UNPKG

dtamind-components

Version:

Apps integration for Dtamind. Contain Nodes and Credentials.

284 lines (281 loc) 11.9 kB
"use strict"; Object.defineProperty(exports, "__esModule", { value: true }); const nodevm_1 = require("@dtamindai/nodevm"); const utils_1 = require("../../../src/utils"); const messages_1 = require("@langchain/core/messages"); const uuid_1 = require("uuid"); class ExecuteFlow_SeqAgents { constructor() { //@ts-ignore this.loadMethods = { async listFlows(_, options) { const returnData = []; const appDataSource = options.appDataSource; const databaseEntities = options.databaseEntities; if (appDataSource === undefined || !appDataSource) { return returnData; } const searchOptions = options.searchOptions || {}; const chatflows = await appDataSource.getRepository(databaseEntities['ChatFlow']).findBy(searchOptions); for (let i = 0; i < chatflows.length; i += 1) { const data = { label: chatflows[i].name, name: chatflows[i].id }; returnData.push(data); } return returnData; } }; this.label = 'Execute Flow'; this.name = 'seqExecuteFlow'; this.version = 1.0; this.type = 'ExecuteFlow'; this.icon = 'executeflow.svg'; this.category = 'Sequential Agents'; this.description = `Execute chatflow/agentflow and return final response`; this.baseClasses = [this.type]; this.credential = { label: 'Connect Credential', name: 'credential', type: 'credential', credentialNames: ['chatflowApi'], optional: true }; this.inputs = [ { label: 'Sequential Node', name: 'sequentialNode', type: 'Start | Agent | Condition | LLMNode | ToolNode | CustomFunction | ExecuteFlow', description: 'Can be connected to one of the following nodes: Start, Agent, Condition, LLM Node, Tool Node, Custom Function, Execute Flow', list: true }, { label: 'Name', name: 'seqExecuteFlowName', type: 'string' }, { label: 'Select Flow', name: 'selectedFlow', type: 'asyncOptions', loadMethod: 'listFlows' }, { label: 'Input', name: 'seqExecuteFlowInput', type: 'options', description: 'Select one of the following or enter custom input', freeSolo: true, loadPreviousNodes: true, options: [ { label: '{{ question }}', name: 'userQuestion', description: 'Use the user question from the chat as input.' } ] }, { label: 'Override Config', name: 'overrideConfig', description: 'Override the config passed to the flow.', type: 'json', optional: true, additionalParams: true }, { label: 'Base URL', name: 'baseURL', type: 'string', description: 'Base URL to Dtamind. By default, it is the URL of the incoming request. Useful when you need to execute flow through an alternative route.', placeholder: 'http://localhost:3000', optional: true, additionalParams: true }, { label: 'Start new session per message', name: 'startNewSession', type: 'boolean', description: 'Whether to continue the session or start a new one with each interaction. Useful for flows with memory if you want to avoid it.', default: false, optional: true, additionalParams: true }, { label: 'Return Value As', name: 'returnValueAs', type: 'options', options: [ { label: 'AI Message', name: 'aiMessage' }, { label: 'Human Message', name: 'humanMessage' }, { label: 'State Object', name: 'stateObj', description: "Return as state object, ex: { foo: bar }. This will update the custom state 'foo' to 'bar'" } ], default: 'aiMessage' } ]; } async init(nodeData, input, options) { const selectedFlowId = nodeData.inputs?.selectedFlow; const _seqExecuteFlowName = nodeData.inputs?.seqExecuteFlowName; if (!