The Interface
One screen. Review, correct, and submit.
Everything you need is in a single view: the image, its current tags, your edit controls, and a notes field that teaches the system.
dam.handiramp.com/review
↩ Your progress has been restored — pick up where you left off
✓ Approved
Review Tags AI proposed · Batch 92
Product / Hierarchy Tags HR: / PS: / HT: PREFIX
ps:folding-dog-ramp-graphite × ps:side-entry-strap ×
Spec Tags SPEC: PREFIX
Content Tags AI-INFERRED
sc:vehicle × sc:consumer × SUV access ramp × Toyota 4Runner × bi-fold ramp × parking lot × overcast ×
✏ Title
★ Rating
★★★★
🤖 AI Caption
A white Toyota 4Runner SUV is parked in a commercial parking lot with both front and rear passenger doors open. A gray bi-fold PetStep folding dog ramp is deployed from the rear passenger door...
📝 Notes for the Agent
Explain your changes. These become rules for future batches — be specific.
Last saved: Wed, Jun 24, 6:03 PM
A
B
C
D
E
F
G
H
I
J
A

Resume Banner

Appears when you return to a session with unsaved draft work. Your tags, notes, and ratings from last time have been restored from local storage. Click ✕ Dismiss once you've oriented yourself — your drafts stay live until you submit or clear them.

B

Status Badge

Shows the current review state of the image you're viewing: ✓ Approved (already submitted), AI Proposed (untouched — tags from AI only), or nothing (in progress). Matches the dot on the thumbnail strip for the active image.

C

Image + Navigation

The photo being reviewed at full natural size. The filename, file size, and date are shown below the image. Use ← Prev / Next → buttons (or keyboard arrows) to move through the batch. The blue-outlined thumbnail in the strip tracks your position.

D

Product / Hierarchy Tags

Blue chips. Identify the exact product with a brand-appropriate prefix: hr: HandiRamp · ht: HandiTreads · ps: PetStep · gb: GoBuddy. Click × to remove. Start typing in the input field and autocomplete suggestions appear from the live worksheet.

E

Spec Tags

Amber chips with spec: prefix. Capture capacity and dimensions — e.g. spec:52in, spec:50klb. These are assigned by reviewers only; the AI doesn't guess them. Add one when the product's spec is clearly identifiable in the image.

F

Content Tags

Green chips. Describe what's visually in the scene — subject, setting, activity, color, condition. Free-form, no required prefix. AI-generated but fully editable. Grow naturally as new scenarios appear across batches.

G

Title · Rating · AI Caption

Title is the editable image name written to XMP:Title on submit. Rating (1–5 stars) is a quality signal — 5 = hero shot, use anywhere; 1 = technically fine but not preferred. AI Caption is the machine's prose description, read-only, for context only.

H

Notes for the Agent

The most important field. If you changed anything, explain why here. One or two clear sentences. These notes get distilled into reusable rules that apply to every future batch automatically. Vague notes produce vague improvement — be specific.

I

Submit Review · Mark Duplicate · Close Batch

Submit Review writes XMP to the file immediately for the current image only — no auto-advance, so you can verify before moving on. Mark Duplicate flags a near-identical file for removal. Close Batch is permanent — use it only when both reviewers are fully done.

J

Thumbnail Strip

The full batch in one horizontal row at the bottom. Each thumbnail has a colored status dot: 🟢 green = you submitted it, 🔵 blue = another reviewer submitted it, 🟠 orange = you have unsaved draft changes, no dot = untouched. Click any thumb to jump to that image.


Step by Step
How to Review a Batch
A typical session takes 15–30 minutes depending on batch size and how many corrections are needed.
1

Open the portal

Navigate to dam.handiramp.com/review in your browser. The current batch loads automatically. If you have unfinished work from a previous session, the blue resume banner appears at the top — your draft state has been restored.

ℹ️ If the banner says "↩ Your progress has been restored" — that's your previous session's unsaved edits coming back. Click to dismiss it once you've had a look.
2

Scan the thumbnail strip before starting

Check the dots on the bottom thumbnail strip. Blue dots mean another reviewer already submitted that image — you can skip those unless you disagree. Green dots are images you already handled. No dot = needs attention.

