes
recent = memory.recall(recency_hours=24, limit=10)
for episode in recent:
print(f"{episode.event_time}: {episode.content}")
Recall by Relevance#
# Most relevant episodes
relevant = memory.recall(
min_relevance=0.7,
limit=20,
order_by='relevance'
)
Search Content#
# Search by content
results = memory.recall(
content_contains='order #12345',
limit=10
)
Time Range Query#
from datetime import datetime, timedelta
yesterday = datetime.now() - timedelta(days=1)
today = datetime.now()
# Episodes from time range
time_range = memory.recall(
event_time_start=yesterday,
event_time_end=today
)
Memory Types#
# Store semantic memory
fact = Episode.action(
'my_agent',
'Learned fact',
'Python 3.12 released October 2023',
memory_type='semantic'
)
memory.store_episode(fact)
# Query semantic memory
facts = memory.recall(
memory_type='semantic',
content_contains='Python'
)
Forgetting#
# Forget specific episode
memory.forget(episode_id=12345)
# Forget old, low-relevance episodes
memory.forget_old(
older_than_days=365,
max_relevance=0.2
)
Statistics#
# Get memory statistics
stats = memory.statistics()
print(f"Total episodes: {stats['total_episodes']}")
print(f"Avg relevance: {stats['avg_relevance']}")
Consolidation#
# Trigger consolidation
memory.consolidate(force=True)
# Configure consolidation
memory.configure(
min_relevance_threshold=0.1,
max_memory_size=1000000,
forgetting_enabled=True
)
Next Steps#
- Learn about Memory Engine
- Explore Agent Memory
- See AI Agents Use Case