Implementation:Truera Trulens Run Dashboard

Knowledge Sources	TruLens TruLens Docs
Domains	Observability, Visualization
Last Updated	2026-02-14 08:00 GMT

Overview

Concrete tool for launching a Streamlit-based interactive dashboard to explore evaluation results, provided by the trulens-dashboard package.

Description

The run_dashboard function launches a Streamlit dashboard as a subprocess. The dashboard reads from the database configured in the TruSession and provides interactive exploration of application traces, feedback scores, and app version comparisons. It supports multiple deployment modes:

• Local: Standard Streamlit server on localhost
• SiS (Streamlit-in-Snowflake): Compatibility mode for Snowflake-hosted Streamlit
• SPCS (Snowpark Container Services): Container-based deployment in Snowflake

Usage

Call this function after recording traces and running feedback evaluations. The dashboard provides a visual interface for exploring results that complements programmatic analysis via retrieve_feedback_results.

Code Reference

Source Location

• Repository: trulens
• File: src/dashboard/trulens/dashboard/run.py
• Lines: L41-310

Signature

def run_dashboard(
    session: Optional[TruSession] = None,
    port: Optional[int] = None,
    address: Optional[str] = None,
    force: bool = False,
    sis_compatibility_mode: bool = False,
    spcs_mode: bool = False,
    _dev: Optional[Path] = None,
    _watch_changes: bool = False,
) -> Process:
    """Run a streamlit dashboard to view logged results and apps.

    Args:
        session: TruSession to display. Defaults to current session.
        port: Port number for the Streamlit server.
        address: Address to bind to. Cannot be set in Colab.
        force: Kill existing dashboards first. Defaults to False.
        sis_compatibility_mode: Enable Streamlit-in-Snowflake compatibility.
        spcs_mode: Enable Snowpark Container Services mode.

    Returns:
        The Process executing the Streamlit dashboard.

    Raises:
        RuntimeError: If dashboard is already running (unless force=True).
    """

Import

from trulens.dashboard import run_dashboard

I/O Contract

Inputs

Name	Type	Required	Description
session	TruSession	No	Session to display (defaults to current singleton)
port	int	No	Streamlit server port (auto-selects if not given)
address	str	No	Server bind address
force	bool	No	Kill existing dashboards first (default: False)
sis_compatibility_mode	bool	No	Streamlit-in-Snowflake mode (default: False)
spcs_mode	bool	No	Snowpark Container Services mode (default: False)

Outputs

Name	Type	Description
return	multiprocessing.Process	The subprocess running the Streamlit dashboard

Usage Examples

Basic Dashboard Launch

from trulens.core.session import TruSession
from trulens.dashboard import run_dashboard

session = TruSession()
# ... record traces and run feedbacks ...

# Launch dashboard
run_dashboard(session)
# Dashboard opens at http://localhost:<port>

Force Restart on Specific Port

from trulens.dashboard import run_dashboard

run_dashboard(session, port=8501, force=True)

Snowflake-Hosted Dashboard

from trulens.dashboard import run_dashboard

# For Streamlit-in-Snowflake deployment
run_dashboard(session, sis_compatibility_mode=True)

# For SPCS deployment
run_dashboard(session, spcs_mode=True)

Related Pages

Implements Principle

• Principle:Truera_Trulens_Dashboard_Visualization

Requires Environment

• Environment:Truera_Trulens_Python_Core_Environment
• Environment:Truera_Trulens_Streamlit_Dashboard_Environment

Uses Heuristic

• Heuristic:Truera_Trulens_Feedback_Score_Thresholds