💡 If you're working alongside another reviewer, coordinate who starts from which end. One of you works left-to-right, the other right-to-left. The dots make overlap obvious if you meet in the middle.
3

Look at the image, then check the tags

Click a thumbnail to open it. Ask yourself: does the product tag match what's in the frame? Are the content tags accurate? Is anything missing or wrong? Your gut read is the right one — if something feels off, it probably is.

4

If the tags look right — submit as-is

Click Submit Review. Tags are written to the file. A green dot appears on the thumbnail. Done — move to the next image.

💡 Even straight approvals help the training set. They tell the system exactly what correct looks like for that product and scene type.
5

If something is wrong — edit the tags

  • Click × on any chip to remove it — it stays visible as struck-through until you submit, so you can see what changed
  • Type in the add field and click + to add a new tag — autocomplete suggestions appear after 1–2 characters, from the live worksheet and previously used tags
  • Tags are entered exactly as typed — spaces are allowed (e.g. loading dock is valid)
  • Use the brand prefix matching the product: hr: HandiRamp · ht: HandiTreads · ps: PetStep · gb: GoBuddy
  • For spec tags: manually add spec:52in, spec:50klb, etc. when the spec is clearly identifiable
⚡ While you're editing, the thumbnail shows an orange dot. Nothing is written to the server yet — you can change your mind freely before hitting Submit.
6

Edit the title and rating if needed (optional)

The Title field is editable — fix it if the AI generated something wrong or generic. Rating stars (1–5) are optional but useful: 5 = hero shot, use anywhere; 3 = solid but not exceptional; 1 = technically acceptable, avoid for featured use.

7

Add a note explaining your changes

This is the step that compounds over time. If you changed anything, explain why in the Notes field. One or two clear sentences is enough.

✏️ Good note: "Removed hr:stair-tread — pool ramps go into water, stair treads go on steps. If there's a pool in the frame, it's always hr:pool-ramp."

Weak note: "Wrong tag."
8

Click Submit Review

Writes XMP metadata directly to the file on disk for the current image. The thumbnail dot turns green. The Last saved timestamp updates at the bottom of the sidebar. The image stays on screen — no auto-advance, so you can verify before moving on.

9

Repeat for each image

Use arrow keys or click thumbnails to navigate. Skip any image with a blue dot unless you disagree with the tags. When you navigate to a blue-dot image, the other reviewer's tags are pre-loaded — accept them by submitting as-is, or make corrections first.

10

When the batch is done, click Close Batch

Close Batch is permanent — it marks all reviewed images as done and moves them to processed/. They won't appear in future batches. Only click it when both reviewers are fully finished with the batch.

⚠️ Don't click Close Batch mid-session if you're planning to come back. Your submitted images are already saved to XMP — Close Batch is only for finalizing the whole batch.

Collaborative Review
Two reviewers, one batch, no conflicts.
The portal supports live multi-reviewer sessions. Two people can work the same batch simultaneously — the thumbnail strip shows who did what in real time.

There's no queuing, no locking, no coordination required. Open the portal on your machine, and whoever else is reviewing opens it on theirs. The thumbnail strip shows the live state of every image — who's submitted what, what has unsaved drafts, and what's still untouched.

Under the hood: every time you navigate to an image, the portal silently fetches the latest XMP metadata from the server. If another reviewer submitted that image two minutes ago, you'll see their tags pre-loaded and a blue dot on the thumbnail — even though you opened the portal before they did.

Status Dot Reference

🟢 Green — You submitted this

You submitted this image in this session. XMP is written to the file. Live sync won't overwrite your work if you navigate back to it.

🔵 Blue — Other reviewer submitted

Another reviewer has already submitted this image. Their tags load when you navigate to it. You can accept or override.

🟠 Orange — Unsaved draft

You've made changes (tags, notes, or rating) but haven't submitted yet. Nothing is written to disk — your draft is local only.

