0.9-1 unfamiliar user testing

Status: planned — runs after the 0.9.0 package cut (U2 + I0–I6). Sequencing: ROADMAP.md Path to 0.9.0 / 0.9-1 / 1.0.

Purpose: Strangers (or deliberately unfamiliar testers) complete a first successful notebook using only hosted/README docs — not contracts — so findings can harden Transcribe before 1.0 and leave an additive-ready foundation for After 1.0 autobiography work.

This is not an infrastructure track (not I7). It is not autobiography testing.

Inputs

  • Tagged 0.9.0 (or a 0.9.x patch that includes U2 sample + first-run docs)

  • Hosted guide from infrastructure I4/I5 (or README + docs/user_guide.md if Pages is not yet public — prefer hosted)

  • Sample notebook path from usability U2.2

Scripted happy path (15–30 minutes)

Testers use product docs only. Observers may note blockers; do not coach contract vocabulary.

  1. Install — follow the first-run install path (Docker preferred; host Python only if testing that advanced path).

  2. Open sample (or import a few of their own scans if they prefer).

  3. Confirm Ollama — Home / Diagnostics one-line health; pull or select a vision model if needed.

  4. Transcribe — run OCR on the sample (or skip if sample is pre-seeded with text for offline Analyse).

  5. Review — open Review; notice dates / empty text if present; make one edit or approve one date if suggested.

  6. Reading — open Reading; flip a few pages.

  7. Analyse Quick — run Quick (deterministic modules; LLM optional).

  8. Overview / Themes / Mood — glance at View consume pages; note empty or “needs model” states.

  9. Export — export Markdown or transcribe.notebook JSON; notice revision if shown.

  10. Backup — create a workspace backup ZIP via Settings or CLI; optionally verify.

Then 5–10 minutes free exploration (Search, Library, Places, Settings — whatever they try).

Capture

File issues (or a single testing notes doc) for:

Area

Examples

Install

Path mounts, port 8510, Docker/extra_hosts, UID/GID, missing Ollama

Models

Vision vs text confusion; unverified identity; hang/timeout honesty

Review / dates

Unapproved dates on timeline; edit vs Use as current text / merged draft in Review; empty OCR

Analyse

Empty states; “needs text model”; status strip clarity

Navigation

Library → Reading; Search jump; Home recent list

Backup / restore

Confidence; size; restore refuse messaging

Docs

Gaps between README and first success

Also note anything that would later block “life around a page” (Reading centrality, Search/Library date filters, jump-to-page) — without asking testers to import WhatsApp or photo libraries.

Explicitly out of script

  • WhatsApp / Telegram / photo-context imports

  • Person identity store, Slices, Autobiography view, reconstruction chat

  • Contract reading, doctor deep hashing as a required step

  • Live multipass fine-tune export (optional free exploration only)

Outputs and exit

Output

Owner

Issue list + severity

Tester / maintainer

Fix train on 0.9.x

Maintainers

Go/no-go note for 1.0

Maintainer

Foundation checklist sign-off

ROADMAP Path to 0.9.0 Track C

0.9-1 → 1.0 exit: critical install/OCR/review/export/backup issues closed or accepted in known_limitations.md; foundation checklist complete; PRODUCT.md still page-first.