Budibase/budibase · error · HTTPError

Unsupported AI config type

Error message

Unsupported AI config type: ${model.configType}

What it means

validateConfig is the gatekeeper for AI config creation/update against LiteLLM. It only supports configs whose configType is AIConfigType.COMPLETIONS; any other (unknown or future) config type is rejected with HTTP 400 before delegating to validateCompletionsModel. The enum (packages/types/src/documents/global/ai.ts) currently only contains COMPLETIONS, so this fires when a caller supplies a mistyped, stale, or invented configType string.

Solutions

1. Set configType to "completions" in the request body (or omit it where the controller defaults it to AIConfigType.COMPLETIONS).
2. Check the exact enum value in @budibase/types (AIConfigType) rather than guessing the string.
3. If a genuinely new config type is needed, extend AIConfigType and add a matching validator in validateConfig.

Example fix

// before
await api.post("/ai/configs", { name: "gpt4o", configType: "chat", ... })
// after
await api.post("/ai/configs", { name: "gpt4o", configType: "completions", ... })

const ALLOWED = ["completions"]
if (!ALLOWED.includes(config.configType)) {
  throw new Error(`configType must be one of ${ALLOWED.join(", ")}, got ${config.configType}`)
}

function isCompletionsConfig(c: { configType: string }): c is { configType: "completions" } {
  return c.configType === "completions"
}

try {
  await createConfig(cfg)
} catch (e) {
  if (e.status === 400 && e.message.startsWith("Unsupported AI config type")) {
    cfg.configType = "completions"
    await createConfig(cfg)
  } else throw e
}

When it happens

Trigger: POST/PUT to the AI config endpoints (create/update in sdk/workspace/ai/configs/index.ts) with body.configType set to anything other than "completions" — e.g. "completion" (typo), "chat", "embedding", or a value from an older client.

Common situations: Hand-rolled API calls to the configs endpoint with a wrong configType; a client SDK or script written against an older/preview enum; copying config JSON between environments where the type drifted.

Related errors

• Invalid status query
• A new verification key is required when changing the embed…
• Agent name is required.
• Agent with name ' ' already exists.
• agentId in body does not match path

AI-assisted analysis of Budibase/budibase@a81a902e9a (2026-08-29). Data as JSON: /api/errors/bfb2661f9245d3f9. Report an issue: GitHub.

Appendix: source

Thrown at packages/server/src/sdk/workspace/ai/configs/litellm.ts:373

    const message = [
      "Error validating configuration",
      sanitizeLiteLLMErrorMessage(json.error?.message || json.result?.error),
    ]
      .filter(Boolean)
      .join(": ")

    throw new HTTPError(message, 400)
  }
}

export async function validateConfig(model: {
  provider: string
  name: string
  credentialFields: Record<string, string>
  configType: AIConfigType
}) {
  if (model.configType !== AIConfigType.COMPLETIONS) {
    throw new HTTPError(`Unsupported AI config type: ${model.configType}`, 400)
  }
  return validateCompletionsModel(model)
}

export async function getKeySettings(): Promise<{
  keyId: string
  secretKey: string
  teamId: string
}> {
  const db = context.getWorkspaceDB()
  const keyDocId = docIds.getLiteLLMKeyID()

  let keyConfig = await db.tryGet<LiteLLMKeyConfig>(keyDocId)
  if (!keyConfig || !keyConfig.teamId) {
    const workspaceId = context.getProdWorkspaceId()
    if (!workspaceId) {
      throw new HTTPError("Workspace ID is required to configure LiteLLM", 400)
    }

View on GitHub (pinned to a81a902e9a)