feat: complete tRPC integration and fix platform UI issues

- Implement comprehensive tRPC setup with type-safe API
- Create tRPC routers for dashboard, admin, and auth endpoints
- Migrate frontend components to use tRPC client
- Fix platform dashboard Settings button functionality
- Add platform settings page with profile and security management
- Create OpenAI API mocking infrastructure for cost-safe testing
- Update tests to work with new tRPC architecture
- Sync database schema to fix AIBatchRequest table errors
This commit is contained in:
2025-07-11 15:37:53 +02:00
committed by Kaj Kowalski
parent f2a3d87636
commit fa7e815a3b
38 changed files with 4285 additions and 518 deletions

View File

@ -0,0 +1,416 @@
/**
* OpenAI API Mock Server
*
* Provides a drop-in replacement for OpenAI API calls during development
* and testing to prevent unexpected costs and enable offline development.
*/
import {
calculateMockCost,
generateBatchResponse,
generateSessionAnalysisResponse,
MOCK_RESPONSE_GENERATORS,
type MockBatchResponse,
type MockChatCompletion,
type MockResponseType,
} from "./openai-responses";
interface MockOpenAIConfig {
enabled: boolean;
baseDelay: number; // Base delay in ms to simulate API latency
randomDelay: number; // Additional random delay (0 to this value)
errorRate: number; // Probability of simulated errors (0.0 to 1.0)
logRequests: boolean; // Whether to log mock requests
}
class OpenAIMockServer {
private config: MockOpenAIConfig;
private totalCost = 0;
private requestCount = 0;
private activeBatches: Map<string, MockBatchResponse> = new Map();
constructor(config: Partial<MockOpenAIConfig> = {}) {
this.config = {
enabled: process.env.OPENAI_MOCK_MODE === "true",
baseDelay: 500, // 500ms base delay
randomDelay: 1000, // 0-1000ms additional delay
errorRate: 0.02, // 2% error rate
logRequests: process.env.NODE_ENV === "development",
...config,
};
}
/**
* Check if mock mode is enabled
*/
isEnabled(): boolean {
return this.config.enabled;
}
/**
* Simulate network delay
*/
private async simulateDelay(): Promise<void> {
const delay =
this.config.baseDelay + Math.random() * this.config.randomDelay;
await new Promise((resolve) => setTimeout(resolve, delay));
}
/**
* Simulate random API errors
*/
private shouldSimulateError(): boolean {
return Math.random() < this.config.errorRate;
}
/**
* Log mock requests for debugging
*/
private logRequest(endpoint: string, data: any): void {
if (this.config.logRequests) {
console.log(`[OpenAI Mock] ${endpoint}:`, JSON.stringify(data, null, 2));
}
}
/**
* Check if this is a session analysis request (comprehensive JSON format)
*/
private isSessionAnalysisRequest(prompt: string): boolean {
const promptLower = prompt.toLowerCase();
return (
promptLower.includes("session_id") &&
(promptLower.includes("sentiment") ||
promptLower.includes("category") ||
promptLower.includes("language"))
);
}
/**
* Extract processing type from prompt
*/
private extractProcessingType(prompt: string): MockResponseType {
const promptLower = prompt.toLowerCase();
if (
promptLower.includes("sentiment") ||
promptLower.includes("positive") ||
promptLower.includes("negative")
) {
return "sentiment";
}
if (promptLower.includes("category") || promptLower.includes("classify")) {
return "category";
}
if (promptLower.includes("summary") || promptLower.includes("summarize")) {
return "summary";
}
if (promptLower.includes("question") || promptLower.includes("extract")) {
return "questions";
}
// Default to sentiment analysis
return "sentiment";
}
/**
* Mock chat completions endpoint
*/
async mockChatCompletion(request: {
model: string;
messages: Array<{ role: string; content: string }>;
temperature?: number;
max_tokens?: number;
}): Promise<MockChatCompletion> {
this.requestCount++;
await this.simulateDelay();
if (this.shouldSimulateError()) {
throw new Error("Mock OpenAI API error: Rate limit exceeded");
}
