Request for sending data via telemetry

The goal is to collect metrics about openshift lightspeed because we want to understand how users are making use of the product(configuration options they enable) as well as the experience they are having when using it(e.g. response times)

selected_model_info

Represents the llm provider+model the customer is currently using

Labels

• <provider> one of the supported provider names (e.g. openai, watsonx, azureopenai)
• <model> one of the supported llm model names (e.g. gpt-3.5-turbo, granite-13b-chat-v2)

The cardinality of the metric is around 6 currently, may grow somewhat as we add supported providers+models in the future (not all provider + model combinations are valid, so it's not cardinality of models*providers)

model_enabled

Represents all the provider/model combinations the customer has configured in ols (but are not necessarily currently using)

Labels

• <model>, one of the supported llm model names (e.g. gpt-3.5-turbo, granite-13b-chat-v2)
• <provider>, one of our 3 supported providers (openai, watsonx, azureopenai)

The cardinality of the metric is around 4 currently since not all provider/model combinations are valid. May grow somewhat as we add supported models in the future.

rest_api_calls_total

number of api calls with path + response code

Labels

• <status_code>, the http response code returned (e.g. 200, 401, 403, 500)
• <path>, one of our request paths, e.g. /v1/query, /v1/feedback

cardinality is around 12 (paths times number of likely response codes)