Skip to main content

Snowflake Quickstart Guide

Overview​

This guide walks you through setting up and using the core functionalities of the Tempo application, a comprehensive system for model inference, optimization, and automated detection.

Prerequisites​

  • Ensure you have the same permissions for warehouse, compute pool, and task management to use Tempo.

Initial Setup​

  1. Create Resources

    CALL management.create_resources();

This initializes the application by loading model weights and configurations.

Core Functionalities​

1. Static Detection​

Running Inference​

Run inference by specifying only your service name:

CALL static_detection.inference('your_service_name');

Deep Dive Analysis​

To investigate specific sequences flagged as anomalies, use the sequence ID from the output table. This command helps you identify the flagged communication between IPs and enables further investigation of associated data points:

CALL inspect.deepdive(sequence_id);

2. Automated Detection​

Setting Up Automated Inference​

To start automated inference, provide the fully qualified name (FQN) of the database.schema.table_name and the slot number where the table is referenced. The time period for auto inference is set internally to 8:00am UTC.

  1. Enable Change Tracking and Set Permissions Before running automated inference, ensure that change tracking and permissions are configured for the table you want to use for time-based automated detections. Replace database.schema.table_name with your own database, schema, and table name that was set to be referenced the specific slot you chose.

    ALTER TABLE database.schema.table_name SET CHANGE_TRACKING = TRUE;
    GRANT USAGE ON DATABASE database TO APPLICATION tempo;
    GRANT USAGE ON SCHEMA database.schema TO APPLICATION tempo;
    GRANT SELECT ON TABLE database.schema.table_name TO APPLICATION tempo;

  2. Call the prcoedure

    CALL automated_detection.start_automated_inference(
        'database.schema.table_name',
        slot_number
    );

Managing Automated Detection​

Control automated inference streams by adjusting their state. Streams are tagged to references by slot number (e.g., stream_one_interactions for slot 1,stream_two_interactions for slot 2) and stream_three_interactions for slot 3).

  • Suspend: Temporarily halts inference but continues updating the stream.
  • Resume: Restarts the inference task.
  • Stop: Ends the inference task and clears data from the source table.
CALL automated_detection.alter_automated_inference('stream_name', 'suspend');
CALL automated_detection.alter_automated_inference('stream_name', 'resume');
CALL automated_detection.alter_automated_inference('stream_name', 'stop');

3. Model Optimization​

Fine-tune the model using a chosen service name. This process updates the model based on data referenced in the tuning log on the app's reference page.

CALL model_optimization.tune_model('service_name');

4. Reference Management​

Manage system references and access the web-based management interface. Placeholder for the image: Reference Management Console Placeholder

The management console can be accessed via the Streamlit dashboard.