semillero-2-AI-ATS/scripts/deploy-n8n-v2.ts
2026-06-09 10:37:31 -04:00

651 lines
21 KiB
TypeScript

import * as fs from "fs";
import * as path from "path";
import * as readline from "readline";
// Load .env variables
const envPath = path.join(__dirname, "../.env");
if (fs.existsSync(envPath)) {
const envContent = fs.readFileSync(envPath, "utf8");
for (const line of envContent.split("\n")) {
const match = line.match(/^\s*([\w.-]+)\s*=\s*(.*)\s*$/);
if (match) {
const key = match[1];
let value = match[2].trim();
if (value.startsWith('"') && value.endsWith('"')) value = value.slice(1, -1);
else if (value.startsWith("'") && value.endsWith("'")) value = value.slice(1, -1);
process.env[key] = value;
}
}
}
const N8N_HOST = process.env.N8N_HOST || "https://n8n.gaboggamer.online";
const N8N_API_KEY = process.env.N8N_API_KEY;
const SUPABASE_URL = process.env.NEXT_PUBLIC_SUPABASE_URL;
const SUPABASE_SECRET_KEY = process.env.SUPABASE_SECRET_KEY;
if (!N8N_API_KEY) {
console.error("Error: N8N_API_KEY is not defined in .env");
process.exit(1);
}
// Interactive prompt helper
function askQuestion(query: string): Promise<string> {
const rl = readline.createInterface({
input: process.stdin,
output: process.stdout,
});
return new Promise((resolve) =>
rl.question(query, (ans) => {
rl.close();
resolve(ans.trim());
})
);
}
async function n8nRequest(endpoint: string, method: string = "GET", body?: any) {
const response = await fetch(`${N8N_HOST}${endpoint}`, {
method,
headers: {
"X-N8N-API-KEY": N8N_API_KEY!,
"Content-Type": "application/json",
},
body: body ? JSON.stringify(body) : undefined,
});
if (!response.ok) {
const text = await response.text();
throw new Error(`n8n API request failed: ${response.status} ${response.statusText} - ${text}`);
}
return response.json();
}
async function getOrCreateCredential(name: string, type: string, data: any) {
try {
const credsList = await n8nRequest("/api/v1/credentials");
const existingCred = credsList.data.find((c: any) => c.name === name && c.type === type);
if (existingCred) {
console.log(`Reusing existing credential: ${name} (ID: ${existingCred.id})`);
return existingCred.id;
} else {
const newCred = await n8nRequest("/api/v1/credentials", "POST", {
name,
type,
data,
});
console.log(`Created new credential: ${name} (ID: ${newCred.id})`);
return newCred.id;
}
} catch (err: any) {
console.error(`Error setting up credential ${name}:`, err.message);
process.exit(1);
}
}
interface ProviderConfig {
nodeType: string;
credentialType: string;
credentialData: any;
nodeParameters: any;
}
function getProviderConfig(provider: string, apiKey: string, modelName: string): ProviderConfig {
switch (provider) {
case "1": // Deepseek (Native)
return {
nodeType: "@n8n/n8n-nodes-langchain.lmChatDeepSeek",
credentialType: "deepSeekApi",
credentialData: {
apiKey,
allowedHttpRequestDomains: "none",
},
nodeParameters: {
model: modelName || "deepseek-chat",
options: {},
},
};
case "2": // OpenAI
return {
nodeType: "@n8n/n8n-nodes-langchain.lmChatOpenAi",
credentialType: "openAiApi",
credentialData: {
apiKey,
header: false,
allowedHttpRequestDomains: "none",
},
nodeParameters: {
model: modelName || "gpt-4o-mini",
options: {},
},
};
case "3": // Google Gemini
return {
nodeType: "@n8n/n8n-nodes-langchain.lmChatGoogleGemini",
credentialType: "googlePalmApi",
credentialData: {
apiKey,
host: "https://generativelanguage.googleapis.com",
allowedHttpRequestDomains: "none",
},
nodeParameters: {
model: modelName || "gemini-1.5-flash",
options: {},
},
};
case "4": // Anthropic
return {
nodeType: "@n8n/n8n-nodes-langchain.lmChatAnthropic",
credentialType: "anthropicApi",
credentialData: {
apiKey,
allowedHttpRequestDomains: "none",
},
nodeParameters: {
model: modelName || "claude-3-5-sonnet-latest",
options: {},
},
};
default:
throw new Error("Invalid provider chosen");
}
}
async function main() {
console.log("\n==================================================");
console.log("Welcome to interactive n8n workflow deployment");