this.logRequest("/v1/chat/completions", request);
// Extract the user content to analyze
const userMessage =
request.messages.find((msg) => msg.role === "user")?.content || "";
const systemMessage =
request.messages.find((msg) => msg.role === "system")?.content || "";
let response: MockChatCompletion;
let processingType: string;
// Check if this is a comprehensive session analysis request
if (this.isSessionAnalysisRequest(systemMessage)) {
// Extract session ID from system message for session analysis
const sessionIdMatch = systemMessage.match(/"session_id":\s*"([^"]+)"/);
const sessionId = sessionIdMatch?.[1] || `mock-session-${Date.now()}`;
response = generateSessionAnalysisResponse(userMessage, sessionId);
processingType = "session_analysis";
} else {
// Use simple response generators for other types
const detectedType = this.extractProcessingType(
systemMessage + " " + userMessage
);
response = MOCK_RESPONSE_GENERATORS[detectedType](userMessage);
processingType = detectedType;
}
// Track costs
const cost = calculateMockCost(response.usage);
this.totalCost += cost;
if (this.config.logRequests) {
console.log(
`[OpenAI Mock] Generated ${processingType} response. Cost: $${cost.toFixed(6)}, Total: $${this.totalCost.toFixed(6)}`
);
}
return response;
}
/**
* Mock batch creation endpoint
*/
async mockCreateBatch(request: {
input_file_id: string;
endpoint: string;
completion_window: string;
metadata?: Record<string, string>;
}): Promise<MockBatchResponse> {
await this.simulateDelay();
if (this.shouldSimulateError()) {
throw new Error("Mock OpenAI API error: Invalid file format");
}
this.logRequest("/v1/batches", request);
const batch = generateBatchResponse("validating");
this.activeBatches.set(batch.id, batch);
// Simulate batch processing progression
this.simulateBatchProgression(batch.id);
return batch;
}
/**
* Mock batch retrieval endpoint
*/
async mockGetBatch(batchId: string): Promise<MockBatchResponse> {
await this.simulateDelay();
const batch = this.activeBatches.get(batchId);
if (!batch) {
throw new Error(`Mock OpenAI API error: Batch ${batchId} not found`);
}
this.logRequest(`/v1/batches/${batchId}`, { batchId });
return batch;
}
/**
* Mock file upload endpoint
*/
async mockUploadFile(request: {
file: string; // File content
purpose: string;
}): Promise<{
id: string;
object: string;
purpose: string;
filename: string;
}> {
await this.simulateDelay();
if (this.shouldSimulateError()) {
throw new Error("Mock OpenAI API error: File too large");
}
const fileId = `file-mock-${Date.now()}-${Math.random().toString(36).substr(2, 9)}`;
this.logRequest("/v1/files", {
purpose: request.purpose,
size: request.file.length,
});
return {
id: fileId,
object: "file",
purpose: request.purpose,
filename: "batch_input.jsonl",
};
}
/**
* Mock file content retrieval
*/
async mockGetFileContent(fileId: string): Promise<string> {
await this.simulateDelay();
// Find the batch that owns this file
const batch = Array.from(this.activeBatches.values()).find(
(b) => b.output_file_id === fileId
);
if (!batch) {
throw new Error(`Mock OpenAI API error: File ${fileId} not found`);
}
// Generate mock batch results
const results: any = [];
for (let i = 0; i < batch.request_counts.total; i++) {
const response = MOCK_RESPONSE_GENERATORS.sentiment(`Sample text ${i}`);
results.push({
id: `batch-req-${i}`,
custom_id: `req-${i}`,
response: {
status_code: 200,
request_id: `req-${Date.now()}-${i}`,
body: response,
},
});
}
return results.map((r) => JSON.stringify(r)).join("\n");
}
/**
* Simulate batch processing progression over time
*/
private simulateBatchProgression(batchId: string): void {
const batch = this.activeBatches.get(batchId);
if (!batch) return;
// Validating -> In Progress (after 30 seconds)
setTimeout(() => {
const currentBatch = this.activeBatches.get(batchId);
if (currentBatch && currentBatch.status === "validating") {
currentBatch.status = "in_progress";
