Advanced Configuration

If you are feeling fancy you can customize how you send information to Empromptu. Use this config guide to add customizations to your prompt registry function.

If you have not replaced your static prompts with Empromptu's prompt registry function, visit the quickstart guide here.

In the prompt registry function

#Remember this is the base function that you use to feed prompts from Empromptu in to your code 
prompt_context = prompt_registry.get_prompt(prompt_family_name,input_data)
prompt_registry.set_autolog()
ModelUtils.score_and_log(prompt_family_name,prompt_context)

You can optionally choose to make the following modifications

Config	Description	Code
Model	You can change the model	prompt_context = prompt_registry.get_prompt(prompt_family_name,input_data)
Temperature	You can change the temperature
Prompt Language

In the eval statement

#Remember if you have defined your eval in your code, it looks like this

You can optionally choose to make the following modifications

Config	Description	Code
Random	You can set your initial task to random	prompt_context = prompt_registry.get_prompt(prompt_family_name,input_data)
Request	You get the best prompt from among the choices that exist
Custom Request	You can tell the request function that you want to optimize using something other than the main eval score

Embeddings

You can optionally choose to make the following modifications

Config	Description	Code
Embedding	You can set a custom embedder	prompt_context = prompt_registry.get_prompt(prompt_family_name,input_data)

General

You can optionally choose to make the following modifications

Config	Description	Code
Latency	You can change the latency	prompt_context = prompt_registry.get_prompt(prompt_family_name,input_data)
Run Token Stats	You can run token stats separately
Turn off Debugger	You can turn off the debugger
Turn off Auto Logger	You can turn off the Auto Logger