console.log("==================================================");
// 1. Ask for Primary Provider
console.log("\nSelect Primary LLM Provider:");
console.log("1. Deepseek (Native Node)");
console.log("2. OpenAI (Standard)");
console.log("3. Google Gemini");
console.log("4. Anthropic");
const primaryProviderChoice = (await askQuestion("Enter choice (1-4) [default: 3]: ")) || "3";
let defaultModel = "gemini-1.5-flash";
if (primaryProviderChoice === "1") defaultModel = "deepseek-chat";
else if (primaryProviderChoice === "2") defaultModel = "gpt-4o-mini";
else if (primaryProviderChoice === "4") defaultModel = "claude-3-5-sonnet-latest";
const primaryModelName = (await askQuestion(`Enter primary model name [default: ${defaultModel}]: `)) || defaultModel;
let defaultKey = "";
if (primaryProviderChoice === "1") defaultKey = process.env.DEEPSEEK_API_KEY || "";
else if (primaryProviderChoice === "3") defaultKey = process.env.GEMINI_API_KEY || "";
const primaryApiKey = (await askQuestion(`Enter API key [default: ${defaultKey ? "Loaded from .env" : "None"}]: `)) || defaultKey;
if (!primaryApiKey) {
console.error("Primary API Key is required.");
process.exit(1);
}
// 2. Ask for Fallback Provider
const configureFallback = ((await askQuestion("\nDo you want to configure a Fallback LLM Model? (y/n) [default: n]: ")) || "n").toLowerCase() === "y";
let fallbackProviderChoice = "";
let fallbackModelName = "";
let fallbackApiKey = "";
if (configureFallback) {
console.log("\nSelect Fallback LLM Provider:");
console.log("1. Deepseek (Native Node)");
console.log("2. OpenAI (Standard)");
console.log("3. Google Gemini");
console.log("4. Anthropic");
fallbackProviderChoice = (await askQuestion("Enter choice (1-4) [default: 1]: ")) || "1";
let defaultFallbackModel = "deepseek-chat";
if (fallbackProviderChoice === "2") defaultFallbackModel = "gpt-4o-mini";
else if (fallbackProviderChoice === "3") defaultFallbackModel = "gemini-1.5-flash";
else if (fallbackProviderChoice === "4") defaultFallbackModel = "claude-3-5-sonnet-latest";
fallbackModelName = (await askQuestion(`Enter fallback model name [default: ${defaultFallbackModel}]: `)) || defaultFallbackModel;
let defaultFallbackKey = "";
if (fallbackProviderChoice === "1") defaultFallbackKey = process.env.DEEPSEEK_API_KEY || "";
else if (fallbackProviderChoice === "3") defaultFallbackKey = process.env.GEMINI_API_KEY || "";
fallbackApiKey = (await askQuestion(`Enter fallback API key [default: ${defaultFallbackKey ? "Loaded from .env" : "None"}]: `)) || defaultFallbackKey;
if (!fallbackApiKey) {
console.error("Fallback API Key is required.");
process.exit(1);
}
}
console.log("\nDeploying credentials to n8n...");
const supabaseCredId = await getOrCreateCredential("Semillero2_Supabase_V3", "supabaseApi", {
host: SUPABASE_URL,
serviceRole: SUPABASE_SECRET_KEY,
allowedHttpRequestDomains: "none",
});
const primaryConfig = getProviderConfig(primaryProviderChoice, primaryApiKey, primaryModelName);
const primaryCredId = await getOrCreateCredential(
`Semillero2_Primary_${primaryConfig.credentialType}`,
primaryConfig.credentialType,
primaryConfig.credentialData
);
let fallbackConfig: ProviderConfig | null = null;
let fallbackCredId = "";
if (configureFallback) {
fallbackConfig = getProviderConfig(fallbackProviderChoice, fallbackApiKey, fallbackModelName);
fallbackCredId = await getOrCreateCredential(
`Semillero2_Fallback_${fallbackConfig.credentialType}`,
fallbackConfig.credentialType,
fallbackConfig.credentialData
);
}
// 3. Define workflow nodes dynamically
console.log("\nBuilding workflow nodes...");
const webhookTriggerNode = {
parameters: {
httpMethod: "POST",
path: "evaluate-candidate",
responseMode: "responseNode",
options: {},
},
id: "webhook-trigger",
name: "Webhook Trigger",
type: "n8n-nodes-base.webhook",
typeVersion: 1.1,
position: [100, 300],
};
const primaryChainNode = {
parameters: {
promptType: "defineBelow", // Fixed casing: camelCase!
