Architecture & Protocol Engine

ExoRoute Architecture & Overview

ExoRoute is a local-first AI protocol gateway and model router engineered in Rust. It provides applications with a unified, authenticated endpoint while seamlessly translating between OpenAI Chat Completions, OpenAI Responses, and Anthropic Messages.

What ExoRoute Does

Single Local HTTP Gateway

Listens at 127.0.0.1:8686 by default, serving both high-throughput API endpoints and an embedded, reactive Svelte 5 dashboard.

Protocol Transpilation

Send OpenAI-formatted payloads to Anthropic Claude models (or vice-versa) with zero client-side code modifications.

Multi-Key Rotation & Failover

Configure multiple keys per provider with round-robin rotation and automatic failover upon receiving upstream rate limit errors.

Embedded LMDB Storage

Zero external database daemon dependencies. High-speed embedded ACID storage secured with ChaCha20-Poly1305 encryption.

Supported Protocols Matrix

Client applications connect seamlessly to standard endpoints, translated automatically to target providers:

Client Request Endpoint	Target Provider Format	SSE Streaming	Tools / Function Calling
POST /v1/chat/completions	OpenAI, Anthropic Claude, OpenRouter, Ollama	✓ Native SSE	✓ Auto-mapped
POST /v1/messages	Anthropic Claude, OpenAI GPT-4o, DeepSeek	✓ Native SSE	✓ Auto-mapped
POST /v1/responses	OpenAI Responses Format	✓ Native SSE	✓ Supported
GET /v1/models	Aggregated Catalog + Combos	—	—

Model Capability Boundary: ExoRoute normalizes protocols and routes requests, but it cannot synthesize capabilities not supported by the upstream model. For example, sending an image through the gateway does not make a text-only model vision-capable.

Architectural Metrics

+0.42 ms

P50 Added Latency

1.208 s

Time-to-Port Ready

15.0 MiB

Private Bytes RAM

Ready to Get Started?

Check the step-by-step installation instructions and initialize your gateway.

Read Quickstart