AI Support Bot

A scalable AI Support Assistant built with Laravel 12 with 29 production-ready API endpoints.

🚀 Features

• Multi-Format Document Processing: PDF, DOCX, TXT with parallel batch processing
• Intelligent Vector Search: OpenAI embeddings with automatic storage optimization (JSON/pgvector/Qdrant)
• Conversation Memory: Context-aware follow-up questions with streaming responses (SSE)
• Quota Management: Track usage, messages, and token consumption
• Multi-Tenant Workspaces: Isolated environments with filtering
• Production Ready: Laravel Sanctum auth, Redis queues, rate limiting

🏗️ Tech Stack

• Framework: Laravel 12 (PHP 8.4)
• AI Model: GPT-4o-mini (chat completions)
• Embeddings: text-embedding-3-small (1536 dimensions)
• Database: MariaDB/MySQL (JSON vector storage) or PostgreSQL (pgvector support)
• Cache & Queue: Redis (Required for batch processing)
• Authentication: Laravel Sanctum with API tokens
• Vector Storage: Auto-detecting (JSON/pgvector/Qdrant)

Installation

1. Clone & Install

   git clone https://github.com/3idey/ai-support-bot.git
   cd ai-support-bot
   composer install

2. Environment Setup

   cp .env.example .env
   php artisan key:generate

   Configure your .env file:

   OPENAI_API_KEY=sk-...
   DB_CONNECTION=mariadb
   QUEUE_CONNECTION=redis
   CACHE_DRIVER=redis

3. Database & Migration

   touch database/database.sqlite # If using SQLite
   php artisan migrate

4. Start Services Ensure Redis is running:

   redis-server

   Start the queue worker (Critical for batch processing):

   php artisan queue:work

   Start the dev server:

   php artisan serve

API Documentation

Authentication

All API endpoints require a Bearer token:

Authorization: Bearer <your-token>

Generate a token: php artisan tinker → User::first()->createToken('dev')->plainTextToken

---

🎯 Quota Management (8 Endpoints)

Monitor and track user usage limits.

Check Full Quota Status

curl http://localhost:8000/api/quota \
  -H "Authorization: Bearer <token>"

Response:

{
  "message_quota": { "total": 50, "used": 15, "remaining": 35, "percentage": 30 },
  "token_quota": { "total": 100000, "used": 12450, "remaining": 87550, "percentage": 12.45 },
  "document_quota": { "total": 100, "used": 5, "remaining": 95, "percentage": 5 }
}

Other Quota Endpoints

• GET /api/quota/check - Simple boolean check
• GET /api/quota/remaining - Get remaining messages
• GET /api/quota/tokens - Get remaining tokens
• GET /api/quota/usage - Detailed usage breakdown
• POST /api/quota/track - Manually track token usage
• GET /api/quota/documents - Check document quota
• POST /api/quota/documents/validate - Validate document size

---

🗂️ Workspace Management (7 Endpoints)

Multi-tenant workspace isolation and filtering.

List Workspaces

curl http://localhost:8000/api/workspaces \
  -H "Authorization: Bearer <token>"

Create Workspace

curl -X POST http://localhost:8000/api/workspaces \
  -H "Authorization: Bearer <token>" \
  -H "Content-Type: application/json" \
  -d '{"name": "Customer Support", "description": "Support documentation"}'

Other Workspace Endpoints

• GET /api/workspaces/{id} - Show workspace details
• PUT /api/workspaces/{id} - Update workspace
• DELETE /api/workspaces/{id} - Delete workspace
• GET /api/workspaces/{id}/documents - List workspace documents
• GET /api/workspaces/{id}/conversations - List workspace conversations

---

💬 Conversation Management (4 Endpoints)

Manage chat history and conversations.

List Conversations

curl http://localhost:8000/api/conversations?workspace_id=1 \
  -H "Authorization: Bearer <token>"

Get Conversation Messages

curl http://localhost:8000/api/conversations/1/messages \
  -H "Authorization: Bearer <token>"

Other Conversation Endpoints

• GET /api/conversations/{id} - Show conversation details
• DELETE /api/conversations/{id} - Delete conversation

---

📄 Document Management (4 Endpoints)

Upload and manage knowledge base documents with filtering.

List Documents (with filtering)

curl http://localhost:8000/api/documents?workspace_id=1 \
  -H "Authorization: Bearer <token>"

Upload Document

curl -X POST http://localhost:8000/api/documents \
  -H "Authorization: Bearer <token>" \
  -F "[email protected]" \
  -F "workspace_id=1"

Other Document Endpoints

• GET /api/documents/{id} - Show document details
• DELETE /api/documents/{id} - Delete document

---

🤖 Chat Endpoints (1 Endpoint)

Ask Question (JSON Response)

curl -X POST http://localhost:8000/api/ask \
  -H "Authorization: Bearer <token>" \
  -H "Content-Type: application/json" \
  -d '{
    "question": "How do I reset my password?",
    "workspace_id": 1
  }'

Ask with Streaming + Memory

curl -N -X POST http://localhost:8000/api/ask \
  -H "Authorization: Bearer <token>" \
  -H "Content-Type: application/json" \
  -d '{
    "question": "Tell me more about that.",
    "conversation_id": 1,
    "stream": true
  }'

Stream Events:

• event: sources - Related document chunks
• event: conversation_id - ID for follow-ups
• data: {"content": "..."} - Answer tokens
• data: [DONE] - Stream complete

---

📊 Vector Database Management (4 Endpoints)

Intelligent vector storage with automatic optimization recommendations.

Check Vector Database Status

curl http://localhost:8000/api/vectors/status \
  -H "Authorization: Bearer <token>"

Response:

{
  "system": "mariadb_json",
  "status": "active",
  "total_embeddings": 150,
  "avg_dimensions": 1536,
  "database": "MariaDB 11.6.2",
  "storage_method": "JSON columns with PHP cosine similarity"
}

Get Optimization Recommendations

curl http://localhost:8000/api/vectors/recommendations \
  -H "Authorization: Bearer <token>"

Performance Benchmark

curl -X POST http://localhost:8000/api/vectors/benchmark \
  -H "Authorization: Bearer <token>" \
  -H "Content-Type: application/json" \
  -d '{"sample_size": 10}'

Migrate to pgvector (PostgreSQL only)

curl -X POST http://localhost:8000/api/vectors/migrate \
  -H "Authorization: Bearer <token>"

---

🏥 Health Check

curl http://localhost:8000/api/health

---

🚀 Performance & Scalability

Benchmarks

• Vector Search: ~30ms (JSON), ~5ms (pgvector), ~1ms (Qdrant)
• API Response: <100ms average
• Scaling: Stateless design with horizontal scaling support

---

📄 License

MIT License