hasOutputParser: true, // Enforce "Require Specific Output Format"
text: "={{ $('Webhook Trigger').item.json.body.text }}",
systemMessage: "You are an AI recruitment assistant evaluating a candidate's CV for a job vacancy. Analyze the candidate's CV text. You MUST respond with a raw JSON object containing exactly these five keys:\n- summary: a brief candidate summary (max 3 sentences).\n- classification: 'Qualified', 'Unqualified', or 'Review'.\n- suggestions: an array of recommendations for next steps (e.g. ['Schedule interview', 'Reject', 'Verify references']).\n- riskLevel: 'Low', 'Medium', or 'High'.\n- ai_score: a number between 0 and 100 representing general suitability.\n\nDo not include markdown code blocks or any text outside the JSON.",
},
id: "llm-chain-primary",
name: "LLM Chain Evaluation (Primary)",
type: "@n8n/n8n-nodes-langchain.chainLlm",
typeVersion: 1.4,
position: [400, 300],
continueOnFail: configureFallback, // continue on fail only if fallback exists
};
const primaryModelNode = {
parameters: primaryConfig.nodeParameters,
id: "primary-model",
name: "Primary Chat Model",
type: primaryConfig.nodeType,
typeVersion: 1,
position: [350, 480],
credentials: {
[primaryConfig.credentialType]: {
id: primaryCredId,
name: `Semillero2_Primary_${primaryConfig.credentialType}`,
},
},
};
const jsonParserNode = {
parameters: {
jsonSchema: "{\n \"type\": \"object\",\n \"properties\": {\n \"summary\": {\n \"type\": \"string\"\n },\n \"classification\": {\n \"type\": \"string\",\n \"enum\": [\"Qualified\", \"Unqualified\", \"Review\"]\n },\n \"suggestions\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\"\n }\n },\n \"riskLevel\": {\n \"type\": \"string\",\n \"enum\": [\"Low\", \"Medium\", \"High\"]\n },\n \"ai_score\": {\n \"type\": \"number\"\n }\n },\n \"required\": [\"summary\", \"classification\", \"suggestions\", \"riskLevel\", \"ai_score\"]\n}",
},
id: "json-parser",
name: "Structured Output Parser",
type: "@n8n/n8n-nodes-langchain.outputParserStructured",
typeVersion: 1,
position: [480, 480],
};
// Replace Code node with a native Edit Fields (Set) node to avoid code blocks
const setNode = {
parameters: {
assignments: {
assignments: [
{
name: "candidate_id",
value: "={{ $('Webhook Trigger').item.json.body.candidateId }}",
type: "string",
},
{
name: "interview_id",
value: "={{ $('Webhook Trigger').item.json.body.interviewId }}",
type: "string",
},
{
name: "ai_score",
value: "={{ $json.ai_score }}",
type: "number",
},
{
name: "evaluation",
value: "={{ { summary: $json.summary, classification: $json.classification, suggestions: $json.suggestions, riskLevel: $json.riskLevel } }}",
type: "object",
},
],
},
options: {
includeOtherFields: false, // Drop other fields to cleanly match schema
},
},
id: "format-data",
name: "Format Evaluation Data",
type: "n8n-nodes-base.set",
typeVersion: 3,
position: [850, 300],
};
const supabaseInsertNode = {
parameters: {
operation: "insert",
table: "scores",
options: {},
},
id: "supabase-insert",
name: "Insert Score to Supabase",
type: "n8n-nodes-base.supabase",
typeVersion: 1,
position: [1050, 300],
credentials: {
supabaseApi: {
id: supabaseCredId,
name: "Semillero2_Supabase_V3",
},
},
};
const respondWebhookNode = {
parameters: {
options: {},
},
id: "respond-webhook",
name: "Respond to Webhook",
type: "n8n-nodes-base.respondToWebhook",
typeVersion: 1.1,
