Session Management

How to properly manage threads, fetch history, and clear user context.

Arivu’s Memory backends store a serialized graph state of the Text-to-SQL logic against a session_id.

When engaging with the pipeline, simply passing a session_id into the invoke function will automatically fetch, utilize, and update the associated conversational thread.

Every conversation with Arivu is automatically linked to a session_id. All state is managed transparently by the memory system.

Managing Sessions

You can directly interact with the Memory Store API to fetch history, manage sessions, or build custom dashboard views.

SQLite Backend

from arivu.memory import SQLiteBackend

store = SQLiteBackend(db_path="arivu_memory.db")

# Fetch entire historical state for a session
history = store.get_session("user_session_123")
print(history)

# Clear/Delete a session (e.g., on user logout or /reset)
store.clear_session("user_session_123")

Redis Backend

from arivu.memory import RedisBackend

store = RedisBackend(redis_url="redis://localhost:6379/0")

# Fetch session history
history = store.get_session("user_session_123")

# Clear session
store.clear_session("user_session_123")

# List all sessions
all_sessions = store.list_sessions()

Context Window Management

LLMs have finite context windows. Arivu automatically handles this by intelligently pruning older, less-relevant messages when the state grows too large.

How Context Truncation Works

• Automatic pruning - Oldest messages are removed first
• Intelligent selection - Preserves semantically important interactions
• Recent emphasis - Always keeps the latest 5-10 messages
• No data loss - Full history still available in database

Best Practices

• Use short session IDs for efficiency
• Periodically archive completed sessions
• Monitor memory usage for long-running sessions
• Clear old sessions regularly to maintain performance