mirror of
https://github.com/hanzoai/cdn.git
synced 2026-10-30 15:06:53 +00:00
zeekayandHanzo Dev
edabf3842c
The Market read plane is files and nothing else. This makes that actually true
and actually deployable.
Catalog: was 10 mock "zoo-eco-agent" placeholders + 5 mock tools, which every
app would have shown to every user. Now generated from the curated flatfile
catalog (hanzoai/store data/{agents,tools}) -> 23 agents + 179 tools, each with
its own detail document. Drops purchases.json: per-user state has no business
on a public CDN. Strips upstream icon URLs, which are expired presigned R2 links
that carry an AWS credential + signature in the query string.
Versioning: index.json gains `schema` (breaking-shape gate, so a client that
does not recognise it keeps its bundle) and a `version` that is a content digest
over every served document. It changes iff what we serve changes, which is what
makes a deploy checkable rather than hopeful.
One writer: gen-market.py -> market.py with `gen` and `check`. It emits the CDN
documents AND each app's offline bundle, so the two cannot disagree about the
schema; `check` re-derives the digest and validates counts, ids, dangling
featured references and per-tool details, with no network and no other checkout.
Deploy: upload, then purge the Cloudflare edge for exactly the URLs written,
then verify the live manifest matches the repo and fail if it does not. Also
resolves wrangler once instead of re-resolving it per object.
CI: hanzo.yml + a 7-line cicd.yml importing hanzoai/ci. Static assets, so no
images:/deploy: — test: is the gate.
Co-authored-by: Hanzo Dev <[email protected]>
50 lines
2.3 KiB
JSON
50 lines
2.3 KiB
JSON
{
|
|
"author": "hanzo.ai",
|
|
"category": {
|
|
"description": "",
|
|
"examples": "",
|
|
"id": "photo-video",
|
|
"name": "Photo & Video"
|
|
},
|
|
"changelog": "",
|
|
"created_at": "2025-04-30T15:52:12.893Z",
|
|
"description": "This tool uses a vision capable LLM to classify images in a folder based on your custom labels, organizes them into subfolders (optional), and creates a COCO-format JSON dataset. It provides a clear summary of results, including any issues like low-confidence labels.\n\nHow to Use :\n\n1. Put your images (.jpg, .png, etc.) in a folder.\n\n2. Specify the folder path, a list of labels (e.g., \"cat\", \"dog\"), and your vision AI model.\n\n3. Choose how to handle images : just label them, copy to label subfolders, or move to subfolders.\n\n4. Set optional preferences, like picking the highest-confidence label instead of all labels, or defining a minimum confidence score.\n\n5. Run the tool to classify images, save a COCO JSON file, and organize images (if selected).\n\n6. Check the JSON file for classifications, subfolders with classified images, and a summary of results.\n\nNotes :\n\n- The tool checks AI outputs for errors, implement retries, and reports issues like invalid labels.\n\n- Supports multiple labels per image.\n\n- If moving images, originals are deleted only after successful organization.\n\n- In configuration you can pick which types of errors to include in the output.\n\nWARNING : possible high LLM usage, depending on the number of images in the folder.\n\n",
|
|
"downloads": 17,
|
|
"icon_url": "",
|
|
"id": "images-folder-classifier-using-vision-llm",
|
|
"install_command": "",
|
|
"license": "MIT",
|
|
"mcp_config": {
|
|
"args": [
|
|
"local:::__gilles_test2_sep_shinkai:::images_folder_classifier___using_vision_llm"
|
|
],
|
|
"command": "any",
|
|
"env": {}
|
|
},
|
|
"name": "Images folder classifier - using vision LLM",
|
|
"operating_system": [
|
|
"linux",
|
|
"macos",
|
|
"windows"
|
|
],
|
|
"path": "/tools/images-folder-classifier-using-vision-llm",
|
|
"price": 0,
|
|
"readme": "",
|
|
"repository": "github.com/hanzoai/tools",
|
|
"router_key": "@hanzo/images-folder-classifier---using-vision-llm",
|
|
"runner": "any",
|
|
"screenshots": [],
|
|
"tags": [
|
|
"image",
|
|
"classification",
|
|
"vision LLM",
|
|
"COCO",
|
|
"folder",
|
|
"labeling",
|
|
"sorting"
|
|
],
|
|
"type": "Tool",
|
|
"updated_at": "2025-04-30T15:52:12.893Z",
|
|
"version": "1.0.0"
|
|
}
|