# Local run only. Production is Google Cloud Run — see commands.md.
#
#   docker compose run --rm ingest -> ingest the corpus into Qdrant (once)
#   docker compose up              -> the Streamlit app at http://localhost:8501
#   docker compose run --rm eval   -> one LangSmith evaluation run, then exit
#
# `docker compose up` also bootstraps ingestion on first run if the Qdrant
# collection is missing, so the explicit `ingest` step is optional.
#
# All read secrets from .env. Calls to Vertex AI Gemini and Cloud Storage
# use your local gcloud Application Default Credentials, mounted read-only
# below — run `gcloud auth application-default login` first.
#
#   Linux / macOS : ADC lives at ~/.config/gcloud  (the default below)
#   Windows       : set GCLOUD_CONFIG=%APPDATA%\gcloud before `docker compose up`
#
# The app runs in "open local mode" (no Google login) unless you also mount
# a Streamlit [auth] secrets file at /app/.streamlit/secrets.toml.

x-gcloud-adc: &gcloud-adc
  - ${GCLOUD_CONFIG:-~/.config/gcloud}:/root/.config/gcloud:ro

services:
  app:
    build: .
    command: >
      streamlit run app.py
      --server.port=8501 --server.address=0.0.0.0 --server.headless=true
      --server.enableCORS=false --server.enableXsrfProtection=false
    ports:
      - "8501:8501"
    env_file: [.env]
    volumes: *gcloud-adc

  # One-off jobs, not services. profiles:[tools] keeps them out of
  # `docker compose up`.
  ingest:
    build: .
    command: python ingest.py
    env_file: [.env]
    volumes: *gcloud-adc
    profiles: [tools]

  eval:
    build: .
    command: python evaluate.py
    env_file: [.env]
    volumes: *gcloud-adc
    profiles: [tools]