Skip to content

[codex] document photo processing recovery behavior#99

Merged
andreichenchik merged 2 commits into
mainfrom
codex/docs-photo-processing-recovery
Feb 14, 2026
Merged

[codex] document photo processing recovery behavior#99
andreichenchik merged 2 commits into
mainfrom
codex/docs-photo-processing-recovery

Conversation

@andreichenchik

Copy link
Copy Markdown
Owner

Issue

The documentation around item photo processing and AI classification had unresolved gaps and contradictory behavior across Item, Background Removal, AI Classification, and Add Item. The open questions made implementation and QA expectations unclear, especially for failure and retry paths.

User Effect

Without explicit rules, teams can implement different fallback behavior for failed cutout/classification cases, and users may get inconsistent outcomes when processing fails or network is unavailable.

Root Cause

The docs described only the happy path and left key operational decisions undefined:

  • whether original photo is preserved,
  • what happens when cutout fails,
  • how classification behaves on low confidence,
  • how offline classification should be deferred and retried.

Fix

This PR aligns the processing model across documentation pages and separates target behavior from current gaps.

Updated target behavior

  • docs/domain/item.md
    • Defines two photo variants per item (original + cutout).
    • Adds explicit processing lifecycle states for failure/deferred flows.
    • Documents independent retryability of background removal and classification.
  • docs/features/background-removal.md
    • Defines failure behavior (Processing Failed + retry) and one-item-per-photo expectation.
  • docs/features/ai-classification.md
    • Defines input selection (cutout preferred, original fallback).
    • Defines low-confidence policy (all-or-nothing, no uncertain auto-fill).
    • Defines deferred/failed error handling and retry behavior.
  • docs/flows/add-item.md
    • Updates pipeline to reflect immediate item creation, fallback paths, and recovery actions.
    • Replaces undefined error table with concrete expected behaviors.

Current-state gaps tracked explicitly

  • docs/constraints/current-limitations.md
    • Adds limitations for original photo preservation, background-removal failure UX, classification recovery lifecycle, and multi-item segmentation reliability.
  • docs/constraints/open-questions.md
    • Removes a now-resolved privacy question about where photos are stored.

Validation

Documentation-only change.

  • Verified clean patch application and commit.
  • Verified changed files and staged diff consistency.
  • Ran git diff --cached --check before commit (no whitespace issues).

@andreichenchik
andreichenchik merged commit 5bff9e5 into main Feb 14, 2026
3 checks passed
@andreichenchik
andreichenchik deleted the codex/docs-photo-processing-recovery branch February 14, 2026 15:10
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant