Budibase/budibase · error

Tool name must be under 64 characters long

Error message

Tool name must be under 64 characters long

What it means

sanitiseToolName enforces LLM tool-name constraints: names must match ^[a-zA-Z0-9_-]+$ and be at most 64 characters. It sanitizes illegal characters to underscores but throws outright if the name exceeds 64 characters because truncation could collide or break mapping to the original tool.

Source

Thrown at packages/pro/src/ai/utils.ts:6

import { SupportedFileType } from "@budibase/types"

// tool names must align with regex "^[a-zA-Z0-9_-]+$"
export function sanitiseToolName(name: string) {
  if (name.length > 64) {
    throw new Error("Tool name must be under 64 characters long")
  }
  return name.replace(/[^a-zA-Z0-9_-]/g, "_")
}

export function normalizeContentType(contentType?: string): string {
  if (!contentType) {
    return "application/octet-stream"
  }

  // If it's already a proper MIME type, return it
  if (contentType.includes("/")) {
    return contentType
  }

  // Map SupportedFileType values to proper MIME types
  const mimeTypeMap: Record<string, string> = {
    [SupportedFileType.PDF]: "application/pdf",
    [SupportedFileType.JPG]: "image/jpeg",

View on GitHub (pinned to a81a902e9a)

Solutions

1. Shorten the underlying resource name (table/query/datasource/automation) to keep the derived tool name under 64 chars.
2. Truncate or hash the tool name before calling sanitiseToolName, keeping a mapping back to the full id.
3. Add pre-validation in the UI/API that warns when a resource name will exceed the tool-name limit.

Example fix

// before
const name = sanitiseToolName(`table_${table.name}`) // may exceed 64
// after
const raw = `table_${table.name}`
const name = sanitiseToolName(raw.length > 64 ? `table_${raw.slice(6, 60)}_${hash(table.id)}` : raw)

if (`table_${name}`.length > 64) throw new Error(`Resource name "${name}" is too long for AI tool registration (max 64 incl. prefix)`)

function isValidToolName(name: string): boolean {
  return name.length <= 64 && /^[a-zA-Z0-9_-]+$/.test(name.replace(/[^a-zA-Z0-9_-]/g, "_").slice(0, 64))
}

try {
  const tool = sanitiseToolName(rawName)
} catch (e) {
  if (e.message.includes("under 64 characters")) {
    const tool = sanitiseToolName(truncateWithHash(rawName, 64))
  } else throw e
}

When it happens

Trigger: Registering an AI tool whose derived name (e.g. from a table, datasource, query, or automation name) exceeds 64 characters — long resource names get concatenated into tool identifiers and blow the limit.

Common situations: Very long datasource/query names combined with prefixes like "query_" or "table_"; SCIM-synced or imported resources with long names; other providers (OpenAI/Anthropic) then reject the tool at the API level if the check were skipped.

Related errors

• Cannot create a query tool without a query ID
• Failed to generate a unique name for the row action
• AI message content must be a string
• Variable name has characters that are not allowed
• ${failureMessage}

AI-assisted analysis of Budibase/budibase@a81a902e9a (2026-08-29). Data as JSON: /api/errors/86864a271fc4232a. Report an issue: GitHub.