#111 — Generate LlmEngine “history” from dashboard and widget.

Repo: Twill-AI/facade State: closed | Status: done Assignee: Unassigned

Created: 2024-09-19 · Updated: 2025-03-24

Description

Separated from https://github.com/Twill-AI/facade/issues/106 (also see https://twill-network.slack.com/archives/D06UREE39J5/p1726736230263629 from private communication with Martin).

Blocked by some parts of https://github.com/Twill-AI/twill-llm-engine/issues/56

Context is expected to be provided for answer_in_chat call into “run_config” dictionary as new fields

  • tenant_info: str # this context is always expected
  • widget_info: dict | None # this is expected only for widget chat
  • dashboard_info: dict | None # this is expected ony for dashboard chat
  • origin: general_chat | widget_chat | dashboard_chat # type of place answer_in_chat is executed in

tenant_info field is populated from tenant.llm_info as is.

widget_info field contains following fields (matching widget table fields):

  • name
  • type
  • sql_query
  • echarts_code (use “code” name in dict)
  • values_headers
  • values_rows
  • values_update_interval_sec
  • created_at
  • updated_at

dashboard_info field

  • name
  • description
  • created_at
  • updated_at
  • widgets : list of widget_info objects that are part of the dashboard

AC:

  • Facade may generate information (context) about dashboards and widgets compatible with LlmEngine to make it context-aware for responds to prompts in relevant serving LLM Chats.
  • (unrelated, from separate task) remove “tenant_role” from items provided to LlmEngine answer_in_chat cal

Notes

Add implementation notes, blockers, and context here

Add wikilinks to related people, meetings, or other tickets