Using LocalAI from China — connecting to upstream OpenAI/Anthropic via relay

[holysheep123]

Use Case

LocalAI can act as a drop-in OpenAI replacement, but if you want to use it as a proxy to forward requests to the real OpenAI/Anthropic APIs, those endpoints are blocked in mainland China.

Solution: Configure LocalAI to route through an API relay

In LocalAI config YAML (model config)

name: gpt-5-china
backend: openai
parameters:
  model: gpt-5
  api_key: cr_your_holysheep_key
  api_base: https://api.holysheep.ai/v1

Or using environment variables

export OPENAI_API_KEY=cr_your_key
export OPENAI_API_BASE=https://api.holysheep.ai/v1

docker run -p 8080:8080 \
  -e OPENAI_API_KEY=cr_your_key \
  -e OPENAI_API_BASE=https://api.holysheep.ai/v1 \
  localai/localai:latest

For Anthropic/Claude backends

export ANTHROPIC_API_KEY=cr_your_key
export ANTHROPIC_BASE_URL=https://api.holysheep.ai

The relay at holysheep.ai provides OpenAI/Anthropic-compatible API access from China with pay-as-you-go pricing (¥1=$1). Free registration: holysheep.ai/register

Does LocalAI's proxy mode work well with custom API bases? Happy to share more details about the setup.