Use custom open ai api compatible endpoints with vscode copilot via local proxy

config.yaml

# config.yaml

# This section defines the models that your local proxy will advertise.
model_list:
  - model_name: openai/gpt-oss-120b
    litellm_params:
      model: nvidia_nim/openai/gpt-oss-120b # The actual model name on OpenRouter
      supports_function_calling: true
      supports_tool_choice: true
      supports_parallel_function_calling: true
      supports_streaming: true
      supports_chat: true
      rpm: 30
  - model_name: nvidia/llama-3.3-nemotron-super-49b-v1.5
    litellm_params:
      model: nvidia_nim/nvidia/llama-3.3-nemotron-super-49b-v1.5 # The actual model name on OpenRouter
      supports_function_calling: true
      supports_tool_choice: true
      supports_parallel_function_calling: true
      supports_streaming: true
      supports_chat: true
      rpm: 30

run.sh

#!/usr/bin/env bash

set -euo pipefail

# Export environment variables from .env if it exists
if [ -f .env ]; then
    echo "Exporting environment variables from .env"
    set -a
    source .env
    set +a
fi

uv run litellm --config config.yaml &
LITELLM_PID=$!
echo "Started litellm with PID $LITELLM_PID"

uvx oai2ollama &
OAI2OLLAMA_PID=$!
echo "Started oai2ollama (via uvx) with PID $OAI2OLLAMA_PID"

# Forward signals and cleanup
cleanup() {
    echo "\nStopping background processes..."
    kill $LITELLM_PID $OAI2OLLAMA_PID 2>/dev/null
    wait $LITELLM_PID $OAI2OLLAMA_PID 2>/dev/null
    exit 0
}
trap cleanup SIGINT SIGTERM

# Wait for both background processes
wait $LITELLM_PID $OAI2OLLAMA_PID