position: [1250, 300],
};
const wNodes: any[] = [
webhookTriggerNode,
primaryChainNode,
primaryModelNode,
jsonParserNode,
setNode,
supabaseInsertNode,
respondWebhookNode,
];
const wConnections: any = {
"Webhook Trigger": {
main: [
[
{
node: "LLM Chain Evaluation (Primary)",
type: "main",
index: 0,
},
],
],
},
"Primary Chat Model": {
ai_languageModel: [
[
{
node: "LLM Chain Evaluation (Primary)",
type: "ai_languageModel",
index: 0,
},
],
],
},
"Structured Output Parser": {
outputParser: [
[
{
node: "LLM Chain Evaluation (Primary)",
type: "outputParser",
index: 0,
},
],
],
},
"Format Evaluation Data": {
main: [
[
{
node: "Insert Score to Supabase",
type: "main",
index: 0,
},
],
],
},
"Insert Score to Supabase": {
main: [
[
{
node: "Respond to Webhook",
type: "main",
index: 0,
},
],
],
},
};
if (configureFallback && fallbackConfig) {
console.log("Configuring Fallback LLM Route...");
const checkErrorNode = {
parameters: {
conditions: {
options: {
caseSensitive: true,
leftValue: "",
typeValidation: "strict",
},
conditions: [
{
id: "err-cond",
leftValue: "={{ $json.hasOwnProperty('error') }}",
rightValue: "true",
operator: {
type: "boolean",
operation: "equals",
},
},
],
combinator: "and",
},
},
id: "check-error",
name: "Check Primary Error",
type: "n8n-nodes-base.if",
typeVersion: 2.2,
position: [600, 300],
};
const fallbackChainNode = {
parameters: {
promptType: "defineBelow", // Fixed casing: camelCase!
hasOutputParser: true, // Enforce "Require Specific Output Format"
text: "={{ $('Webhook Trigger').item.json.body.text }}",
systemMessage: "You are an AI recruitment assistant evaluating a candidate's CV for a job vacancy. Analyze the candidate's CV text. You MUST respond with a raw JSON object containing exactly these five keys:\n- summary: a brief candidate summary (max 3 sentences).\n- classification: 'Qualified', 'Unqualified', or 'Review'.\n- suggestions: an array of recommendations for next steps (e.g. ['Schedule interview', 'Reject', 'Verify references']).\n- riskLevel: 'Low', 'Medium', or 'High'.\n- ai_score: a number between 0 and 100 representing general suitability.\n\nDo not include markdown code blocks or any text outside the JSON.",
},
id: "llm-chain-fallback",
name: "LLM Chain Evaluation (Fallback)",
type: "@n8n/n8n-nodes-langchain.chainLlm",
typeVersion: 1.4,
position: [800, 450],
};
const fallbackModelNode = {
parameters: fallbackConfig.nodeParameters,
id: "fallback-model",
name: "Fallback Chat Model",
type: fallbackConfig.nodeType,
typeVersion: 1,
position: [750, 630],
credentials: {
[fallbackConfig.credentialType]: {
id: fallbackCredId,
name: `Semillero2_Fallback_${fallbackConfig.credentialType}`,
},
},
};
const jsonParserFallbackNode = {
parameters: {
jsonSchema: "{\n \"type\": \"object\",\n \"properties\": {\n \"summary\": {\n \"type\": \"string\"\n },\n \"classification\": {\n \"type\": \"string\",\n \"enum\": [\"Qualified\", \"Unqualified\", \"Review\"]\n },\n \"suggestions\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\"\n }\n },\n \"riskLevel\": {\n \"type\": \"string\",\n \"enum\": [\"Low\", \"Medium\", \"High\"]\n },\n \"ai_score\": {\n \"type\": \"number\"\n }\n },\n \"required\": [\"summary\", \"classification\", \"suggestions\", \"riskLevel\", \"ai_score\"]\n}",
},
id: "json-parser-fallback",
name: "Structured Output Parser (Fallback)",
type: "@n8n/n8n-nodes-langchain.outputParserStructured",
