Intake & Provenance Pipeline

No material skips intake.

Nothing lands in the catalog without passing every step of a defined procedure. Account exports from ChatGPT and Gemini are treated as unverified and unscoped until they clear collection, privacy screening, dedupe, and verification.

Required First Step

To begin a work, provide the account export.

The Library of Alexandria cannot catalog, describe, or verify a ChatGPT- or Gemini-derived work until the official account export is obtained and placed in its drop zone. No export means no verified source, and no verified source means no catalog entry.

Export from ChatGPT Export from Gemini (Google Takeout)

This site has no upload form and collects no data. Exports are delivered privately to the Intake Coordinator through the owner\'s own channel, never through this public page. The links above open each provider\'s official export tool in a new tab.

Coming Works — Current State

Two placeholders, awaiting source

ChatGPT export

Pending

Not yet obtained. The chatgpt-export/ drop zone is empty. A placeholder entry exists in the catalog; no editorial claims about content, quality, or scope are permitted until intake completes.

Source path when available: Settings → Data Controls → Export data.

Gemini export

Pending

Not yet obtained. The gemini-export/ drop zone is empty. A placeholder entry exists in the catalog; no editorial claims are permitted until intake completes.

Source path when available: Google Takeout → Gemini Apps Activity scope.

No steps beyond placeholder creation have been performed for either export. Advancing a placeholder to development-corpus or published-preview is a separate, deliberate editorial decision — never done in the same step that admits new intake.

The Six-Step Procedure

From raw archive to cataloged source

Applies to both ChatGPT and Gemini exports. A failure at any step routes the item to quarantine — held, not deleted.

STEP 01

Obtain

Use the official export mechanism — never screenshots or retyping. Place the untouched archive in its drop zone; extract into a working subfolder.

STEP 02

Record metadata

Log provider, request/download dates, filename, size, and SHA-256 hash before opening the content.

STEP 03

Scope & privacy screen

Scan for personal/sensitive data; keep only threads relevant to a Library of Alexandria work. Anything sensitive stays quarantined and never enters the catalog.

STEP 04

Dedupe

Compare against existing sources. Verbatim duplicates become corroborating copies, not new sources; contradictions get flagged for a Publisher decision.

STEP 05

Verify

Extract exact, quotable passages preserving wording. Save a scoped excerpt to _processed/ — never point sources at the raw archive.

STEP 06

Assign status

New material enters as pending-source-ingestion first. Corroborating material extends an existing work's sources as a dated, additive change.

Drop zones

chatgpt-export/
Raw, as-downloaded ChatGPT archives
gemini-export/
Raw Google Takeout / Gemini archives
_processed/
Items that completed every step and were admitted
_quarantine/
Items that failed a step — held, not deleted

Dedupe rules

  • Hash-first. Identical hash means the same file, not a new source.
  • Near-duplicates are successor versions — keep both, cite the most complete.
  • Timestamp precedence is a provisional judgment when verified sources conflict — logged, never an automatic overwrite.
Rights gate at intake

Before any export-derived content is used in a published work, a counsel-review gate applies regarding each provider's terms of service and output-ownership terms. This pipeline makes that gate visible; it does not resolve it. No legal conclusions are issued here.