Merge pull request FlowiseAI#1012 from FlowiseAI/feature/SimilarityThresholdRetriever

Feature/Similarity Threshold Retriever

Author: HenryHengZJ

packages/components/nodes/retrievers/SimilarityThresholdRetriever/SimilarityThresholdRetriever.ts

@@ -0,0 +1,107 @@
+import { VectorStore } from 'langchain/vectorstores/base'
+import { INode, INodeData, INodeParams, INodeOutputsValue } from '../../../src/Interface'
+import { handleEscapeCharacters } from '../../../src'
+import { ScoreThresholdRetriever } from 'langchain/retrievers/score_threshold'
+
+class SimilarityThresholdRetriever_Retrievers implements INode {
+    label: string
+    name: string
+    version: number
+    description: string
+    type: string
+    icon: string
+    category: string
+    baseClasses: string[]
+    inputs: INodeParams[]
+    outputs: INodeOutputsValue[]
+
+    constructor() {
+        this.label = 'Similarity Score Threshold Retriever'
+        this.name = 'similarityThresholdRetriever'
+        this.version = 1.0
+        this.type = 'SimilarityThresholdRetriever'
+        this.icon = 'similaritythreshold.svg'
+        this.category = 'Retrievers'
+        this.description = 'Return results based on the minimum similarity percentage'
+        this.baseClasses = [this.type, 'BaseRetriever']
+        this.inputs = [
+            {
+                label: 'Vector Store',
+                name: 'vectorStore',
+                type: 'VectorStore'
+            },
+            {
+                label: 'Minimum Similarity Score (%)',
+                name: 'minSimilarityScore',
+                description: 'Finds results with at least this similarity score',
+                type: 'number',
+                default: 80,
+                step: 1
+            },
+            {
+                label: 'Max K',
+                name: 'maxK',
+                description: `The maximum number of results to fetch`,
+                type: 'number',
+                default: 20,
+                step: 1
+            },
+            {
+                label: 'K Increment',
+                name: 'kIncrement',
+                description: `How much to increase K by each time. It'll fetch N results, then N + kIncrement, then N + kIncrement * 2, etc.`,
+                type: 'number',
+                default: 2,
+                step: 1
+            }
+        ]
+        this.outputs = [
+            {
+                label: 'Similarity Threshold Retriever',
+                name: 'retriever',
+                baseClasses: this.baseClasses
+            },
+            {
+                label: 'Document',
+                name: 'document',
+                baseClasses: ['Document']
+            },
+            {
+                label: 'Text',
+                name: 'text',
+                baseClasses: ['string', 'json']
+            }
+        ]
+    }
+
+    async init(nodeData: INodeData, input: string): Promise<any> {
+        const vectorStore = nodeData.inputs?.vectorStore as VectorStore
+        const minSimilarityScore = nodeData.inputs?.minSimilarityScore as number
+        const maxK = nodeData.inputs?.maxK as string
+        const kIncrement = nodeData.inputs?.kIncrement as string
+
+        const output = nodeData.outputs?.output as string
+
+        const retriever = ScoreThresholdRetriever.fromVectorStore(vectorStore, {
+            minSimilarityScore: minSimilarityScore ? minSimilarityScore / 100 : 0.9,
+            maxK: maxK ? parseInt(maxK, 10) : 100,
+            kIncrement: kIncrement ? parseInt(kIncrement, 10) : 2
+        })
+
+        if (output === 'retriever') return retriever
+        else if (output === 'document') return await retriever.getRelevantDocuments(input)
+        else if (output === 'text') {
+            let finaltext = ''
+
+            const docs = await retriever.getRelevantDocuments(input)
+
+            for (const doc of docs) finaltext += `${doc.pageContent}\n`
+
+            return handleEscapeCharacters(finaltext, false)
+        }
+
+        return retriever
+    }
+}
+
+module.exports = { nodeClass: SimilarityThresholdRetriever_Retrievers }