Sansa as a Small ML Gate

Sansa is essentially a lightweight machine learning model trained specifically on your data. It acts as an intelligent routing gate that makes real-time decisions about which AI model to use for each request.

Individual Training on Your Data

Unlike generic routing systems, Sansa is trained individually for each customer. It learns from your specific usage patterns, understanding which types of requests perform equally well on less expensive models versus when your premium baseline model is truly needed.

Learning Phase

Sansa analyzes your actual usage by shadow-testing requests on less expensive models against your baseline (premium) model, building a personalized routing model based on your unique data patterns.

Real-Time Routing Decisions

For each incoming request, Sansa's trained ML gate analyzes the prompt content in real-time, determines semantic complexity and requirements, and routes to the most cost-effective model that can deliver equivalent quality. Routing decisions happen in 50ms with negligible latency.

Confidence-Based Routing

Users set confidence thresholds for routing decisions, ensuring conservative routing where accuracy matters most.

Smart Model Routing

Intelligent routing that optimizes for cost, performance, and accuracy across multiple AI providers

Low Latency

Selects the most efficient model in 50ms with no meaningful performance impact on applications

Complete Control

Full control over routing decisions with configurable confidence thresholds and performance monitoring

No Evaluations Needed

Starts learning and routing immediately from actual usage patterns without evaluation setup

Developer Experience

Sansa is a separate SDK that replaces existing AI provider SDKs. Install the npm package, configure your API keys, and replace existing SDK calls with Sansa calls. No major infrastructure changes required.

Code Example

// Install and import Sansa
import { Sansa } from 'sansa';

// Initialize with your API keys
const sansa = new Sansa({
  apiKey: process.env.SANSA_API_KEY,
  providers: {
    anthropic: { apiKey: process.env.ANTHROPIC_API_KEY },
    openai: { apiKey: process.env.OPENAI_API_KEY },
  }
});

// Make your call - Sansa handles the routing
const result = await sansa.create({
  callName: 'customer-support',
  messages: [...],
  llmProviders: 'auto',
  llmBaseline: {
    provider: 'anthropic',
    model: 'claude-sonnet-4-5-20240929'
  }
});