TRACING

For investigating LLM requests and responses, you can use Langtrace, an open-source Observability tool for LLMs. It's based on OTel. It allows us to easily see what requests are being sent to the LLM and what responses are received.

Here's how to set up locally:

checkout out the Langtrace repository locally
run docker-compose up (for further instructions, check the repository)
go to http://localhost:3000 (or where you've configured Langtrace to run)
create a project, and generate an API key
run Kibana with the following command: `$ LANGTRACE_API_KEY=<my-api-key> LANGTRACE_API_HOST=<langtrace-host> npm start

If you use the AI Assistant, you should then start seeing traces and metrics in the project you've configured.