Skip to content

OpenAI-Compatible Backend

Point eyeroll at any OpenAI-compatible API endpoint. Use this for self-hosted models, custom proxies, or any provider that implements the OpenAI API format.

Setup

export EYEROLL_BACKEND=openai-compat

You must provide a base URL via the --base-url flag:

eyeroll watch video.mp4 --backend openai-compat --base-url https://my-server/v1

For authentication, set whichever environment variable your endpoint expects (e.g., OPENAI_API_KEY).

Capabilities

Feature Supported
Direct video upload No
Multi-frame batch Yes (all frames in one API call)
Audio transcription No
Text generation Yes
Preflight health check Yes

Model

Override with the --model flag:

eyeroll watch video.mp4 --backend openai-compat --base-url https://my-server/v1 -m my-model

How it works

Uses the openai SDK pointed at your custom base URL. eyeroll sends all extracted frames as base64-encoded images in a single API call, with timestamps per frame. No audio transcription is available.

Use cases

  • Self-hosted models -- vLLM, TGI, llama.cpp with OpenAI-compatible server
  • Custom proxies -- LiteLLM proxy, API gateways
  • Niche providers -- any service with an OpenAI-compatible vision endpoint

Install

pip install eyeroll[openai]

Uses the openai SDK with a custom base URL.