Workers glossary

Nassila's Ouroboros architecture uses seven worker codenames — each a loop stage and code module. They are product vocabulary, not seven equal app tabs. The user-facing journey is the Manuscript loop.

Worker table

Worker	Arabic	Loop stage	End-state role	Current status
Maktab	مكتب	Ingest	Manuscript upload and segmentation	Stub — required for full Tier 3 loop
Masdar	مصدر	Sources	Cited-paper text (OA PDF + planned user attach)	OA PDF grounding shipped (1.2.0); user PDF attach still planned
Sanad	سند	Ground	Passage vs source excerpt → verdicts	Live — L3 when Passage grounding enabled
Shahid	شاهد	Evidence	Tables and figures as evidence	Disabled — Tier 3+
Raqim	رقيم	Records	Import, verify L1/L2, export (citeproc)	Live — Bibliography mode + loop feed
Tasnif	تصنيف	Risk	Dedupe, predatory flags, issue triage	Live — inline in loop + Bibliography
Sharh	شرح	Explain	Issue and mismatch explanations	Partial — deterministic copy today

Ouroboros vs Hydra

Ouroboros (ship)	Hydra (reject)
One manuscript audit journey	Seven peer tabs as the main product
Workers complete each other as stages	User copies text between modules
Upload → sources → audit → export	Pick a worker and hope the rest catches up
Unfinished stages shown as pipeline gaps	Stubs that look like finished apps

Where workers appear in the UI

Raqim — Bibliography mode (primary reference workspace)
Sanad — Manuscript Sanad bar and Passage grounding settings
Tasnif / Sharh — Inline in loop audit detail and bibliography drawer
Maktab / Masdar / Shahid — Honest stubs or pipeline labels until Tier 3 ships

Task IDs (reference)

Engineering task constants (for cross-reference with release notes):

Worker	Task id
Sanad	`l3_grounding`
Maktab	`doc_extract`
Masdar	`source_pdf_extract`
Shahid	`table_figure_grounding`
Raqim	`webpage_metadata`
Tasnif	`webpage_classify`
Sharh	`issue_explain`

Manuscript loop
Roadmap
Local models — Sanad tiers only