Model listing

listModels() fetches the model inventory visible to the configured Mistral client and normalizes it to Anvia's ModelList.

const models = await mistral.listModels()

for (const model of models.data) {
  console.log(model.id)
  console.log(model.name)
  console.log(model.contextLength)
}

Every normalized entry has an id. When supplied by Mistral, it can also include a name, description, type, creation time, owner, and context length.

Use model listing for administration, inventory, or deployment diagnostics. A listed ID is not proof that it supports a workflow's tools, schemas, context size, OCR, or embedding configuration.

Keep selection controlled

Do not turn the complete provider inventory into an unrestricted user-controlled selector. Keep a reviewed application allow-list:

const allowedModels = new Set([
  'mistral-large-latest',
  'mistral-small-latest',
])

export function selectCompletionModel(modelId: string) {
  if (!allowedModels.has(modelId)) {
    throw new Error('Unsupported model selection')
  }

  return mistral.completionModel({
      modelId: modelId
  })
}

These IDs illustrate application policy, not a universal recommendation. Select and test models for the deployment's quality, latency, cost, account, and endpoint requirements.

Treat inventory as fallible

Listing can fail, change independently of the application, or include models inappropriate for the current workflow. Provider failures are converted to an Anvia ModelListingError that includes the provider name and an available status code.

Avoid making every user request depend on a fresh listing call. Cache inventory for administrative views and keep production model selection in reviewed configuration.