At ToRun AI, we're building an AI Orchestration Platform that lets teams access 100+ AI models from 10+ providers through a single interface.
The Problem
Every major AI provider has its own API, auth, pricing, and rate limits. For teams using multiple providers, this means multiple integrations, separate billing, and no unified routing.
Our Solution: A 9-Step Execution Pipeline
Every AI request in ToRun goes through:
- Request Validation - Input sanitization
- Tenant Isolation - Multi-tenant security
- Subscription Verification - Plan checks
- Budget Checking - Dual-source billing
- Rule Engine - Pre-prompt rules, PII detection
- Model Routing - Dynamic access control
- Provider Execution - Adapter pattern
- Response Streaming - Real-time via SignalR
- Cost Calculation - Token-level tracking
Tech Stack
- .NET 10.0 on ABP.IO Framework
- MongoDB + Redis
- Angular frontend
- Domain-Driven Design (23 entities, 6 bounded contexts)
Why It Matters
No vendor lock-in. Switch providers seamlessly.
Dynamic access. Model access calculated in real-time, not hardcoded.
Smart budgets. Subscription allowance + wallet system.
We're building in public. Follow us for more technical deep-dives!
Top comments (0)