currentBatch.in_progress_at = Math.floor(Date.now() / 1000);
this.activeBatches.set(batchId, currentBatch);
}
}, 30000);
// In Progress -> Finalizing (after 2 minutes)
setTimeout(() => {
const currentBatch = this.activeBatches.get(batchId);
if (currentBatch && currentBatch.status === "in_progress") {
currentBatch.status = "finalizing";
currentBatch.finalizing_at = Math.floor(Date.now() / 1000);
this.activeBatches.set(batchId, currentBatch);
}
}, 120000);
// Finalizing -> Completed (after 3 minutes)
setTimeout(() => {
const currentBatch = this.activeBatches.get(batchId);
if (currentBatch && currentBatch.status === "finalizing") {
currentBatch.status = "completed";
currentBatch.completed_at = Math.floor(Date.now() / 1000);
currentBatch.output_file_id = `file-mock-output-${batchId}`;
currentBatch.request_counts.completed =
currentBatch.request_counts.total;
this.activeBatches.set(batchId, currentBatch);
}
}, 180000);
}
/**
* Get mock statistics
*/
getStats(): {
totalCost: number;
requestCount: number;
activeBatches: number;
isEnabled: boolean;
} {
return {
totalCost: this.totalCost,
requestCount: this.requestCount,
activeBatches: this.activeBatches.size,
isEnabled: this.config.enabled,
};
}
/**
* Reset statistics (useful for tests)
*/
resetStats(): void {
this.totalCost = 0;
this.requestCount = 0;
this.activeBatches.clear();
}
/**
* Update configuration
*/
updateConfig(newConfig: Partial<MockOpenAIConfig>): void {
this.config = { ...this.config, ...newConfig };
}
}
// Global instance
export const openAIMock = new OpenAIMockServer();
/**
* Drop-in replacement for OpenAI client that uses mocks when enabled
*/
export class MockOpenAIClient {
private realClient: any;
constructor(realClient: any) {
this.realClient = realClient;
}
get chat() {
return {
completions: {
create: async (params: any) => {
if (openAIMock.isEnabled()) {
return openAIMock.mockChatCompletion(params);
}
return this.realClient.chat.completions.create(params);
},
},
};
}
get batches() {
return {
create: async (params: any) => {
if (openAIMock.isEnabled()) {
return openAIMock.mockCreateBatch(params);
}
return this.realClient.batches.create(params);
},
retrieve: async (batchId: string) => {
if (openAIMock.isEnabled()) {
return openAIMock.mockGetBatch(batchId);
}
return this.realClient.batches.retrieve(batchId);
},
};
}
get files() {
return {
create: async (params: any) => {
if (openAIMock.isEnabled()) {
return openAIMock.mockUploadFile(params);
}
return this.realClient.files.create(params);
},
content: async (fileId: string) => {
if (openAIMock.isEnabled()) {
return openAIMock.mockGetFileContent(fileId);
}
return this.realClient.files.content(fileId);
},
};
}
}
export default openAIMock;

View File

@ -0,0 +1,583 @@
/**
* OpenAI API Mock Response Templates
*
* Provides realistic response templates for cost-safe testing
* and development without actual API calls.
*/
export interface MockChatCompletion {
id: string;
object: "chat.completion";
created: number;
model: string;
choices: Array<{
index: number;
message: {
role: "assistant";
content: string;
};
finish_reason: "stop" | "length" | "content_filter";
}>;
usage: {
prompt_tokens: number;
completion_tokens: number;
total_tokens: number;
};
}
export interface MockBatchResponse {
id: string;
object: "batch";
endpoint: string;
errors: {
object: "list";
data: Array<{
code: string;
message: string;
param?: string;
type: string;
}>;
};
input_file_id: string;
completion_window: string;
status:
| "validating"
| "in_progress"
| "finalizing"
| "completed"
| "failed"
| "expired"
| "cancelling"
| "cancelled";
output_file_id?: string;
error_file_id?: string;
created_at: number;
in_progress_at?: number;
expires_at?: number;
finalizing_at?: number;
completed_at?: number;
failed_at?: number;
expired_at?: number;
cancelling_at?: number;
cancelled_at?: number;
request_counts: {
total: number;
completed: number;
