Part of a /al/editor ↔ Weaver-wizard parity review. This is one of seven gaps that keep the graphical editor from being a full replacement for assembly_line.yml.
What happens
The wizard detects a PDF whose fields are unlabelled (or a DOCX with no Jinja tags), offers to run ALDashboard's labelling tools in the background, and shows every suggestion for review before applying it:
data/questions/assembly_line.yml lines 403-470 — detection and the offerdata/questions/assembly_line.yml lines 588-652 — the background task and the DOCX review screen
/al/editor has no counterpart. Searching api_editor.py and editor.js for autolabel, auto_label, or label_fields returns nothing.
Why it matters
Unlabelled scanned court PDFs are common. Without this, the editor's answer for one of them is "go back to the wizard" — which defeats the point of treating /al/editor as the replacement pathway.
Suggested approach
Reuse the background-job machinery the editor already has. weaver_editor_new_project_task already models queued work with an owner-scoped Redis progress record, status polling, and structured errors — auto-labelling fits that shape exactly.
Surface the results in a reviewable table rather than applying them directly, matching the wizard's behaviour. The review table is shared with the bulk field-mapping view, so building that first is worthwhile.
Related
- Best built after the bulk field-mapping view, which supplies the review UI
Part of a
/al/editor↔ Weaver-wizard parity review. This is one of seven gaps that keep the graphical editor from being a full replacement forassembly_line.yml.What happens
The wizard detects a PDF whose fields are unlabelled (or a DOCX with no Jinja tags), offers to run ALDashboard's labelling tools in the background, and shows every suggestion for review before applying it:
data/questions/assembly_line.ymllines 403-470 — detection and the offerdata/questions/assembly_line.ymllines 588-652 — the background task and the DOCX review screen/al/editorhas no counterpart. Searchingapi_editor.pyandeditor.jsforautolabel,auto_label, orlabel_fieldsreturns nothing.Why it matters
Unlabelled scanned court PDFs are common. Without this, the editor's answer for one of them is "go back to the wizard" — which defeats the point of treating
/al/editoras the replacement pathway.Suggested approach
Reuse the background-job machinery the editor already has.
weaver_editor_new_project_taskalready models queued work with an owner-scoped Redis progress record, status polling, and structured errors — auto-labelling fits that shape exactly.Surface the results in a reviewable table rather than applying them directly, matching the wizard's behaviour. The review table is shared with the bulk field-mapping view, so building that first is worthwhile.
Related