{{define "content"}}

Language models

Providers are cognitive service suppliers. Roles decide which model does which job; minds keep their identity whatever you change here.

{{$admin := .Principal.User.Role.AtLeast "administrator"}}

Providers

{{if .Data.Providers}} {{range .Data.Providers}}{{end}}
NameKindEndpointKeyStatusLast test
{{.Name}}{{.Kind}}{{.Endpoint}} {{.KeyDescribed}} {{if .Enabled}}enabled{{else}}disabled{{end}}{{if .AllowPrivate}} private ok{{end}} {{if .LastTestAt}}{{if deref .LastTestOK}}ok{{else}}failed{{end}} {{trunc .LastTestMsg 60}}{{else}}never{{end}}
{{else}}

No providers yet. Add one on the right; local Ollama needs “allow private addresses”.

{{end}}

Models

{{if .Data.Models}} {{range .Data.Models}}{{end}}
ProviderModelKindContextCost in/out per MState
{{.Provider}}{{.Name}}{{.Kind}}{{if .Dims}} ({{.Dims}}d){{end}}{{if .ContextWindow}}{{.ContextWindow}}{{else}}—{{end}} ${{.InputCostPerM}} / ${{.OutputCostPerM}} {{if .Enabled}}on{{else}}off{{end}} {{if $admin}}
{{end}}
{{else}}

No models yet.

{{end}} {{if and $admin .Data.Providers}}
{{end}}

Roles which model does which cognitive job

{{if $admin}}{{end}}{{range .Data.Roles}}{{$a := .Assignment}} {{if $admin}}{{end}} {{end}}
RolePurposePrimaryFallbacksTempMax tokens
{{.Info.Role}}{{.Info.Description}} ({{.Info.Suggested}})

Usage last 24 hours

{{if .Data.Usage}}{{range .Data.Usage}}{{end}}
ProviderModelRoleCallsErrorsTokens inTokens outCostAvg ms
{{.Provider}}{{.Model}}{{.Role}}{{.Calls}}{{.Errors}}{{.PromptTokens}}{{.OutputTokens}}${{printf "%.4f" .Cost}}{{printf "%.0f" .AvgLatencyMS}}
{{else}}

No calls yet.

{{end}} {{if .Data.Errors}}

Recent errors

    {{range .Data.Errors}}
  • {{.CreatedAt.Local.Format "01-02 15:04:05"}} {{.Provider}}/{{.Model}} ({{.Role}}): {{trunc .Error 160}}
  • {{end}}
{{end}}
{{if $admin}}

Add provider

Pasted keys are encrypted with the master key and never shown again. References (env:, file:) are resolved at call time.

{{end}}

Budgets

Global
{{.Data.GlobalBudget}} tokens/day
Used today
{{.Data.UsedToday}}
Concurrency
{{.Data.MaxConcurrent}} calls

Per-mind budgets are in each mind's configuration. When a budget is exhausted, deterministic cognition continues and model-assisted steps are skipped.

{{end}}