Merge pull request #511 from FlowiseAI/feature/ConversationalRetrievalQAChain-Memory

Feature/conversational retrieval qa chain memory

Author: HenryHengZJ

packages/components/nodes/chains/ConversationalRetrievalQAChain/ConversationalRetrievalQAChain.ts

@@ -2,7 +2,7 @@ import { BaseLanguageModel } from 'langchain/base_language'
 import { ICommonObject, IMessage, INode, INodeData, INodeParams } from '../../../src/Interface'
 import { CustomChainHandler, getBaseClasses } from '../../../src/utils'
 import { ConversationalRetrievalQAChain } from 'langchain/chains'
-import { AIChatMessage, BaseRetriever, HumanChatMessage } from 'langchain/schema'
+import { AIMessage, BaseRetriever, HumanMessage } from 'langchain/schema'
 import { BaseChatMemory, BufferMemory, ChatMessageHistory } from 'langchain/memory'
 import { PromptTemplate } from 'langchain/prompts'
 
@@ -20,6 +20,20 @@ const qa_template = `Use the following pieces of context to answer the question
 Question: {question}
 Helpful Answer:`
 
+const CUSTOM_QUESTION_GENERATOR_CHAIN_PROMPT = `Given the following conversation and a follow up question, return the conversation history excerpt that includes any relevant context to the question if it exists and rephrase the follow up question to be a standalone question.
+Chat History:
+{chat_history}
+Follow Up Input: {question}
+Your answer should follow the following format:
+\`\`\`
+Use the following pieces of context to answer the users question.
+If you don't know the answer, just say that you don't know, don't try to make up an answer.
+----------------
+<Relevant chat history excerpt as context here>
+Standalone question: <Rephrased question here>
+\`\`\`
+Your answer:`
+
 class ConversationalRetrievalQAChain_Chains implements INode {
     label: string
     name: string
@@ -49,6 +63,13 @@ class ConversationalRetrievalQAChain_Chains implements INode {
                 name: 'vectorStoreRetriever',
                 type: 'BaseRetriever'
             },
+            {
+                label: 'Memory',
+                name: 'memory',
+                type: 'DynamoDBChatMemory | RedisBackedChatMemory | ZepMemory',
+                optional: true,
+                description: 'If no memory connected, default BufferMemory will be used'
+            },
             {
                 label: 'Return Source Documents',
                 name: 'returnSourceDocuments',
@@ -99,22 +120,33 @@ class ConversationalRetrievalQAChain_Chains implements INode {
         const systemMessagePrompt = nodeData.inputs?.systemMessagePrompt as string
         const returnSourceDocuments = nodeData.inputs?.returnSourceDocuments as boolean
         const chainOption = nodeData.inputs?.chainOption as string
+        const memory = nodeData.inputs?.memory
 
         const obj: any = {
             verbose: process.env.DEBUG === 'true' ? true : false,
             qaChainOptions: {
                 type: 'stuff',
                 prompt: PromptTemplate.fromTemplate(systemMessagePrompt ? `${systemMessagePrompt}\n${qa_template}` : default_qa_template)
             },
-            memory: new BufferMemory({
+            questionGeneratorChainOptions: {
+                template: CUSTOM_QUESTION_GENERATOR_CHAIN_PROMPT
+            }
+        }
+        if (returnSourceDocuments) obj.returnSourceDocuments = returnSourceDocuments
+        if (chainOption) obj.qaChainOptions = { ...obj.qaChainOptions, type: chainOption }
+        if (memory) {
+            memory.inputKey = 'question'
+            memory.outputKey = 'text'
+            memory.memoryKey = 'chat_history'
+            obj.memory = memory
+        } else {
+            obj.memory = new BufferMemory({
                 memoryKey: 'chat_history',
                 inputKey: 'question',
                 outputKey: 'text',
                 returnMessages: true
             })
         }
-        if (returnSourceDocuments) obj.returnSourceDocuments = returnSourceDocuments
-        if (chainOption) obj.qaChainOptions = { ...obj.qaChainOptions, type: chainOption }
 
