#65 — Form algorithm of converting on-boarding questionnaire to tenant profile

Repo: Twill-AI/twill-ai-ui State: closed | Status: done Assignee: meliascosta

Created: 2024-10-14 · Updated: 2025-01-09

Description

Initiated by https://twill-network.slack.com/archives/C06UCQP1BFC/p1728665647581349.

See https://docs.google.com/document/d/13nUViSVA_3aGgoCLHEslCh9LryyKrOGCco5GvZY46lI/edit - this set of questions would be presented to business owner during on-boarding. We may get a lot of valuable information about tenant here. But this information would be mostly in user-friendly “predefined options” view gathered in UI code.

Facade persistence has string “llm_info” field for each tenant and it expected to be passed to LlmEngine with each request.

Note that better to convert on-boarding questionnaire to/from “llm_info” in UI TypeScript code because otherwise need to invent protocol of passing on-boarding questionnaire UI ←> Facade which would add decent amount of effort to Facade both on initial implementation+tests and on following support (because this conversion logic expected to be changed frequently). UI is expected to be hardcoded on this logic of LlmEngine anyway (new questions, new options to answer, etc.), while Facade’s participation may be kept limited to “pass string” feature.

It looks like the best format for “llm_engine” is JSON.

AC:

• Wiki page describing how to convert user answers from on-boarding questionnaire into llm_info field inside UI in a way to:
  1. Be digestible by LlmEngine withou

Notes

Add implementation notes, blockers, and context here

Related

Add wikilinks to related people, meetings, or other tickets