Set up the support email project

Use Python 3.12 and a source checkout of Foliqant. From the repository root:

uv sync --locked --extra openai --extra mcp
mkdir -p my_support/config/support_email/classify

Create files below my_support/. The final configuration is a comparison point. Create my_support/config/settings.yaml:

models:
  local_qwen:
    provider: openai_compatible
    model: $MODEL_ID
    base_url: $MODEL_BASE_URL
    allow_insecure_http: true
    output_mode: native
    supports_tools: false
execution:
  concurrency: 1
  queue_limit: 0

These complete environment references resolve when clients open. validate compiles offline without an endpoint. Set them only when making a live model request, for example in an untracked my_support/config/.env:

MODEL_ID=your-served-qwen-model-id
MODEL_BASE_URL=http://127.0.0.1:8000/v1

Match those values to your local server. The finished example uses a scripted model by default and opens no model connection:

uv run --no-sync python -m examples.support_email_tutorial.run

Expect execution.status: completed, payload.queue: billing, and payload.account_reference: A-100. The next chapters show the files to write before reaching that snapshot. Continue with classification. For installation choices, see Install and run.