Chore/cohere embed models (#2035)

* Update of Cohere embedding models

* Update of Cohere embedding models

* adding official docs link to inputType
This commit is contained in:
Vinod Paidimarry
2024-03-26 11:50:50 +05:30
committed by GitHub
parent 794818b434
commit f116dba84e
@@ -17,7 +17,7 @@ class CohereEmbedding_Embeddings implements INode {
constructor() {
this.label = 'Cohere Embeddings'
this.name = 'cohereEmbeddings'
this.version = 1.0
this.version = 2.0
this.type = 'CohereEmbeddings'
this.icon = 'Cohere.svg'
this.category = 'Embeddings'
@@ -35,27 +35,82 @@ class CohereEmbedding_Embeddings implements INode {
name: 'modelName',
type: 'options',
options: [
{
label: 'embed-english-v3.0',
name: 'embed-english-v3.0',
description: 'Embedding Dimensions: 1024'
},
{
label: 'embed-english-light-v3.0',
name: 'embed-english-light-v3.0',
description: 'Embedding Dimensions: 384'
},
{
label: 'embed-multilingual-v3.0',
name: 'embed-multilingual-v3.0',
description: 'Embedding Dimensions: 1024'
},
{
label: 'embed-multilingual-light-v3.0',
name: 'embed-multilingual-light-v3.0',
description: 'Embedding Dimensions: 384'
},
{
label: 'embed-english-v2.0',
name: 'embed-english-v2.0'
name: 'embed-english-v2.0',
description: 'Embedding Dimensions: 4096'
},
{
label: 'embed-english-light-v2.0',
name: 'embed-english-light-v2.0'
name: 'embed-english-light-v2.0',
description: 'Embedding Dimensions: 1024'
},
{
label: 'embed-multilingual-v2.0',
name: 'embed-multilingual-v2.0'
name: 'embed-multilingual-v2.0',
description: 'Embedding Dimensions: 768'
}
],
default: 'embed-english-v2.0',
optional: true
},
{
label: 'Type',
name: 'inputType',
type: 'options',
description:
'Specifies the type of input passed to the model. Required for embedding models v3 and higher. <a target="_blank" href="https://docs.cohere.com/reference/embed">Official Docs</a>',
options: [
{
label: 'search_document',
name: 'search_document',
description: 'Use this to encode documents for embeddings that you store in a vector database for search use-cases'
},
{
label: 'search_query',
name: 'search_query',
description: 'Use this when you query your vector DB to find relevant documents.'
},
{
label: 'classification',
name: 'classification',
description: 'Use this when you use the embeddings as an input to a text classifier'
},
{
label: 'clustering',
name: 'clustering',
description: 'Use this when you want to cluster the embeddings.'
}
],
default: 'search_query',
optional: true
}
]
}
async init(nodeData: INodeData, _: string, options: ICommonObject): Promise<any> {
const modelName = nodeData.inputs?.modelName as string
const inputType = nodeData.inputs?.inputType as string
const credentialData = await getCredentialData(nodeData.credential ?? '', options)
const cohereApiKey = getCredentialParam('cohereApiKey', credentialData, nodeData)
@@ -65,6 +120,7 @@ class CohereEmbedding_Embeddings implements INode {
}
if (modelName) obj.model = modelName
if (inputType) obj.inputType = inputType
const model = new CohereEmbeddings(obj)
return model