https://github.com/paulkinlan/transformerjs-breadboard-kit
https://github.com/paulkinlan/transformerjs-breadboard-kit
Last synced: about 2 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/paulkinlan/transformerjs-breadboard-kit
- Owner: PaulKinlan
- License: apache-2.0
- Created: 2023-10-06T10:49:25.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2024-06-16T09:39:07.000Z (11 months ago)
- Last Synced: 2025-03-23T23:50:08.751Z (about 2 months ago)
- Language: TypeScript
- Size: 187 KB
- Stars: 2
- Watchers: 2
- Forks: 1
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# "Google Labs Breadboard" TransformerJS Nodes Kit
This is a collection of [Breadboard](https://github.com/google/labs-prototypes/tree/main/seeds/breadboard) nodes that let you integrate `transformerjs` in to your apps built with Breadboard.
## Installing
This Kit requires Node version >=v19.0.0. To install:
```sh
npm install @paulkinlan/transformerjs-breadboard-kit
```
## Node Types
Here are all node handlers that are included in the OpenAI Breadboard Kit
### The `pipeline` node
Takes an `input`
#### Inputs:
- `input` required. The value to be checked if undefined.
- `task` required. The transformerjs task to be used (tested: `sentiment-analysis`, `summarize`)
- `model` optional.
#### Outputs:
- `modelResult` - likely an array of results returned by the `transfomerjs` logic as defined by `task` and `model`
### The `summarize` node
- `input` required. The text to be summarized.
- `model` optional. The model to use. This will be loaded by transformjs. The default is `Xenova/distilbart-cnn-6-6`.
#### Outputs:
- `modelResult` - an array of `{ "summary_text": "[some text]" }`
### The `sentiment-analysis` node
- `input` required. The text to be analysed for sentiment.
- `model` optional. The model to use. This will be loaded by transformjs. The default is `Xenova/distilbert-base-uncased-finetuned-sst-2-english`.
#### Outputs:
- `modelResult` - an array of objects of the form: `{ "label": "POSITIVE", "confidence": 0.9995 }`