Decision Tree Overview

How each feature's name, status, and classification were determined. ← Back to Features

Features Changelog Decision Tree Overview
Legend — How to read the decision tree

Status Trail Nodes

Docs Banner — Status extracted by regex from page's Important/Note admonition
LLM Verdict — Claude's analysis when regex was ambiguous or page has sub-features
Preview API — Status from Databricks workspace/account preview API
Cross-Reference — Feature also mentioned on another page (confirms status independently)
Final Status — Reconciled status used in output (green border = winning)
Carried Forward — Preserved from previous run (feature not detected in current pipeline)

Name Sources

page_title — Page title (single-page feature)
nav_tree — Navigation tree (deepest path component)
api_display_name — From Databricks Preview API
llm_sub_feature — LLM-identified sub-feature on a parent page
name_override — Manual override in config/name_overrides.json
ai_function_norm — Normalized AI function name
generic_prefix — Generic name prefixed with parent category

Confidence Levels

confirmed — API and docs agree
api_only — Only API has status
api_override — API overrides conflicting docs
llm_sub_feature — LLM identified on parent page
llm_sub_feature_xref — LLM sub-feature confirmed by cross-reference
high / low — Regex extraction confidence

Cross-References

When a feature appears on both a dedicated page and a parent overview page, the duplicates are merged.
The dedicated (deeper) page is kept as canonical. The overview mention becomes a cross_reference.
Cross-references add confidence — the feature was independently identified on multiple pages.