#!/usr/bin/env bash

export INFERENCE_MODEL="inference-llama4-maverick"
export EMBEDDING_MODEL="inference-bge-m3"
export EMBEDDING_DIMENSION="1024"
export LLAMA_STACK_PORT=8321
export OPENAI_BASE_URL=https://maas.ai-2.kvant.cloud/v1
export OPENAI_API_KEY=sk-ZqAWqBKFXjb6y3tVej2AaA
export VLLM_MAX_TOKENS=125000

docker run -it \
  -p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \
  -v $(pwd)/data:/root/.llama \
  --mount type=bind,source="$(pwd)"/llama_stack/templates/kvant/run.yaml,target=/root/.llama/config.yaml,readonly \
  --entrypoint python \
  distribution-kvant:dev \
  -m llama_stack.distribution.server.server  --config /root/.llama/config.yaml \
  --port $LLAMA_STACK_PORT \
  --env VLLM_URL=$OPENAI_BASE_URL \
  --env VLLM_API_TOKEN=$OPENAI_API_KEY \
  --env PASSTHROUGH_URL=$OPENAI_BASE_URL \
  --env PASSTHROUGH_API_KEY=$OPENAI_API_KEY \
  --env INFERENCE_MODEL=$INFERENCE_MODEL \
  --env EMBEDDING_MODEL=$EMBEDDING_MODEL \
  --env EMBEDDING_DIMENSION=$EMBEDDING_DIMENSION \