⚫ No dot — Untouched

Neither you nor any other reviewer has interacted with this image. Tags shown are AI-generated only.

How the Same Batch Looks to Each Reviewer
N
Nathan's view
Image 01
✓ I submitted
Image 02
Holly submitted
Image 03
My draft
Image 04
Untouched
Image 05
✓ I submitted
H
Holly's view — same batch, same time
Image 01
Nathan submitted
Image 02
✓ I submitted
Image 03
Untouched
Image 04
Untouched
Image 05
Nathan submitted

🔒 Built-In Safeguards

Practical Tips for Working Together


Fixing Mistakes
Re-editing tags on a reviewed image
Already submitted an image but the tags need a correction? Use the Search portal — no need to touch the review queue.
1

Open the Search portal

Go to dam.handiramp.com/search. This searches all reviewed images in the processed/ folder — the review queue is unaffected.

2

Find the image

Type any tag, product name, or descriptive term into the search box and click Search. Hyphens, spaces, and underscores are treated as equivalent — pool ramp and pool-ramp return the same results.

3

Click ✏️ Edit Tags on the image card

Each result card has an ✏️ Edit Tags button. Clicking it expands an inline form showing the current tags, star rating, and notes — all pre-loaded.

4

Make your corrections and save

Edit the tags, rating, or notes as needed, then click Save. Changes are written directly to the image file via XMP — the file never moves out of processed/. The form auto-collapses after a successful save and the card updates immediately.

💡 No queue disruption. Re-editing via Search writes directly to the reviewed file. It does not affect the active review batch or anyone else currently in the review portal.

Tag Reference
Three Tag Types
All three types are written directly into each image file as XMP keywords — cross-platform, no database, works in Bridge, Lightroom, and any XMP-aware tool.

Product / Hierarchy Tags brand prefix

Identify the specific product in the image. The prefix matches the brand. Every image should have at least one. Autocomplete suggestions come from the live worksheet — start typing to see options.

hr:yard-ramp hr:pool-ramp hr:van-ramp hr:threshold-ramp hr:stair-tread ht:stair-tread ht:nosing ps:folding-dog-ramp ps:puppy-treads ps:pool-leg-kit gb:gobuddy

Specification Tags spec: prefix

Capture product capacity and dimensions — assigned by reviewers only, not AI-inferred. Add a spec: tag when the product's dimension or rating is clearly identifiable in the image.

spec:1000lb spec:1500lb spec:30in spec:36in spec:45in spec:52in spec:20klb spec:30klb spec:50klb spec:78in

Content Tags AI-inferred + reviewer additions

Describe what's visually in the image — subject, setting, activity, condition, color. Free-form, no prefix required. AI-generated but fully editable. For yard ramps, color tags use the rail/frame color only (sc:black = black rails; sc:orange = orange rails — surface color is always dark and is not a basis for a color tag).

dog golden retriever large dog pool outdoor in use lifestyle product only loading dock rental installation wheelchair sc:black sc:orange level off 4 legs

How It Learns
Your Notes Become Rules
Every note you write gets distilled into a reusable rule that applies to every future batch automatically — for both reviewers, on every product line.

📝 You write this in the Notes field:

"Removed hr:stair-tread — the ramp is entering a pool, not a staircase. Pool ramps and stair treads are completely different products. If there's water + a dog + a ramp, it's always hr:pool-ramp."
↓   System distills this into a tagging rule   ↓
## Pool Ramp vs. Stair Tread - If water is visible in the frame AND a ramp is present: tag hr:pool-ramp, NOT hr:stair-tread - Pool ramps and stair treads are different products — never apply both - Trigger: water + ramp + dog → pool context, even if pool edge is not visible

That rule now applies to every image in every future batch — automatically, for both the local AI model and the cloud fallback.


Quick Reference

⌨️ Keyboard Shortcuts

Next image
Previous image
⌘ Enter Submit Review
Tab Jump to Notes field
Esc Clear tag input / close autocomplete
Navigate autocomplete list
Enter Select autocomplete suggestion
⌘ R Refresh batch state from server