> ## Documentation Index
> Fetch the complete documentation index at: https://docs.trychannel3.com/llms.txt
> Use this file to discover all available pages before exploring further.

# v4.2.0 — October 2026

`@channel3/sdk@4.2.0` and `channel3-sdk==4.2.0` are released. Every change is additive.

**New**

* [Product stickers](/sdk/sticker) (alpha): [`POST /experimental/sticker`](/api-reference/v1/sticker) returns a transparent-background cutout of a catalog product, from a `product_id` or a product page `url`. It costs 1 credit per call. In the SDKs it is `experimental.getSticker` / `experimental.get_sticker`.

**Websites**

* `websites.search` ([`GET /v1/websites/search`](/api-reference/v1/search-websites)) finds a website by retailer name or domain.
* `websites.list` ([`GET /v1/websites`](/api-reference/v1/list-websites)) pages through supported websites, optionally narrowed to those selling given `category_ids`.
* See [Websites](/sdk/websites). [`GET /v1/websites/{website_id}`](/api-reference/v1/website-detail) accepts an ID or a domain. It is REST-only until the next major version, when it takes over the `websites.retrieve` name.

**Find products in image**

* `return_masks: true` adds `things[].masks`: one segmentation mask per physical object (a pair of shoes gets two), each a `bbox` with a COCO RLE `rle` over it.
* `things[].category_ids` and `things[].gender` report the category and gender filters applied to each thing's catalog search.

**Products**

* `colors`: the cleaned main image's palette as `{hex, percentage}`, ordered by pixel share.
* `ratings`: a cross-merchant rollup as `{value, count, source_count}`. `value` is an average on a 0–5 scale, weighted by review count. `ratings` is null when no merchant has a usable rating.

**Conversations**

* `ToolPart.input` is typed for the agent's image search, URL lookup and similar-products tools: `ImageSearchInput`, `UrlLookupInput` and `SimilarProductsInput`.
* Image search tool output groups matches by detected thing: `output.things[]`, each with a `thing` and its `matches`.

**Deprecated**

* `websites.retrieve` (`GET /v0/websites`). Use `websites.search`, or `GET /v1/websites/{website_id}` over REST, which accepts domains as well as website ids. The method keeps working until the next major version.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.