Model Prism

v1.0

Open Model Prism is a multi-tenant OpenAI-compatible LLM gateway with intelligent routing, cost tracking, and admin UI. Connect multiple LLM providers, route requests automatically to the optimal model, and track costs per team.

Multi-Provider

Connect OpenAI, Anthropic, Ollama, Azure, Bedrock, OpenRouter, vLLM, and any OpenAI-compatible endpoint simultaneously.

Intelligent Routing

Send model: "auto" and let the routing engine classify requests into 45 categories and select the optimal model.

Cost Tracking

Per-request cost calculation, daily aggregates, baseline vs actual comparisons, and real-time dashboards showing savings from intelligent routing.

Multi-Tenant

Isolated gateway endpoints per team with their own API keys, rate limits, model access controls, and routing configuration.

Documentation

Getting Started

Installation, setup wizard, first tenant, and gateway usage.

Architecture

System overview, request flow, directory structure, and security.

Providers

Supported provider types, model discovery, and connection management.

Tenants

Tenant configuration, API keys, model access control, and aliases.

Routing

Intelligent model routing pipeline, cost tiers, signal extraction, and categories.

Analytics

Cost tracking, dashboards, request logs, and Prometheus metrics.

Operations

Deployment modes, scaling, load balancing, security, and monitoring.

API Reference

Gateway API, Admin API, Tenant Portal API, and global endpoints.