Skip to main content
Enclave
Local models

Connect self-hosted vLLM inference in Enclave

Enclave supports vLLM: point the model layer at your own vLLM inference service (OpenAI-compatible API) and characters reply using your private, high-throughput local deployment. Ideal for advanced users with GPU resources who want privacy and high concurrency.

Last reviewed:

The short answer

Point Enclave's model layer at your self-hosted vLLM (OpenAI-compatible endpoint) to drive characters with private, high-throughput inference.

Key points at a glance

  • Private inference: models run on your own GPU/server.
  • OpenAI-compatible API — configured much like connecting a cloud API.
  • Suits advanced deployments with high concurrency and latency/cost requirements.

How to connect

First spin up an OpenAI-compatible inference endpoint with vLLM, then in Enclave's model settings choose the OpenAI-compatible provider, point the base URL at your vLLM address, and enter the corresponding key (if any). Then set the model as the default or assign it per character.

Who it's for

Teams and advanced users with GPU resources who want to fully privatize inference, or need stable throughput for multi-user/high-concurrency scenarios. With Enclave self-hosting, models and data both live in your own infrastructure.

Key facts

Integration
OpenAI-compatible endpoint (base URL)
Billing
Self-hosted, no third-party API fees
Best for
High concurrency / private inference
Data flow
Stays within your own infrastructure

Ready to try it?

Open it in your browser — no credit card, no install.

Related

Connect other models

Connect self-hosted vLLM inference in Enclave · Enclave