Apex Backend

Express.js API server with AI-powered document processing and RAG (Retrieval-Augmented Generation) capabilities for the Apex project management platform.

Tech Stack

Technology	Purpose
Node.js	Runtime environment
Express.js	Web framework
TypeScript	Type safety
Prisma	ORM for MongoDB
MongoDB	Database (with Atlas Vector Search)
LangGraph	AI agent orchestration
Google GenAI	LLM and embeddings
Better Auth	Authentication
Multer	File uploads
Zod	Schema validation

Project Structure

backend/
├── src/
│   ├── agents/
│   │   ├── graphs/                    # LangGraph workflow definitions
│   │   │   ├── assistantGraph.ts      # Main query answering agent
│   │   │   ├── informationGraph.ts    # Document summarization agent
│   │   │   ├── vectorGraph.ts         # Vector embedding pipeline
│   │   │   ├── textInfoGraph.ts       # Text data processing
│   │   │   ├── textVectorGraph.ts     # Text vectorization
│   │   │   ├── subAssistantDatabaseGraph.ts  # Database query sub-agent
│   │   │   └── subAssistantRagGraph.ts       # RAG sub-agent
│   │   │
│   │   ├── nodes/
│   │   │   ├── informatinUpload/      # Document upload processing nodes
│   │   │   │   ├── extractText.ts     # Extract text from files
│   │   │   │   ├── chunkText.ts       # Split text into chunks
│   │   │   │   ├── embedChunks.ts     # Generate embeddings
│   │   │   │   ├── storeChunks.ts     # Store in vector DB
│   │   │   │   ├── summarizeFile.ts   # AI summarization
│   │   │   │   ├── summarizeTextData.ts
│   │   │   │   ├── saveFileData.ts    # Save to database
│   │   │   │   └── saveTextData.ts
│   │   │   │
│   │   │   └── InformationRetrieve.ts/  # Query processing nodes
│   │   │       ├── decideInformationSource.ts  # Route queries
│   │   │       ├── embedQuery.ts       # Embed user query
│   │   │       ├── vectorSearch.ts     # Semantic search
│   │   │       ├── cleanVectorSearchInfo.ts
│   │   │       ├── cleanDatabaseInfo.ts
│   │   │       └── responseSummary.ts  # Generate response
│   │   │
│   │   ├── tools/
│   │   │   └── databaseAccess.ts       # Database query tool
│   │   │
│   │   ├── functionDeclaration.ts/
│   │   │   └── assitant.function.ts    # Function declarations
│   │   │
│   │   └── zodSchema/
│   │       └── documentSchema.ts       # Zod schemas for AI output
│   │
│   ├── modules/                        # Feature modules
│   │   ├── projects/
│   │   │   ├── projects.controller.ts
│   │   │   ├── projects.routes.ts
│   │   │   ├── projects.schema.ts
│   │   │   └── projects.sevice.ts
│   │   │
│   │   ├── data/                       # Document upload module
│   │   │   ├── data.controller.ts
│   │   │   ├── data.routes.ts
│   │   │   ├── data.schema.ts
│   │   │   └── data.service.ts
│   │   │
│   │   ├── assistant/                  # AI chat module
│   │   │   ├── assistant.controller.ts
│   │   │   ├── assistant.routes.ts
│   │   │   ├── assistant.schema.ts
│   │   │   └── assistant.service.ts
│   │   │
│   │   ├── tasks/
│   │   │   ├── tasks.controller.ts
│   │   │   ├── tasks.routes.ts
│   │   │   ├── tasks.schema.ts
│   │   │   └── tasks.service.ts
│   │   │
│   │   ├── projectData/                # Requirements, conflicts, etc.
│   │   │   ├── projectData.controller.ts
│   │   │   ├── projectData.routes.ts
│   │   │   ├── projectData.schema.ts
│   │   │   └── projectData.service.ts
│   │   │
│   │   └── progress/
│   │       ├── progress.controller.ts
│   │       ├── progress.route.ts
│   │       └── progress.service.ts
│   │
│   ├── config/
│   │   ├── index.ts                    # Config exports
│   │   ├── database.ts                 # Prisma client
│   │   ├── genAI.ts                    # Google GenAI setup
│   │   ├── mongoDB.ts                  # MongoDB native client
│   │   └── multer.ts                   # File upload config
│   │
│   ├── middlewares/
│   │   ├── authMiddleware.ts           # Auth verification
│   │   ├── errorHandler.ts             # Global error handler
│   │   ├── validate.ts                 # Request validation
│   │   ├── notFound.ts                 # 404 handler
│   │   └── middleware.ts               # Middleware exports
│   │
│   ├── utils/
│   │   ├── sse.ts                      # Server-Sent Events helper
│   │   ├── chunk.text.ts               # Text chunking utility
│   │   ├── embedChunk.ts               # Embedding utility
│   │   └── extractTextFromFile.ts      # File text extraction
│   │
│   ├── types/
│   │   └── express.d.ts                # Express type extensions
│   │
│   ├── routes/
│   │   └── index.ts                    # Route aggregation
│   │
│   ├── auth.ts                         # Better Auth configuration
│   ├── env.ts                          # Environment validation
│   ├── routes.ts                       # Main router
│   └── server.ts                       # Express app entry
│
├── prisma/
│   └── schema.prisma                   # Database schema
│
├── uploads/                            # Uploaded files (temporary)
├── tsconfig.json                       # TypeScript config
└── package.json                        # Dependencies & scripts

