4. Visualize the Log

Login to kibana

Access https://localhost:5601 in your web browser. To log in to Kibana, use elastic as the username and enter the password you have selected, which is the one you defined in the PASSWORD environment variable.

Kibana Login to Review LLM Analytics

NOTE: Please note that the gecholog-ek-dev bundle, being a development setup running on your localhost, utilizes self-signed certificates. When accessing the https://localhost:5601 website, which may be flagged as insecure, proceed by accepting the security prompt.

NOTE: For information about basic licensing for Elastic & Kibana (free for dev use at point of writing this) visit Elastic License 2.0 (ELv2) and Kibana Basic License.

View the dashboard

You can find the pre-built gecholog Dashboard DevOps in (at least) two ways

  • Go to https://localhost:5601/app/dashboards and click on gecholog Dashboard DevOps
  • Click on the "hamburger-menu" in the upper left corner. Select Analytics -> Dashboards and then click on gecholog Dashboard DevOps

And here is the pre-build dashboard:

LLM Analytics Dashboard from LLM Gateway gecholog.ai

NOTE: The dashboard will be empty until you have completed the Section Test your configuration and actually sent some traffic through gecholog.

What we measure

The gecholog Dashboard DevOps pre-configured in Kibana serves as a demonstration of the simplicity in crafting visualizations tailored to your needs using gecholog log data.

Included features are:

  • Word cloud generated from NLP post-processor analyzing request and response payloads.
  • Detailed breakdown of response times, encompassing actual requests and comparisons of pre- and post-processing durations.
  • Trends in token usage and API call patterns.
  • Monitoring of traffic and token usage per router.
  • Comprehensive tracking of errors and response codes.
  • Analysis of processor executions.

Kibana offers a user-friendly interface for experimenting and creating your own graphs and dashboards.

NOTE: Kibana is included primarily for convenience. gecholog is not reliant on Elastic/Kibana, and its logs can be integrated into any log analytics tool.

Subscribe to the log

If you are curious to explore how the log looks like, configuring gecholog to direct logs to a mounted volume is straightforward. This can be done using the nats2log processor in file_writer mode, a standard library feature pre-installed in gecholog container. For more details, refer to the section nats2log.