_seqExecuteFlowName) throw new Error('Execute Flow node name is required!'); const seqExecuteFlowName = _seqExecuteFlowName.toLowerCase().replace(/\s/g, '_').trim(); const startNewSession = nodeData.inputs?.startNewSession; const appDataSource = options.appDataSource; const databaseEntities = options.databaseEntities; const sequentialNodes = nodeData.inputs?.sequentialNode; const seqExecuteFlowInput = nodeData.inputs?.seqExecuteFlowInput; const overrideConfig = typeof nodeData.inputs?.overrideConfig === 'string' && nodeData.inputs.overrideConfig.startsWith('{') && nodeData.inputs.overrideConfig.endsWith('}') ? JSON.parse(nodeData.inputs.overrideConfig) : nodeData.inputs?.overrideConfig; if (!sequentialNodes || !sequentialNodes.length) throw new Error('Execute Flow must have a predecessor!'); const baseURL = nodeData.inputs?.baseURL || options.baseURL; const returnValueAs = nodeData.inputs?.returnValueAs; const credentialData = await (0, utils_1.getCredentialData)(nodeData.credential ?? '', options); const chatflowApiKey = (0, utils_1.getCredentialParam)('chatflowApiKey', credentialData, nodeData); if (selectedFlowId === options.chatflowid) throw new Error('Cannot call the same agentflow!'); let headers = {}; if (chatflowApiKey) headers = { Authorization: `Bearer ${chatflowApiKey}` }; const chatflowId = options.chatflowid; const sessionId = options.sessionId; const chatId = options.chatId; const executeFunc = async (state) => { const variables = await (0, utils_1.getVars)(appDataSource, databaseEntities, nodeData, options); let flowInput = ''; if (seqExecuteFlowInput === 'userQuestion') { flowInput = input; } else if (seqExecuteFlowInput && seqExecuteFlowInput.startsWith('{{') && seqExecuteFlowInput.endsWith('}}')) { const nodeId = seqExecuteFlowInput.replace('{{', '').replace('}}', '').replace('$', '').trim(); const messageOutputs = (state.messages ?? []).filter((message) => message.additional_kwargs && message.additional_kwargs?.nodeId === nodeId); const messageOutput = messageOutputs[messageOutputs.length - 1]; if (messageOutput) { flowInput = JSON.stringify(messageOutput.content); } } const flow = { chatflowId, sessionId, chatId, input: flowInput, state }; const body = { question: flowInput, chatId: startNewSession ? (0, uuid_1.v4)() : chatId, overrideConfig: { sessionId: startNewSession ? (0, uuid_1.v4)() : sessionId, ...(overrideConfig ?? {}) } }; const callOptions = { method: 'POST', headers: { 'Content-Type': 'application/json', ...headers }, body: JSON.stringify(body) }; let sandbox = { $input: flowInput, $callOptions: callOptions, $callBody: body, util: undefined, Symbol: undefined, child_process: undefined, fs: undefined, process: undefined }; sandbox['$vars'] = (0, utils_1.prepareSandboxVars)(variables); sandbox['$flow'] = flow; const code = ` const fetch = require('node-fetch'); const url = "${baseURL}/api/v1/prediction/${selectedFlowId}"; const body = $callBody; const options = $callOptions; try { const response = await fetch(url, options); const resp = await response.json(); return resp.text; } catch (error) { console.error(error); return ''; } `; const builtinDeps = process.env.TOOL_FUNCTION_BUILTIN_DEP ? utils_1.defaultAllowBuiltInDep.concat(process.env.TOOL_FUNCTION_BUILTIN_DEP.split(',')) : utils_1.defaultAllowBuiltInDep; const externalDeps = process.env.TOOL_FUNCTION_EXTERNAL_DEP ? process.env.TOOL_FUNCTION_EXTERNAL_DEP.split(',') : []; const deps = utils_1.availableDependencies.concat(externalDeps); const nodeVMOptions = { console: 'inherit', sandbox, require: { external: { modules: deps }, builtin: builtinDeps }, eval: false, wasm: false, timeout: 10000 }; const vm = new nodevm_1.NodeVM(nodeVMOptions); try { let response = await vm.run(`module.exports = async function() {${code}}()`, __dirname); if (typeof response === 'object') { response = JSON.stringify(response); } if (returnValueAs === 'humanMessage') { return { messages: [ new messages_1.HumanMessage({ content: response, additional_kwargs: { nodeId: nodeData.id } }) ] }; } return { messages: [ new messages_1.AIMessage({ content: response, additional_kwargs: { nodeId: nodeData.id } }) ] }; } catch (e) { throw new Error(e); } }; const startLLM = sequentialNodes[0].startLLM; const returnOutput = { id: nodeData.id, node: executeFunc, name: seqExecuteFlowName, label: _seqExecuteFlowName, type: 'utilities', output: 'ExecuteFlow', llm: startLLM, startLLM, multiModalMessageContent: sequentialNodes[0]?.multiModalMessageContent, predecessorAgents: sequentialNodes }; return returnOutput; } } module.exports = { nodeClass: ExecuteFlow_SeqAgents }; //# sourceMappingURL=ExecuteFlow.js.map