Overview
The end_session function completes the current active session, optionally setting final evaluation metrics and success status. This sends all remaining data to Lucidic and cleans up the local session state.
Syntax
lai.end_session(
session_eval: Optional[float] = None,
session_eval_reason: Optional[str] = None,
is_successful: Optional[bool] = None,
is_successful_reason: Optional[str] = None
) -> None
Parameters
Overall evaluation score for the session (typically 0.0 to 1.0). This provides a quantitative measure of session quality or success.
Text explanation for the session evaluation score. This provides context for the numerical score.
Boolean flag indicating whether the session achieved its intended goal.
Text explanation for why the session was or wasn’t successful.
Returns
This function returns None.
Examples
Basic Usage
import lucidicai as lai
# Start session
lai.init(session_name="Customer Query")
# Do work...
handle_customer_query()
# End session
lai.end_session()
With Success Evaluation
import lucidicai as lai
lai.init(session_name="Data Processing Pipeline")
try:
# Process data through multiple steps
data = load_data()
processed = transform_data(data)
results = analyze_data(processed)
# Successful completion
lai.end_session(
session_eval=0.95,
session_eval_reason="Pipeline completed with high quality results",
is_successful=True,
is_successful_reason="All data processed without errors"
)
except Exception as e:
# Failed completion
lai.end_session(
session_eval=0.2,
session_eval_reason=f"Pipeline failed: {str(e)}",
is_successful=False,
is_successful_reason="Critical error during processing"
)
Conditional Success
import lucidicai as lai
lai.init(session_name="Goal-Based Task")
# Define success criteria
target_accuracy = 0.90
target_speed = 100 # items per second
# Execute task
results = execute_task()
# Evaluate against criteria
success = (results['accuracy'] >= target_accuracy and
results['speed'] >= target_speed)
lai.end_session(
session_eval=results['accuracy'],
session_eval_reason=f"Accuracy: {results['accuracy']:.2%}, Speed: {results['speed']} items/sec",
is_successful=success,
is_successful_reason=(
"Met all performance targets" if success
else f"Below targets - need {target_accuracy:.0%} accuracy and {target_speed} items/sec"
)
)
Auto-End Behavior
import lucidicai as lai
# With auto_end enabled (default)
lai.init(session_name="Auto-End Example", auto_end=True)
# Even if you forget to call end_session or the program crashes,
# the session will automatically end when the process exits
# With auto_end disabled
lai.init(session_name="Manual-End Example", auto_end=False)
# You must explicitly call end_session
# If you don't, the session data may be lost
lai.end_session()
Integration with Monitoring
import lucidicai as lai
from your_metrics import track_metric
lai.init(session_name="Production Service", production_monitoring=True)
try:
result = process_production_request()
# Track success metrics
track_metric("request_success", 1)
lai.end_session(
session_eval=1.0,
is_successful=True,
is_successful_reason="Request processed successfully"
)
except Exception as e:
# Track failure metrics
track_metric("request_failure", 1)
lai.end_session(
session_eval=0.0,
is_successful=False,
is_successful_reason=f"Request failed: {type(e).__name__}"
)
raise
Notes
- Only affects the current active session
- Automatically ends any active step before ending the session
- Clears all session state after ending
- If
auto_end=True (default), session ends automatically on process exit
- Cannot be called if no session is active
- All buffered data is sent to Lucidic before the session ends
Error Handling
from lucidicai.errors import LucidicNotInitializedError
try:
lai.end_session()
except Exception:
# Session might not be active
print("No active session to end")
Best Practices
- Always End Sessions: Ensure sessions are properly ended to avoid data loss
- Use Auto-End: Keep
auto_end=True for automatic cleanup
- Meaningful Evaluation: Provide both score and reason for clarity
- Success Criteria: Define clear success criteria for
is_successful
- Error Handling: End session even when errors occur
See Also