         const chain = ConversationalRetrievalQAChain.fromLLM(model, vectorStoreRetriever, obj)
         return chain
@@ -123,6 +155,8 @@ class ConversationalRetrievalQAChain_Chains implements INode {
     async run(nodeData: INodeData, input: string, options: ICommonObject): Promise<string | ICommonObject> {
         const chain = nodeData.instance as ConversationalRetrievalQAChain
         const returnSourceDocuments = nodeData.inputs?.returnSourceDocuments as boolean
+        const memory = nodeData.inputs?.memory
+
         let model = nodeData.inputs?.model
 
         // Temporary fix: https://github.com/hwchase17/langchainjs/issues/754
@@ -131,16 +165,17 @@ class ConversationalRetrievalQAChain_Chains implements INode {
 
         const obj = { question: input }
 
-        if (chain.memory && options && options.chatHistory) {
+        // If external memory like Zep, Redis is being used, ignore below
+        if (!memory && chain.memory && options && options.chatHistory) {
             const chatHistory = []
             const histories: IMessage[] = options.chatHistory
             const memory = chain.memory as BaseChatMemory
 
             for (const message of histories) {
                 if (message.type === 'apiMessage') {
-                    chatHistory.push(new AIChatMessage(message.message))
+                    chatHistory.push(new AIMessage(message.message))
                 } else if (message.type === 'userMessage') {
-                    chatHistory.push(new HumanChatMessage(message.message))
+                    chatHistory.push(new HumanMessage(message.message))
                 }
             }
             memory.chatHistory = new ChatMessageHistory(chatHistory)

packages/components/nodes/memory/DynamoDb/DynamoDb.ts

@@ -13,8 +13,9 @@ class DynamoDb_Memory implements INode {
     inputs: INodeParams[]
 
     constructor() {
-        this.label = 'DynamoDB Memory'
-        this.name = 'DynamoDbMemory'
+        this.label = 'DynamoDB Chat Memory'
+        this.name = 'DynamoDBChatMemory'
+        this.type = 'DynamoDBChatMemory'
         this.icon = 'dynamodb.svg'
         this.category = 'Memory'
         this.description = 'Stores the conversation in dynamo db table'

packages/components/nodes/memory/ZepMemory/ZepMemory.ts

@@ -18,7 +18,7 @@ class ZepMemory_Memory implements INode {
         this.label = 'Zep Memory'
         this.name = 'ZepMemory'
         this.type = 'ZepMemory'
-        this.icon = 'memory.svg'
+        this.icon = 'zep.png'
         this.category = 'Memory'
         this.description = 'Summarizes the conversation and stores the memory in zep server'
         this.baseClasses = [this.type, ...getBaseClasses(ZepMemory)]

packages/components/nodes/memory/ZepMemory/memory.svg

@@ -539,6 +539,14 @@
                         "name": "vectorStoreRetriever",
                         "type": "BaseRetriever",
                         "id": "conversationalRetrievalQAChain_0-input-vectorStoreRetriever-BaseRetriever"
+                    },
+                    {
+                        "label": "Memory",
+                        "name": "memory",
+                        "type": "DynamoDBChatMemory | RedisBackedChatMemory | ZepMemory",
+                        "optional": true,
+                        "description": "If no memory connected, default BufferMemory will be used",
+                        "id": "conversationalRetrievalQAChain_0-input-memory-DynamoDBChatMemory | RedisBackedChatMemory | ZepMemory"
                     }
                 ],
                 "inputs": {

