Listing the current list of models and the loaded model.

[FGDumitru]

Hi again,

I'm working on a library that leverages llama-swap at remote API level.

In this context:

1. Is there a way for llama-swap to list the currently configured models in a JSON format?

2. I could not find a reliable way of detecting the currently loaded model also in JSON format (for now I'm scrubbing the /logs for that which, as you may already know, is not ideally.

I think these two improvements would increasingly add value to llama-swap all around.

Thank you.

[mostlygeek]

Hi,

There is /v1/models which is openAI compatible, but it won’t list any model that is configured with unlisted:true.

There’s currently no API that lists the loaded models. Maybe adding a llama-swap propriety API to allow programmatic control would be useful.

Take a look at proxy/proxymanager.go. You may be able to add the http handlers you need quickly.

[FGDumitru]

I've added a PR that adds an additional /running endpoint:

The endpoint returns either an empty JSON object if no model has been loaded so far, or the last model loaded (model key) and it's current state (state key). Possible state values are: stopped, starting, ready and stopping.

Example output if the endpoint is called right after llama-swap has been started:
{}

Example output if the endpoint is called and a model is still being loaded:
{"model":"DeepSeek-V3-Q4_K_M","state":"starting"}

Example output if the endpoint is called and a model has been loaded:
{"model":"DeepSeek-V3-Q4_K_M","state":"ready"}

Example output if the endpoint is called and a model is being unloaded due to TTL:
{"model":"DeepSeek-V3-Q4_K_M","state":"stopping"}

Example output if the endpoint is called and a model has been unloaded due to TTL:
{"model":"DeepSeek-V3-Q4_K_M","state":"stopped"}

Note: It returns an empty JSON object if the model is marked as Unlisted.

[mostlygeek]

fixed in #65