Skip to content

feat(vision): unified workflow — alt+title+description+tags+classify in one call#84

Merged
gfargo merged 1 commit into
mainfrom
feat/vision-unified
May 11, 2026
Merged

feat(vision): unified workflow — alt+title+description+tags+classify in one call#84
gfargo merged 1 commit into
mainfrom
feat/vision-unified

Conversation

@gfargo
Copy link
Copy Markdown
Owner

@gfargo gfargo commented May 11, 2026

Closes #83. See issue.

`localpress vision <ids>` is the unified AI-metadata workflow. Generates
alt, title, description, tags, and classification for one or more
attachments in a single pass and either prints proposals for review or,
with --apply, writes them all back to WordPress.

This is the "I just uploaded a new image, fill in everything" command,
both for humans inspecting one attachment and for MCP agents that want
all the metadata in a single tool call.

- Print-only default — bulk metadata writes require deliberate --apply
- --fields <list> to subset (alt, title, description, tags, classify)
- Idempotent unless --overwrite (won't clobber existing field values)
- Time-machine snapshot before any write
- Classification result is cached the same way as `classify`, so
  subsequent `optimize` calls pick smarter format defaults
- Tag results write as `[tags: …]` caption block, matching `tag` command
- MCP tool `vision` registered with same surface

This is intentionally a per-attachment workflow — N Ollama calls per
image (one per field) makes it expensive at library scale. Use the
dedicated commands with their `--missing-*` filters for bulk passes.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@gfargo gfargo merged commit 1e643f7 into main May 11, 2026
4 checks passed
@gfargo gfargo deleted the feat/vision-unified branch May 11, 2026 21:00
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.

feat(vision): unified workflow — alt+title+description+tags+classify in one call

1 participant