Skip to main content
AI Architecture Deep Dive

How Our AI Works

AI-First. Model-Agnostic. The intelligence is in our architecture.

We Are Not a Wrapper Around GPT

Most "AI platforms" add ChatGPT to existing software and call it AI-powered. That locks you to one vendor. Solid# is fundamentally different.

Others: API Wrappers

  • Locked to one AI provider
  • Every request = expensive LLM call
  • Generic responses, no business context
  • AI suggests, you execute

Solid#: ABOS Architecture

  • Use ANY LLM provider (or none)
  • 80% of tasks need no LLM call
  • 258 KB templates know your industry
  • AI executes, not just suggests

The Architecture That Makes This Real

66-Table Multi-Tenant Core

CRM, E-commerce, Payments, Inventory, Analytics - the foundation

14 Autonomous AI Agents

ADA - Orchestrator
Sage - Customer Chat
Marcus - Growth
Devon - DevOps/SEO
Jordan - Sales
Sarah - Support

Custom ML Celery Workers

No LLM Needed
Intent ClassificationSentiment AnalysisEntity ExtractionRouting DecisionsAnomaly DetectionFraud DetectionLead ScoringChurn Prediction

258 Knowledge Base Templates

No LLM Needed

Pattern matching + rules = instant answers in <100ms. Industry-specific from day one.

LLM Layer - YOUR CHOICE

The LLM is a plugin, not the product. 80% of tasks don’t even need an LLM call.

GPT-4
Claude
Gemini
Llama
Mistral
Local
Any API

Supported LLM Providers

Connect your preferred AI. Switch anytime. No vendor lock-in.

OpenAI

Production
GPT-4GPT-4oGPT-3.5 Turbo

Anthropic

Production
Claude 3 OpusClaude 3 SonnetClaude 3 Haiku

Google

Production
Gemini 2.0 ProGemini 2.0 Flash

Meta

Production
Llama 3Llama 3.1

Mistral

Production
Mistral LargeMistral Medium

Local/Self-hosted

Production
Ollamallama.cppvLLM

Any OpenAI-compatible

Production
Custom endpoints

Why Model-Agnostic Matters

No Vendor Lock-in

Switch from GPT to Claude to Gemini anytime. Your business data and workflows stay intact.

Cost Control

Route simple tasks to cheap models, complex tasks to powerful ones. Some tasks need no LLM at all.

Future-Proof

When GPT-5 or Claude 4 launches, upgrade instantly. No migration needed.

Privacy Options

Run local models (Ollama, llama.cpp) for sensitive data. Keep everything on-premise if needed.

Ready to See It in Action?

Try our MCP endpoint right now. No signup required.

curl -X POST https://solidnumber.com/api/mcp \ -H "Content-Type: application/json" \ -d '{"type": "list_tools"}'
How Our AI Works | Solid# - Model-Agnostic ABOS | SolidNumber