failed: number;
};
metadata?: Record<string, string>;
}
/**
* Generate realistic session analysis response matching the expected JSON schema
*/
export function generateSessionAnalysisResponse(
text: string,
sessionId: string
): MockChatCompletion {
// Extract session ID from the text if provided in system prompt
const sessionIdMatch = text.match(/"session_id":\s*"([^"]+)"/);
const extractedSessionId = sessionIdMatch?.[1] || sessionId;
// Simple sentiment analysis logic
const positiveWords = [
"good",
"great",
"excellent",
"happy",
"satisfied",
"wonderful",
"amazing",
"pleased",
"thanks",
];
const negativeWords = [
"bad",
"terrible",
"awful",
"unhappy",
"disappointed",
"frustrated",
"angry",
"upset",
"problem",
];
const words = text.toLowerCase().split(/\s+/);
const positiveCount = words.filter((word) =>
positiveWords.some((pos) => word.includes(pos))
).length;
const negativeCount = words.filter((word) =>
negativeWords.some((neg) => word.includes(neg))
).length;
let sentiment: "POSITIVE" | "NEUTRAL" | "NEGATIVE";
if (positiveCount > negativeCount) {
sentiment = "POSITIVE";
} else if (negativeCount > positiveCount) {
sentiment = "NEGATIVE";
} else {
sentiment = "NEUTRAL";
}
// Simple category classification
const categories: Record<string, string[]> = {
SCHEDULE_HOURS: ["schedule", "hours", "time", "shift", "working", "clock"],
LEAVE_VACATION: [
"vacation",
"leave",
"time off",
"holiday",
"pto",
"days off",
],
SICK_LEAVE_RECOVERY: [
"sick",
"ill",
"medical",
"health",
"doctor",
"recovery",
],
SALARY_COMPENSATION: [
"salary",
"pay",
"compensation",
"money",
"wage",
"payment",
],
CONTRACT_HOURS: ["contract", "agreement", "terms", "conditions"],
ONBOARDING: [
"onboard",
"new",
"start",
"first day",
"welcome",
"orientation",
],
OFFBOARDING: ["leaving", "quit", "resign", "last day", "exit", "farewell"],
WORKWEAR_STAFF_PASS: [
"uniform",
"clothing",
"badge",
"pass",
"equipment",
"workwear",
],
TEAM_CONTACTS: ["contact", "phone", "email", "reach", "team", "colleague"],
PERSONAL_QUESTIONS: ["personal", "family", "life", "private"],
ACCESS_LOGIN: [
"login",
"password",
"access",
"account",
"system",
"username",
],
SOCIAL_QUESTIONS: ["social", "chat", "friendly", "casual", "weather"],
};
const textLower = text.toLowerCase();
let bestCategory: keyof typeof categories | "UNRECOGNIZED_OTHER" =
"UNRECOGNIZED_OTHER";
let maxMatches = 0;
for (const [category, keywords] of Object.entries(categories)) {
const matches = keywords.filter((keyword) =>
textLower.includes(keyword)
).length;
if (matches > maxMatches) {
maxMatches = matches;
bestCategory = category as keyof typeof categories;
}
}
// Extract questions (sentences ending with ?)
const questions = text
.split(/[.!]+/)
.map((s) => s.trim())
.filter((s) => s.endsWith("?"))
.slice(0, 5);
// Generate summary (first sentence or truncated text)
const sentences = text.split(/[.!?]+/).filter((s) => s.trim().length > 0);
let summary = sentences[0]?.trim() || text.substring(0, 100);
if (summary.length > 150) {
summary = summary.substring(0, 147) + "...";
}
if (summary.length < 10) {
summary = "User inquiry regarding company policies";
}
// Detect language (simple heuristic)
const dutchWords = [
"het",
"de",
"een",
"en",
"van",
"is",
"dat",
"te",
"met",
"voor",
];
const germanWords = [
"der",
"die",
"das",
"und",
"ist",
"mit",
"zu",
"auf",
"für",
"von",
];
const dutchCount = dutchWords.filter((word) =>
textLower.includes(word)
).length;
const germanCount = germanWords.filter((word) =>
textLower.includes(word)
).length;
let language = "en"; // default to English
if (dutchCount > 0 && dutchCount >= germanCount) {
language = "nl";
} else if (germanCount > 0) {
language = "de";
}
// Check for escalation indicators
const escalated = /escalate|supervisor|manager|boss|higher up/i.test(text);
const forwardedHr = /hr|human resources|personnel|legal/i.test(text);
const analysisResult = {
language,