typeVersion: 1,
position: [880, 630],
};
// Reposition Set node for fallback routing
setNode.position = [1050, 450];
supabaseInsertNode.position = [1250, 450];
respondWebhookNode.position = [1450, 450];
wNodes.push(checkErrorNode, fallbackChainNode, fallbackModelNode, jsonParserFallbackNode);
// Primary chain routes to IF check
wConnections["LLM Chain Evaluation (Primary)"] = {
main: [
[
{
node: "Check Primary Error",
type: "main",
index: 0,
},
],
],
};
// IF Node routes
wConnections["Check Primary Error"] = {
main: [
[
{
node: "LLM Chain Evaluation (Fallback)",
type: "main",
index: 0,
},
], // true branch (Output 0 -> Error happened)
[
{
node: "Format Evaluation Data",
type: "main",
index: 0,
},
], // false branch (Output 1 -> Success)
],
};
// Connect fallback LLM components
wConnections["Fallback Chat Model"] = {
ai_languageModel: [
[
{
node: "LLM Chain Evaluation (Fallback)",
type: "ai_languageModel",
index: 0,
},
],
],
};
wConnections["Structured Output Parser (Fallback)"] = {
outputParser: [
[
{
node: "LLM Chain Evaluation (Fallback)",
type: "outputParser",
index: 0,
},
],
],
};
wConnections["LLM Chain Evaluation (Fallback)"] = {
main: [
[
{
node: "Format Evaluation Data",
type: "main",
index: 0,
},
],
],
};
} else {
// If no fallback, wire directly Primary Chain -> Set node
wConnections["LLM Chain Evaluation (Primary)"].main = [
[
{
node: "Format Evaluation Data",
type: "main",
index: 0,
},
],
];
}
// Define complete workflow object
const workflowDefinition = {
name: "Semillero2: End-to-End Candidate Evaluation",
settings: {},
nodes: wNodes,
connections: wConnections,
};
console.log("\nDeploying workflow to n8n...");
// Check if it already exists
const workflowsList = await n8nRequest("/api/v1/workflows");
const existingWf = workflowsList.data.find(
(w: any) => w.name === "Semillero2: End-to-End Candidate Evaluation"
);
let deployResult;
if (existingWf) {
console.log(`Updating existing workflow (ID: ${existingWf.id})...`);
deployResult = await n8nRequest(`/api/v1/workflows/${existingWf.id}`, "PUT", workflowDefinition);
} else {
deployResult = await n8nRequest("/api/v1/workflows", "POST", workflowDefinition);
}
// Activate the workflow
console.log(`Activating workflow (ID: ${deployResult.id})...`);
await n8nRequest(`/api/v1/workflows/${deployResult.id}/activate`, "POST");
const webhookUrl = `${N8N_HOST}/webhook/${deployResult.id}/webhook/evaluate-candidate`;
console.log("\n==============================================");
console.log("DEPLOYMENT COMPLETE");
console.log("==============================================");
console.log(`Workflow ID: ${deployResult.id}`);
console.log(`Webhook URL: ${webhookUrl}`);
console.log("==============================================");
// Update .env file automatically
const envFilePath = path.join(__dirname, "../.env");
if (fs.existsSync(envFilePath)) {
let envContent = fs.readFileSync(envFilePath, "utf8");
if (envContent.includes("NEXT_PUBLIC_N8N_WEBHOOK_URL=")) {
envContent = envContent.replace(
/NEXT_PUBLIC_N8N_WEBHOOK_URL=.*/,
`NEXT_PUBLIC_N8N_WEBHOOK_URL=${webhookUrl}`
);
} else {
envContent += `\nNEXT_PUBLIC_N8N_WEBHOOK_URL=${webhookUrl}\n`;
}
fs.writeFileSync(envFilePath, envContent, "utf8");
console.log("Updated NEXT_PUBLIC_N8N_WEBHOOK_URL in .env");
}
}
main().catch((err) => {
console.error("Deployment failed:", err);
process.exit(1);
});