Skip to main content

​
First Run

​
Supported First-Run Loop

loong onboard
loong ask --message "Summarize this repository and suggest the best next step."
Onboarding is the supported first-run path. It is expected to detect reusable config where possible, write credential references instead of inline secrets, and hand you off to a first concrete command.

​
Continue The Session

When you need follow-up work: 
loong chat
If you want to inspect the active provider’s model catalog after credentials are healthy: 
loong list-models

​
What Success Looks Like

  • loong onboard finishes without leaving you in raw config editing as step zero.
  • loong ask produces one useful answer.
  • loong chat is available for follow-up work.
  • loong doctor is there as the repair path if something still looks off.

​
If First Run Feels Incomplete

Go straight to Doctor And Health when:
  • credentials were not detected cleanly
  • the provider path still feels ambiguous
  • ask or chat does not behave like a healthy local assistant path

​
Approximate Config Shape

active_provider = "openai"

[providers.openai]
kind = "openai"
api_key = { env = "PROVIDER_API_KEY" }
The exact provider kind and env var will vary. The important part is that onboarding writes the same provider surface the runtime will keep using later. Secret reference note:
  • api_key = { env = "PROVIDER_API_KEY" } means “read the secret from that environment variable”.
  • api_key = "PROVIDER_API_KEY" would instead be treated as the literal API key value itself.
  • The docs standardize on the { env = "..." } shape because it is explicit and matches the wider secret-reference model.

​
Why This Page Stays Short

  • The first-run handoff should get you to one useful answer quickly.
  • Provider matrices, channel matrices, and richer runtime policy details belong in deeper guides.
  • The local CLI success path should come before gateway, channels, or broader product surfaces.

​
What Comes Next

  • Continue to Providers And Models if you want the public provider and model-selection contract.
  • Continue to Doctor And Health if the first run is incomplete or unstable.
  • Continue to Use Loong once the base operator path is healthy.