Dot Loom

Dot Loom is an open-source orchestration runtime that optimizes AI inference by adaptively routing requests through multiple models. Rather than committing to a fixed cost per request, it scales computational depth based on task difficulty using trained decision policies.

Highlights

• Adaptive execution policies that scale from 1-4 model calls (Lean, Balanced, Strict, or Fixed pipelines)
• Provider-agnostic architecture supporting OpenAI, Claude, Dot API, Ollama, and custom endpoints
• Built-in budget controls with hard ceilings on calls, credits, and latency
• Cost tracking and observability with detailed execution receipts and token counts
• Trained Ministral conductor model for intelligent routing decisions

Designed for developers and AI teams building cost-conscious inference systems, Dot Loom reduces unnecessary API calls while maintaining quality. The runtime is in alpha and works well for experiments, demos, and production architectures that need dynamic model selection. Anyone running multiple models or optimizing LLM costs benefits from its adaptive approach.