Model Providers

Roster uses a model during participant resolution to interpret the workflow question, select relevant participants, and produce an auditable answer.

The deployment owns the model choice. Roster records model provider, model name, latency, token usage, and resolution metadata for observability.

Admins can store resolver agent metadata and cost inputs in Settings. Provider credentials, provider family defaults, and model allowlists are still deployment runtime configuration.

Provider Model

The deployment chooses one runtime provider. OpenAI is the default:

OPENAI_API_KEY=<secret>
OPENAI_BASE_URL=https://api.openai.com/v1
ROSTER_MODEL_PROVIDER=openai-responses
ROSTER_MODEL_NAME=gpt-5.4
ROSTER_MODEL_EFFORT=low

Mistral and Anthropic are also supported directly:

MISTRAL_API_KEY=<secret>
MISTRAL_BASE_URL=https://api.mistral.ai/v1
ROSTER_MODEL_PROVIDER=mistral-chat-completions
ROSTER_MODEL_NAME=mistral-medium-3-5

ANTHROPIC_API_KEY=<secret>
ANTHROPIC_BASE_URL=https://api.anthropic.com
ROSTER_MODEL_PROVIDER=anthropic-messages
ROSTER_MODEL_NAME=claude-sonnet-4-20250514

*_BASE_URL values are optional. Set them only when routing through a provider proxy, a regional endpoint, or an approved model gateway.

Supported Provider Families

Roster is designed for bring-your-own-model deployments. The deployment can standardize on a direct provider or route through an approved model gateway.

Provider family	Production use
OpenAI	Direct `openai-responses` provider.
Anthropic Claude	Direct `anthropic-messages` provider.
Mistral	Direct `mistral-chat-completions` provider.
Google Gemini	Use through an approved gateway or adapter with schema-compatible responses.
DeepSeek	Use through an approved gateway or adapter with production rate limits and logging.
Moonshot Kimi	Use through an approved gateway or adapter with approved data residency controls.
Other approved models	Use the same gateway path and restrict with `ROSTER_MODEL_ALLOWED`.

Model Allowlist

Use ROSTER_MODEL_ALLOWED to restrict which provider/model pairs can run resolution:

ROSTER_MODEL_ALLOWED=openai-responses:gpt-5.4,mistral-chat-completions:mistral-medium-3-5,anthropic-messages:claude-sonnet-4-20250514

For model gateways, use the provider/model key your deployment exposes:

ROSTER_MODEL_ALLOWED=openai-responses:anthropic.claude-sonnet,openai-responses:mistral.large

Reasoning Effort

ROSTER_MODEL_EFFORT controls the requested reasoning level for providers that support effort controls. Roster accepts the value for all providers, but direct Mistral and Anthropic requests currently ignore it.

none
minimal
low
medium
high
xhigh

Use lower effort for high-volume routing and higher effort for complex, high-impact workflows.

Production Checklist

Choose the model provider or gateway before exposing resolution to users.
Store model credentials in the deployment secret manager.
Set ROSTER_MODEL_ALLOWED for production deployments.
Verify structured response behavior before approving a non-OpenAI model.
Confirm data residency, retention, and logging requirements with the model provider or gateway.
Monitor model runs for latency, cost, error rate, and resolution quality.

Use Model Runs to inspect recorded invocations and correlate provider/model choices with latency, token usage, cost, and errors.