sentiment,
escalated,
forwarded_hr: forwardedHr,
category: bestCategory,
questions,
summary,
session_id: extractedSessionId,
};
const jsonContent = JSON.stringify(analysisResult);
const promptTokens = Math.ceil(text.length / 4);
const completionTokens = Math.ceil(jsonContent.length / 4);
return {
id: `chatcmpl-mock-${Date.now()}-${Math.random().toString(36).substr(2, 9)}`,
object: "chat.completion",
created: Math.floor(Date.now() / 1000),
model: "gpt-4o-mini",
choices: [
{
index: 0,
message: {
role: "assistant",
content: jsonContent,
},
finish_reason: "stop",
},
],
usage: {
prompt_tokens: promptTokens,
completion_tokens: completionTokens,
total_tokens: promptTokens + completionTokens,
},
};
}
/**
* Generate realistic category classification response
*/
export function generateCategoryResponse(text: string): MockChatCompletion {
// Simple category classification logic
const categories: Record<string, string[]> = {
SCHEDULE_HOURS: ["schedule", "hours", "time", "shift", "working"],
LEAVE_VACATION: ["vacation", "leave", "time off", "holiday", "pto"],
SICK_LEAVE_RECOVERY: ["sick", "ill", "medical", "health", "doctor"],
SALARY_COMPENSATION: ["salary", "pay", "compensation", "money", "wage"],
CONTRACT_HOURS: ["contract", "agreement", "terms", "conditions"],
ONBOARDING: ["onboard", "new", "start", "first day", "welcome"],
OFFBOARDING: ["leaving", "quit", "resign", "last day", "exit"],
WORKWEAR_STAFF_PASS: ["uniform", "clothing", "badge", "pass", "equipment"],
TEAM_CONTACTS: ["contact", "phone", "email", "reach", "team"],
PERSONAL_QUESTIONS: ["personal", "family", "life", "private"],
ACCESS_LOGIN: ["login", "password", "access", "account", "system"],
SOCIAL_QUESTIONS: ["social", "chat", "friendly", "casual"],
};
const textLower = text.toLowerCase();
let bestCategory = "UNRECOGNIZED_OTHER";
let maxMatches = 0;
for (const [category, keywords] of Object.entries(categories)) {
const matches = keywords.filter((keyword) =>
textLower.includes(keyword)
).length;
if (matches > maxMatches) {
maxMatches = matches;
bestCategory = category;
}
}
const promptTokens = Math.ceil(text.length / 4);
const completionTokens = bestCategory.length / 4;
return {
id: `chatcmpl-mock-${Date.now()}-${Math.random().toString(36).substr(2, 9)}`,
object: "chat.completion",
created: Math.floor(Date.now() / 1000),
model: "gpt-4o-mini",
choices: [
{
index: 0,
message: {
role: "assistant",
content: bestCategory,
},
finish_reason: "stop",
},
],
usage: {
prompt_tokens: promptTokens,
completion_tokens: completionTokens,
total_tokens: promptTokens + completionTokens,
},
};
}
/**
* Generate realistic summary response
*/
export function generateSummaryResponse(text: string): MockChatCompletion {
// Simple summarization - take first sentence or truncate
const sentences = text.split(/[.!?]+/).filter((s) => s.trim().length > 0);
let summary = sentences[0]?.trim() || text.substring(0, 100);
if (summary.length > 150) {
summary = summary.substring(0, 147) + "...";
}
const promptTokens = Math.ceil(text.length / 4);
const completionTokens = Math.ceil(summary.length / 4);
return {
id: `chatcmpl-mock-${Date.now()}-${Math.random().toString(36).substr(2, 9)}`,
object: "chat.completion",
created: Math.floor(Date.now() / 1000),
model: "gpt-4o-mini",
choices: [
{
index: 0,
message: {
role: "assistant",
content: summary,
},
finish_reason: "stop",
},
],
usage: {
prompt_tokens: promptTokens,
completion_tokens: completionTokens,
total_tokens: promptTokens + completionTokens,
},
};
}
/**
* Generate realistic sentiment analysis response
*/
export function generateSentimentResponse(text: string): MockChatCompletion {
// Simple sentiment analysis logic
const positiveWords = [
"good",
"great",
"excellent",
"happy",
"satisfied",
"wonderful",
"amazing",
"pleased",
"thanks",
];
const negativeWords = [
"bad",
"terrible",
"awful",
"unhappy",
"disappointed",
"frustrated",
"angry",
"upset",
