Feature Request: Add LLM-Based Classification as a Model Card in OLIM

[isaquepim]

Open Question: Prompt Engineering Strategy

One important thing we need to clarify. Should prompt engineering be handled:

• Internally (managed and versioned inside OLIM), or
• Externally (user-defined prompts provided outside the system)?

If internally: should we add a tab to upload, edit, delete the knowledge base used for the prompts? Can we incorporate the knowledge information for each symptom directly as metadata for the respective label? That would also include the RegEx queries.

Example:

Label: Febre -> {'regex' : '\b[fF]ebr(e|es|il)?\b',
'synonyms' : '...',
'few_shot_examples' : '...',
[...]}