Getting Started

Prerequisites

Node.js 18+
MongoDB (local or MongoDB Atlas with Vector Search)
Google AI API Key

Installation

# Install dependencies
npm install

# Create environment file
cp .env.example .env

Environment Variables

Create a .env file in the backend directory:

# Database
DATABASE_URL="mongodb+srv://username:password@cluster.mongodb.net/apex"

# Server
PORT=5000
BASE_URL="http://localhost:5000"
NODE_ENV="development"
CORS_ORIGIN="http://localhost:5173"

# Authentication
BETTER_AUTH_SECRET="your-secret-key-minimum-32-characters-long"

# AI
GOOGLE_API_KEY="your-google-ai-api-key"

Database Setup

# Generate Prisma client
npm run prisma:generate

# Push schema to database
npm run prisma:push

# Open Prisma Studio (optional)
npm run prisma:studio

MongoDB Atlas Vector Search Index

For semantic search to work, create a vector search index in MongoDB Atlas:

{
  "fields": [
    {
      "type": "vector",
      "path": "vector",
      "numDimensions": 768,
      "similarity": "cosine"
    },
    {
      "type": "filter",
      "path": "projectId"
    }
  ]
}

Running the Server

# Development (with hot reload)
npm run dev

# Production build
npm run build
npm start

The API server will start at http://localhost:5000

API Endpoints

Authentication

Method	Endpoint	Description
`ALL`	`/api/auth/*`	Better Auth handlers (sign up, sign in, sign out, session)

Projects

Method	Endpoint	Description
`GET`	`/api/projects`	List all user projects
`GET`	`/api/projects/:id`	Get project by ID
`POST`	`/api/projects`	Create new project
`PUT`	`/api/projects/:id`	Update project
`DELETE`	`/api/projects/:id`	Delete project

Documents

Method	Endpoint	Description
`GET`	`/api/project/:projectId/documents`	List project documents
`POST`	`/api/project/:projectId/upload`	Upload document (multipart/form-data)
`POST`	`/api/project/:projectId/text`	Add text data
`DELETE`	`/api/project/:projectId/documents/:docId`	Delete document

Project Data

Method	Endpoint	Description
`GET`	`/api/project/:projectId/requirements`	Get all requirements
`GET`	`/api/project/:projectId/tech-stacks`	Get tech stacks
`GET`	`/api/project/:projectId/conflicts`	Get conflicts
`GET`	`/api/project/:projectId/missing-info`	Get missing information
`PUT`	`/api/project/:projectId/conflicts/:id`	Resolve conflict
`PUT`	`/api/project/:projectId/missing-info/:id`	Address missing info

Tasks

Method	Endpoint	Description
`GET`	`/api/tasks/:projectId`	List project tasks
`POST`	`/api/tasks/:projectId`	Create task
`PUT`	`/api/tasks/:taskId`	Update task
`DELETE`	`/api/tasks/:taskId`	Delete task

AI Assistant

Method	Endpoint	Description
`POST`	`/api/assistant/:projectId/chat`	Send message (SSE streaming response)
`GET`	`/api/assistant/:projectId/history`	Get chat history
`DELETE`	`/api/assistant/:projectId/history`	Clear chat history

Progress

Method	Endpoint	Description
`GET`	`/api/project/:projectId/progress`	Get project progress

Database Schema

Core Models

model Project {
  id          String        @id
  title       String
  description String?
  status      ProjectStatus @default(ACTIVE)  // ACTIVE | COMPLETED
  progress    Float         @default(0)
  userId      String
  
  documents                 Document[]
  tasks                     Task[]
  chunks                    Chunk[]
  functionalRequirements    FunctionalRequirement[]
  nonFunctionalRequirements NonFunctionalRequirement[]
  techStacks                TechStack[]
  conflicts                 Conflict[]
  missingInformation        MissingInformation[]
}

model Document {
  id       String         @id
  fileName String
  fileType String
  status   DocumentStatus  // PENDING | PROCESSED | FAILED
  projectId String
  chunks   Chunk[]
}

model Chunk {
  id         String   @id
  content    String
  vector     Float[]   // Embedding vector for semantic search
  source     String?
  projectId  String
  documentId String?
}

model Task {
  id          String     @id
  title       String
  description String?
  status      TaskStatus  // TODO | IN_PROGRESS | DONE
  origin      TaskOrigin  // USER | AGENT
  weight      Int
  projectId   String
}

Requirements & Analysis

model FunctionalRequirement {
  id          String @id
  code        String   // e.g., "FR-1"
  description String
  projectId   String
}

model NonFunctionalRequirement {
  id          String @id
  code        String   // e.g., "NFR-1"
  description String
  projectId   String
}

model TechStack {
  id       String   @id
  category String   // "frontend" | "backend" | "database" | etc.
  stack    String[] // ["React", "TypeScript", "Tailwind"]
  projectId String
}

model Conflict {
  id          String           @id
  description String
  severity    ConflictSeverity  // LOW | MEDIUM | HIGH
  resolved    Boolean
  resolution  String?
  projectId   String
}

model MissingInformation {
  id          String @id
  description String
  addressed   Boolean
  note        String?
  projectId   String
}

AI Agent Workflows

Document Processing Pipeline

When a document is uploaded, it goes through:

┌─────────────────┐
│  Upload File    │
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│  Extract Text   │  (PDF.js / Mammoth)
└────────┬────────┘
         │
         ├──────────────────────┐
         │                      │
         ▼                      ▼
┌─────────────────┐    ┌─────────────────┐
│   Summarize     │    │   Chunk Text    │
│  & Extract      │    │                 │
│  Requirements   │    └────────┬────────┘
└────────┬────────┘             │
         │                      ▼
         │             ┌─────────────────┐
         │             │ Embed Chunks    │  (Google GenAI)
         │             └────────┬────────┘
         │                      │
         ▼                      ▼
┌─────────────────┐    ┌─────────────────┐
│  Save to DB     │    │  Store Vectors  │
│ (Requirements,  │    │  (MongoDB)      │
│  TechStack,     │    └─────────────────┘
│  Conflicts)     │
└─────────────────┘

Query Processing Pipeline

When a user asks a question:

┌─────────────────┐
│  User Query     │
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│ Decide Source   │  (LLM decides: DB, RAG, or both)
└────────┬────────┘
         │
    ┌────┴────┐
    │         │
    ▼         ▼
┌───────┐  ┌───────┐
│  DB   │  │  RAG  │
│ Query │  │Search │
└───┬───┘  └───┬───┘
    │          │
    └────┬─────┘
         │
         ▼
┌─────────────────┐
│   Summarize     │  (Generate answer with sources)
│   Response      │
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│  SSE Stream     │  (Real-time to client)
└─────────────────┘

Real-time Streaming

The chat endpoint uses Server-Sent Events (SSE) for real-time responses:

// SSE event types
interface SSEEvent {
  type: 'progress' | 'chunk' | 'done' | 'error'
  data: string
}

// Progress events show what the agent is doing
{ type: 'progress', data: 'Searching documents...' }

// Chunk events stream the response
{ type: 'chunk', data: 'Based on your requirements...' }

// Done event signals completion
{ type: 'done', data: JSON.stringify({ sources: [...] }) }

Authentication

Better Auth handles all authentication with:

Email/Password authentication
Session-based auth with secure cookies
Cookie caching for performance (5 min TTL)
7-day session expiration with daily refresh

// Protecting routes
import { authMiddleware } from './middlewares/authMiddleware'

router.use(authMiddleware)  // All routes after this are protected

File Uploads

Multer handles file uploads with:

Max file size: 10MB
Supported formats: PDF, DOCX, TXT
Files stored temporarily in /uploads

Scripts

Script	Description
`npm run dev`	Start with hot reload (tsx watch)
`npm run build`	Compile TypeScript
`npm start`	Run compiled JS
`npm run lint`	Run ESLint
`npm run prisma:generate`	Generate Prisma client
`npm run prisma:push`	Push schema to database
`npm run prisma:studio`	Open Prisma Studio

Error Handling

Centralized error handling with proper HTTP status codes:

// Custom error format
{
  "success": false,
  "error": "Error message",
  "code": "ERROR_CODE"  // optional
}

License

ISC License

Navigation Menu

Uh oh!

Apex Backend

Tech Stack

Project Structure

Getting Started

Prerequisites

Installation

Environment Variables

Database Setup

MongoDB Atlas Vector Search Index

Running the Server

API Endpoints

Authentication

Projects

Documents

Project Data

Tasks

AI Assistant

Progress

Database Schema

Core Models

Requirements & Analysis

AI Agent Workflows

Document Processing Pipeline

Query Processing Pipeline

Real-time Streaming

Authentication

File Uploads

Scripts

Error Handling

License

Uh oh!

FilesExpand file tree

README.md

Latest commit

History

README.md

File metadata and controls

Apex Backend

Tech Stack

Project Structure

Getting Started

Prerequisites

Installation

Environment Variables

Database Setup

MongoDB Atlas Vector Search Index

Running the Server

API Endpoints

Authentication

Projects

Documents

Project Data

Tasks

AI Assistant

Progress

Database Schema

Core Models

Requirements & Analysis

AI Agent Workflows

Document Processing Pipeline

Query Processing Pipeline

Real-time Streaming

Authentication

File Uploads

Scripts

Error Handling

License