AI Providers
Use this page to choose provider roles and configure Custom / Local.
Provider model in Genie
Section titled “Provider model in Genie”Genie separates provider settings into two roles:
- Coding: code.
- Logistics: preflight, QUERY, DATA, and ASK coordination.
You can point both roles at the same provider, or split them when you want different cost, speed, or behavior profiles.
Supported provider choices
Section titled “Supported provider choices”The Settings UI currently exposes these provider families for both roles:
| Provider | What it means in Genie | Typical use |
|---|---|---|
| Custom / Local | OpenAI-compatible endpoint with an arbitrary model id | Local models, self-hosted gateways, or compatible proxies |
| OpenAI | Hosted OpenAI account and model selection | General default for daily Genie usage |
| Anthropic | Hosted Anthropic account and model selection | Alternate hosted reasoning/coding profile |
| Hosted Google account and model selection | Alternate hosted provider path |
Bring your own keys
Section titled “Bring your own keys”Genie does not bundle model usage.
- Your Genie activation key unlocks product access.
- Your provider keys authorize the models you choose.
- Billing, quotas, and account policy for hosted models remain with the provider account you configure.
That means activation and provider credentials are separate systems.
Role-scoped settings
Section titled “Role-scoped settings”Each role can carry its own:
- provider
- model
- reasoning effort
- API key
- custom endpoint when
Custom / Localis selected
When a role uses Custom / Local, Genie also exposes:
- a free-text model id field
- an OpenAI-compatible endpoint field
- a Test Connection button
Both roles are first-class. You can run hosted providers, local providers, or a mix of both without changing how Genie presents the main workflow.
Custom / Local behavior
Section titled “Custom / Local behavior”Custom / Local is for OpenAI-compatible endpoints.
Use it when you want Genie to talk to:
- a local model server
- a self-hosted gateway
- a compatible proxy that exposes OpenAI-style request/response behavior
What you configure
Section titled “What you configure”- Model id: any model identifier your endpoint accepts
- Endpoint: per-role base URL such as
http://localhost:1234/v1 - API key: optional or required depending on your endpoint
Connection testing
Section titled “Connection testing”The Test Connection button checks endpoint reachability before you save. Use it whenever you change the endpoint or switch models/providers.
For a step-by-step local setup path, see LM Studio Setup (Custom / Local).
Choosing a provider strategy
Section titled “Choosing a provider strategy”- Use OpenAI when you want the most straightforward hosted default.
- Use Custom / Local when you want local control, self-hosting, or compatibility with an existing gateway.
- Use Anthropic or Google when those providers match your team standard or preferred model behavior.
- Split Coding and Logistics only when you want different models for generation vs planning.
- For a local-model walkthrough, see LM Studio Setup (Custom / Local).
If the UI is locked but provider settings look fine
Section titled “If the UI is locked but provider settings look fine”Activation and provider access are different systems. A correct provider setup does not bypass a locked Genie UI.
Use Licensing and Activation when the interface is locked. Use API and Provider Limits when the interface is available but model calls fail.