packages/components/nodes/memory/ZepMemory/zep.png

@@ -556,6 +556,14 @@
                         "name": "vectorStoreRetriever",
                         "type": "BaseRetriever",
                         "id": "conversationalRetrievalQAChain_0-input-vectorStoreRetriever-BaseRetriever"
+                    },
+                    {
+                        "label": "Memory",
+                        "name": "memory",
+                        "type": "DynamoDBChatMemory | RedisBackedChatMemory | ZepMemory",
+                        "optional": true,
+                        "description": "If no memory connected, default BufferMemory will be used",
+                        "id": "conversationalRetrievalQAChain_0-input-memory-DynamoDBChatMemory | RedisBackedChatMemory | ZepMemory"
                     }
                 ],
                 "inputs": {

packages/server/marketplaces/chatflows/Conversational Retrieval QA Chain.json

@@ -131,6 +131,14 @@
                         "name": "vectorStoreRetriever",
                         "type": "BaseRetriever",
                         "id": "conversationalRetrievalQAChain_0-input-vectorStoreRetriever-BaseRetriever"
+                    },
+                    {
+                        "label": "Memory",
+                        "name": "memory",
+                        "type": "DynamoDBChatMemory | RedisBackedChatMemory | ZepMemory",
+                        "optional": true,
+                        "description": "If no memory connected, default BufferMemory will be used",
+                        "id": "conversationalRetrievalQAChain_0-input-memory-DynamoDBChatMemory | RedisBackedChatMemory | ZepMemory"
                     }
                 ],
                 "inputs": {

packages/server/marketplaces/chatflows/Github Repo QnA.json

@@ -421,6 +421,14 @@
                         "name": "vectorStoreRetriever",
                         "type": "BaseRetriever",
                         "id": "conversationalRetrievalQAChain_0-input-vectorStoreRetriever-BaseRetriever"
+                    },
+                    {
+                        "label": "Memory",
+                        "name": "memory",
+                        "type": "DynamoDBChatMemory | RedisBackedChatMemory | ZepMemory",
+                        "optional": true,
+                        "description": "If no memory connected, default BufferMemory will be used",
+                        "id": "conversationalRetrievalQAChain_0-input-memory-DynamoDBChatMemory | RedisBackedChatMemory | ZepMemory"
                     }
                 ],
                 "inputs": {

packages/server/marketplaces/chatflows/Local QnA.json

@@ -625,6 +625,14 @@
                         "name": "vectorStoreRetriever",
                         "type": "BaseRetriever",
                         "id": "conversationalRetrievalQAChain_0-input-vectorStoreRetriever-BaseRetriever"
+                    },
+                    {
+                        "label": "Memory",
+                        "name": "memory",
+                        "type": "DynamoDBChatMemory | RedisBackedChatMemory | ZepMemory",
+                        "optional": true,
+                        "description": "If no memory connected, default BufferMemory will be used",
+                        "id": "conversationalRetrievalQAChain_0-input-memory-DynamoDBChatMemory | RedisBackedChatMemory | ZepMemory"
                     }
                 ],
                 "inputs": {

packages/server/marketplaces/chatflows/Metadata Filter Load.json

@@ -168,8 +168,10 @@ export const isValidConnection = (connection, reactFlowInstance) => {
     //sourceHandle: "llmChain_0-output-llmChain-BaseChain"
     //targetHandle: "mrlkAgentLLM_0-input-model-BaseLanguageModel"
 
-    const sourceTypes = sourceHandle.split('-')[sourceHandle.split('-').length - 1].split('|')
-    const targetTypes = targetHandle.split('-')[targetHandle.split('-').length - 1].split('|')
+    let sourceTypes = sourceHandle.split('-')[sourceHandle.split('-').length - 1].split('|')
+    sourceTypes = sourceTypes.map((s) => s.trim())
+    let targetTypes = targetHandle.split('-')[targetHandle.split('-').length - 1].split('|')
+    targetTypes = targetTypes.map((t) => t.trim())
 
     if (targetTypes.some((t) => sourceTypes.includes(t))) {
         let targetNode = reactFlowInstance.getNode(target)