fly.toml

# Test deploy of the zorp-web backend for app.zorp.dev, no Ollama sidecar:
# recall/voice are compiled out (FEATURES empty) and chat goes to a real
# hosted OpenAI-compatible endpoint via ZORP_BASE_URL/ZORP_API_KEY instead.
# The frontend (web/) deploys separately, as a Cloudflare Worker. See
# web/wrangler.jsonc and web/README.md's "Deploying the UI" section.
#
# One-time setup from the repo root:
#   flyctl launch --no-deploy --copy-config --name <your-app-name>
#   flyctl volumes create zorp_state --size 1 --region <primary_region below>
#   flyctl secrets set \
#     ZORP_WEB_TOKEN=$(openssl rand -hex 16) \
#     ZORP_ALLOW_ORIGIN=https://app.zorp.dev \
#     ZORP_BASE_URL=<your model provider's OpenAI-compatible base URL> \
#     ZORP_MODEL=<model name> \
#     ZORP_API_KEY=<that provider's key>
#   flyctl deploy
#
# Then point the frontend's ZORP_API_BASE (web/index.html, or wherever it's
# templated) at this app's https://<name>.fly.dev, or a CNAME onto it, and
# start the Worker as documented in web/README.md.

app = "zorp-api-test" # rename: must be globally unique on Fly
primary_region = "ord"

[build]
  dockerfile = "zorp-web/Dockerfile"

[build.args]
  FEATURES = ""

[env]
  ZORP_STATE_DB = "/home/zorp/state/sessions.db"

[[mounts]]
  source = "zorp_state"
  destination = "/home/zorp/state"

[http_service]
  internal_port = 7777
  force_https = true
  auto_stop_machines = "stop"
  auto_start_machines = true
  min_machines_running = 0

[[vm]]
  size = "shared-cpu-1x"
  memory = "512mb"