CD OPS
Not signed in

Workbench

Evidence review and publication workspace

Worklist

Profile blockers, evidence state, review queue, and source capture.

Loading profiles…
Pick a profile to scope sources, claims, readiness, and snapshots.
Choose profile
Workflow state appears after profiles load.
The worklist is the team’s control plane: one row per profile run, with the blocking work stated in operational terms.

Profile Worklist

Loading profile review state...
Profile Identifier Current Stage Evidence Gap Source readiness Action
Sources become usable evidence only after capture, extracted text, and source-packet provenance are visible to reviewers.

Sources & Packets

Batch capture saves uncaptured profile source URLs and extracts searchable text when available.
Filter to a profile to see source health, text coverage, and capture failures.
Pick a profile to see the source-intake inbox.

Source Packet Registry

Reviewer-readable citation packets with hashes, locators, artifact URIs, and backup state.
Packet Locator Hash Backup
No packets loaded.
Select a packet to inspect source provenance, parser fields, and reviewer-readable text/table.
Pick a source from Sources & Packets to open it here.
The review surface is claim-first and packet-gated: approve against exact source text, or use a documented judgment approval when packet-backed evidence supports the call.

Review Queue

Filter by profile to turn this into a real review queue.

Evidence Matrix

Select a claim to review evidence and proposed justifications.
Review the actual profile language paragraph by paragraph. Atomic claims and packet evidence are shown underneath as support, but this surface is about the words a reader will see.

Publication Review

Load a profile to inspect draft prose and evidence coverage.

Language Coverage

The reviewable prose appears here once a profile is loaded.
Publication is a snapshot workflow. A profile can export only when approved claims have packet-backed evidence, explicit review rationale where needed, and a complete source manifest.

Publication Gate

Load a profile slug to see publication blockers, evidence state, and approved claim export.

Snapshot & Manifest

No profile loaded. The approved-claim export and snapshot history appear here once a profile is loaded on the left.
Import turns researcher Markdown into source records and draft atomic claims. The next required step is source intake: capture URLs or manually import files before evidence review.

Import Researcher Markdown

Optional. Selecting a result fills the ID and label above. Broad searches may show OpenAleph duplicates; skip this if the subject is ambiguous.
or paste below — dropping a file on the text area works too
AI extraction (always on)
Ready.

Import Result

Imported sources are handled first. Claims enter the queue as “needs evidence” and should not move into review until their sources have readable packets.
Technical log optional — for troubleshooting
A step-by-step record of what the browser did during import. You can ignore this — it's only useful if an import fails or looks wrong, to see (or send us) exactly what happened.
Waiting for import
Run an import to see the request, API, and extraction steps here.
Operations shows the external systems that affect review quality: optional AI paths, source capture providers, packet gates, and backup state.

Feature State

No feature state loaded.

Backup Policy

No backup status loaded.

LLM Usage & Model Mix

Internal token ledger for model-quality decisions. Prompts and source text are not stored.
No LLM usage loaded.

Audit Trail

No activity loaded.
Account management — admin only. Create accounts, assign roles, reset passwords, and deactivate users. New accounts must change their password on first sign-in.

Accounts

No users loaded.

Create account