Scroll or pinch to zoom · drag to move · double-click to zoom in
Collatorium
Starting… Reader:About this viewer
One viewer for every archive. Choose an archive at the top; the same tools apply to each. The reading engines (Claude, ChatGPT/Codex, Gemini, Transkribus) are plugged in through the Engines tab and can be swapped: any engine can be the primary reader, the others secondary. The pipeline scripts compare whichever readers are assigned.
Process lists the 13 stages per box with their status from the project's Excel log, and lets you request a run. A request is saved here with your name; Claude picks it up in a working session and writes the results back into Box and into this page.
Rules: one main page per project (this viewer); nothing is deleted; every answer is saved with the reader's name, the account and the time; archives stay separate (no name or fact is carried from one to another).
Where the earlier pages went:
Data sources:
Process —
Status per box and stage comes from the project's Excel log (). Tick the stages you want run on a box and click Request run; the request is saved here with your name. Claude runs it in a working session and reports back on this tab.
| Stage | Status | Request |
|---|
Requests
—
Engines —
Assign a role to each engine for this archive. Primary writes the base text; Secondary engines read blind and are compared against it; the Adjudicator settles disagreements before anything goes to the PI. Swap roles to test another engine as primary; the scripts take any assignment. Roles are saved here; only editors can change them.
| Engine | Provider | How it runs | Role in this archive |
|---|
Full specs per reader (exact model string, settings, prompt version, dates) live in the Readers tab of the project's Excel log; this page records only the role.
Progress —
From the project's Excel log, as of . Each bar = share of the 13 stages approved. Seeded statuses are marked "check" in the log.