Table of Contents
AI CMS / Framework / Model
Overview
Yes. In AI, there are equivalent concepts to CMS / Framework / Library similar to traditional programming, but they are divided into multiple packaging layers.
Goal of this page: Explain AI using the mindset of OpenCart / Laravel / SDK so developers can understand it immediately. What it is: In AI, there are also equivalent concepts to CMS / Framework / Library similar to traditional programming.
Key idea: AI is not a single block, but is divided into multiple packaging layers, from “ready to use” to “build from the core.”
1) Direct Mapping: Traditional Programming ↔ AI
| Traditional Programming | AI Equivalent | Meaning |
|---|---|---|
| CMS (OpenCart, WordPress) | AI CMS / Prebuilt AI App | Ready to use |
| Framework (Laravel, Django) | AI Framework / LLM Framework | Build AI app |
| Library / SDK | Model / Inference Library | Assemble components |
| Custom low-level code | Train model from scratch | Expensive & rare |
2) AI CMS Layer – Ready to Use (No-code / Low-code)
What it is: Prepackaged AI applications with built-in UI + workflow + model integration.
Equivalent: OpenCart / WordPress
Examples
Key Characteristics
Built-in UI
Workflow built using forms / nodes
Configured via YAML / JSON
Plug in an LLM and run
Use When
Internal chatbot
Knowledge assistant
Fast PoC
Team without ML skills
Limitations
Hard to deeply customize
Performance not optimized
Weak governance if not configured properly
👉 This is true AI CMS.
3) AI Framework Layer – Like Laravel / Django
What it is: Frameworks for building AI applications using code.
Equivalent: Laravel / Django
Popular Frameworks
What They Do
Prompt management
RAG (Retrieval Augmented Generation)
Tool calling
Memory & agent flow
Mapping
Laravel ≈ LangChain
Django ≈ LlamaIndex
Use When
Building a real AI product
Have a backend team
Need flow control & testing
4) Pretrained Model Layer – Pretrained Models
What it is: AI models trained on large datasets.
Equivalent: Database engine / Search engine
Examples
Characteristics
No UI
No workflow
Can be self-hosted
Can be fine-tuned
🚫 Not a CMS – just the “brain.”
5) AI Inference Engine Layer
What it is: Runtime for executing models efficiently.
Equivalent: JVM / PHP runtime / DB engine core
Examples
Use When
High throughput
Cost optimization
On-prem / air-gapped
6) Managed AI Platform (AI SaaS)
What it is: AI platforms fully operated by vendors.
Equivalent: Shopify / Salesforce
Examples
Pros / Cons
✅ Fast, no infrastructure management
❌ Higher cost, vendor lock-in
7) Overall Layer Diagram
AI CMS (No-code) │ Flowise / AnythingLLM │ AI Framework (Code) │ LangChain / LlamaIndex │ Pretrained Model │ LLaMA / Mistral │ Inference Engine │ vLLM / llama.cpp │ Infrastructure │ Cloud / On-prem GPU
8) What to Choose in Practice?
“Like OpenCart, download and run” → AnythingLLM / Flowise
“Like Laravel” → LangChain / LlamaIndex
“Enterprise, data control” → Open-source model + vLLM + RAG
AI Service Classification – Company & Use Case
Goal
This page classifies popular AI services today using a senior developer mindset (CMS / Framework / Runtime), including:
Company behind it
Real-world use case
Position in system architecture
Layer Overview
AI is not a single block, but consists of 6 layers, from “ready to use” to “infrastructure core.”
AI App / Tool ↓ AI Platform (API) ↓ AI Framework ↓ Pretrained Model ↓ Inference Engine ↓ Infrastructure (GPU)
1) AI App / AI Tool (Ready to Use)
Equivalent: CMS (WordPress / OpenCart)
| Service | Company | Main Use |
|---|---|---|
| ChatGPT | OpenAI | General assistant, Q&A, coding, spec writing |
| Claude Chat | Anthropic | Long document analysis, reasoning |
| GitHub Copilot | GitHub / Microsoft | Code completion, pair programming |
| AnythingLLM | Mintplex Labs | Internal document chatbot |
| Flowise | FlowiseAI (OSS) | Rapid AI workflow building |
| Open WebUI | Open-source | Chat UI for self-hosted models |
| Botpress | Botpress Inc. | Customer service chatbot |
| Rasa | Rasa Technologies | Conversation engine |
Use when: Fast PoC, internal use, small team
2) AI Platform / AI SaaS
Equivalent: Shopify / Firebase
| Platform | Company | Main Use |
|---|---|---|
| OpenAI API | OpenAI | LLM API: chat, embedding, tool calling |
| Azure OpenAI | Microsoft | OpenAI + enterprise security |
| AWS Bedrock | Amazon | Multi-model AI for enterprise |
| Google Vertex AI | End-to-end AI platform | |
| IBM watsonx | IBM | AI + data governance |
Use when: Calling AI via API, no GPU management
3) AI Framework (Build AI Logic)
Equivalent: Laravel / Django
| Framework | Company / Organization | Main Use |
|---|---|---|
| LangChain | LangChain Inc. | Orchestrate prompt, agent, tools |
| LlamaIndex | LlamaIndex Inc. | RAG (document → answer) |
| Haystack | deepset | Search + QA pipeline |
| Semantic Kernel | Microsoft | Enterprise AI orchestration |
| CrewAI | CrewAI Inc. | Multi-agent workflow |
| AutoGen | Microsoft Research | Agent collaboration |
Use when: Building real AI features, need testing & CI/CD
4) Pretrained Model (AI Brain)
Equivalent: Database / Search Engine
| Model | Company | Main Use |
|---|---|---|
| GPT-4 / GPT-5 | OpenAI | Reasoning, general-purpose |
| Claude 3 / 4 | Anthropic | Long context, safety |
| Gemini | Multimodal | |
| LLaMA 3 | Meta | Open-source, self-host |
| Mistral / Mixtral | Mistral AI | Lightweight, fast, low cost |
| Qwen | Alibaba | Multilingual |
| DeepSeek | DeepSeek AI | Reasoning, open |
Note: Models have no UI and no workflow.
5) Inference Engine (Runtime)
Equivalent: JVM / PHP-FPM
| Engine | Company / Organization | Main Use |
|---|---|---|
| vLLM | UC Berkeley | High-throughput LLM serving |
| llama.cpp | Open-source | Run LLM on CPU / edge |
| TensorRT-LLM | NVIDIA | GPU optimization, low latency |
| TGI | Hugging Face | Production LLM endpoint |
| ONNX Runtime | Microsoft | Cross-platform inference |
Use when: Self-host models, high traffic, cost optimization
6) Infrastructure (GPU / Cloud)
Equivalent: Server / Datacenter
| Infra | Company | Main Use |
|---|---|---|
| AWS GPU | Amazon | AI cloud |
| Azure GPU | Microsoft | AI cloud |
| GCP GPU | AI cloud | |
| NVIDIA A100 / H100 | NVIDIA | Large-scale training / inference |
| On-prem GPU | Enterprise | Full privacy |
===== Real Architecture Example =====
React / Vue ↓ Laravel / Go API ↓ LangChain / LlamaIndex ↓ OpenAI API or LLaMA ↓ (vLLM if self-host)
Important Notes
⚠️ AI is not deterministic like traditional code.
Always require:
Guardrails
Evaluation
Human review