"problem",
];
const words = text.toLowerCase().split(/\s+/);
const positiveCount = words.filter((word) =>
positiveWords.some((pos) => word.includes(pos))
).length;
const negativeCount = words.filter((word) =>
negativeWords.some((neg) => word.includes(neg))
).length;
let sentiment: "POSITIVE" | "NEUTRAL" | "NEGATIVE";
if (positiveCount > negativeCount) {
sentiment = "POSITIVE";
} else if (negativeCount > positiveCount) {
sentiment = "NEGATIVE";
} else {
sentiment = "NEUTRAL";
}
const promptTokens = Math.ceil(text.length / 4);
const completionTokens = Math.ceil(sentiment.length / 4);
return {
id: `chatcmpl-mock-${Date.now()}-${Math.random().toString(36).substr(2, 9)}`,
object: "chat.completion",
created: Math.floor(Date.now() / 1000),
model: "gpt-4o-mini",
choices: [
{
index: 0,
message: {
role: "assistant",
content: sentiment,
},
finish_reason: "stop",
},
],
usage: {
prompt_tokens: promptTokens,
completion_tokens: completionTokens,
total_tokens: promptTokens + completionTokens,
},
};
}
/**
* Generate realistic question extraction response
*/
export function generateQuestionExtractionResponse(
text: string
): MockChatCompletion {
// Extract sentences that end with question marks
const questions = text
.split(/[.!]+/)
.map((s) => s.trim())
.filter((s) => s.endsWith("?"))
.slice(0, 5); // Limit to 5 questions
const result =
questions.length > 0 ? questions.join("\n") : "No questions found.";
const promptTokens = Math.ceil(text.length / 4);
const completionTokens = Math.ceil(result.length / 4);
return {
id: `chatcmpl-mock-${Date.now()}-${Math.random().toString(36).substr(2, 9)}`,
object: "chat.completion",
created: Math.floor(Date.now() / 1000),
model: "gpt-4o-mini",
choices: [
{
index: 0,
message: {
role: "assistant",
content: result,
},
finish_reason: "stop",
},
],
usage: {
prompt_tokens: promptTokens,
completion_tokens: completionTokens,
total_tokens: promptTokens + completionTokens,
},
};
}
/**
* Generate mock batch job response
*/
export function generateBatchResponse(
status: MockBatchResponse["status"] = "in_progress"
): MockBatchResponse {
const now = Math.floor(Date.now() / 1000);
const batchId = `batch_mock_${Date.now()}_${Math.random().toString(36).substr(2, 9)}`;
const result: MockBatchResponse = {
id: batchId,
object: "batch",
endpoint: "/v1/chat/completions",
errors: {
object: "list",
data: [],
},
input_file_id: `file-mock-input-${batchId}`,
completion_window: "24h",
status,
created_at: now - 300, // 5 minutes ago
expires_at: now + 86400, // 24 hours from now
request_counts: {
total: 100,
completed:
status === "completed" ? 100 : status === "in_progress" ? 75 : 0,
failed: status === "failed" ? 25 : 0,
},
metadata: {
company_id: "test-company",
batch_type: "ai_processing",
},
};
// Set optional fields based on status
if (status === "completed") {
result.output_file_id = `file-mock-output-${batchId}`;
result.completed_at = now - 30;
}
if (status === "failed") {
result.failed_at = now - 30;
}
if (status !== "validating") {
result.in_progress_at = now - 240; // 4 minutes ago
}
if (status === "finalizing" || status === "completed") {
result.finalizing_at = now - 60;
}
return result;
}
/**
* Mock cost calculation for testing
*/
export function calculateMockCost(usage: {
prompt_tokens: number;
completion_tokens: number;
}): number {
// Mock pricing: $0.15 per 1K prompt tokens, $0.60 per 1K completion tokens (gpt-4o-mini rates)
const promptCost = (usage.prompt_tokens / 1000) * 0.15;
const completionCost = (usage.completion_tokens / 1000) * 0.6;
return promptCost + completionCost;
}
/**
* Response templates for different AI processing types
*/
export const MOCK_RESPONSE_GENERATORS = {
sentiment: generateSentimentResponse,
category: generateCategoryResponse,
summary: generateSummaryResponse,
questions: generateQuestionExtractionResponse,
} as const;
export type MockResponseType = keyof typeof MOCK_RESPONSE_GENERATORS;