alighasami
3d5eaf9445
Security: Sync from Public / sync-from-public (push) Has been cancelled
Test: Benchmark Nightly / build (push) Has been cancelled
Test: Benchmark Nightly / Notify Cats on failure (push) Has been cancelled
CI: Python / Checks (push) Has been cancelled
Test: Evals Python / Workflow Comparison Python (push) Has been cancelled
Util: Check Docs URLs / check-docs-urls (push) Has been cancelled
Test: Visual Storybook / Cloudflare Pages (push) Has been cancelled
Test: E2E Performance / build-and-test-performance (push) Has been cancelled
Test: Workflows Nightly / Run Workflow Tests (push) Has been cancelled
Util: Cleanup CI Docker Images / Delete stale CI images (push) Has been cancelled
Test: Benchmark Destroy Env / build (push) Has been cancelled
Util: Update Node Popularity / update-popularity (push) Has been cancelled
Test: E2E Coverage Weekly / Coverage Tests (push) Has been cancelled
412 lines
11 KiB
TypeScript
412 lines
11 KiB
TypeScript
/* eslint-disable n8n-nodes-base/node-filename-against-convention */
|
|
import { type INodeProperties } from 'n8n-workflow';
|
|
|
|
import { JAILBREAK_PROMPT } from './actions/checks/jailbreak';
|
|
import { NSFW_SYSTEM_PROMPT } from './actions/checks/nsfw';
|
|
import { PII_NAME_MAP, PIIEntity } from './actions/checks/pii';
|
|
import { TOPICAL_ALIGNMENT_SYSTEM_PROMPT } from './actions/checks/topicalAlignment';
|
|
import { LLM_SYSTEM_RULES } from './helpers/model';
|
|
|
|
const THRESHOLD_OPTION: INodeProperties = {
|
|
displayName: 'Threshold',
|
|
name: 'threshold',
|
|
type: 'number',
|
|
default: '',
|
|
description: 'Minimum confidence threshold to trigger the guardrail (0.0 to 1.0)',
|
|
hint: 'Inputs scoring less than this will be treated as violations',
|
|
};
|
|
|
|
const getPromptOption: (
|
|
defaultPrompt: string,
|
|
collapsible?: boolean,
|
|
hint?: string,
|
|
) => INodeProperties[] = (defaultPrompt, collapsible = true, hint) => {
|
|
const promptParameters: INodeProperties = {
|
|
displayName: 'Prompt',
|
|
name: 'prompt',
|
|
type: 'string',
|
|
default: defaultPrompt,
|
|
description:
|
|
'The system prompt used by the guardrail. Thresholds and JSON output are enforced by the node automatically.',
|
|
hint,
|
|
typeOptions: {
|
|
rows: 6,
|
|
},
|
|
};
|
|
if (collapsible) {
|
|
return [
|
|
{ displayName: 'Customize Prompt', name: 'customizePrompt', type: 'boolean', default: false },
|
|
{ ...promptParameters, displayOptions: { show: { customizePrompt: [true] } } },
|
|
];
|
|
}
|
|
return [promptParameters];
|
|
};
|
|
|
|
const wrapValue = (properties: INodeProperties[]) => ({
|
|
displayName: 'Value',
|
|
name: 'value',
|
|
values: properties,
|
|
});
|
|
|
|
export const propertiesDescription: INodeProperties[] = [
|
|
{
|
|
displayName:
|
|
'Use guardrails to validate text against a set of policies (e.g. NSFW, prompt injection) or to sanitize it (e.g. personal data, secret keys)',
|
|
name: 'guardrailsUsage',
|
|
type: 'notice',
|
|
default: '',
|
|
},
|
|
{
|
|
displayName: 'Operation',
|
|
name: 'operation',
|
|
type: 'options',
|
|
noDataExpression: true,
|
|
options: [
|
|
{
|
|
name: 'Check Text for Violations',
|
|
value: 'classify',
|
|
action: 'Check text for violations',
|
|
description: 'Validate text against a set of policies (e.g. NSFW, prompt injection)',
|
|
},
|
|
{
|
|
name: 'Sanitize Text',
|
|
value: 'sanitize',
|
|
action: 'Sanitize text',
|
|
// eslint-disable-next-line n8n-nodes-base/node-param-description-excess-final-period
|
|
description: 'Redact text to mask personal data, secret keys, URLs, etc.',
|
|
},
|
|
],
|
|
default: 'classify',
|
|
},
|
|
{
|
|
displayName: 'Text To Check',
|
|
name: 'text',
|
|
type: 'string',
|
|
required: true,
|
|
default: '',
|
|
typeOptions: {
|
|
rows: 1,
|
|
},
|
|
},
|
|
{
|
|
displayName: 'Guardrails',
|
|
name: 'guardrails',
|
|
placeholder: 'Add Guardrail',
|
|
type: 'collection',
|
|
default: {},
|
|
options: [
|
|
{
|
|
displayName: 'Keywords',
|
|
name: 'keywords',
|
|
type: 'string',
|
|
default: '',
|
|
description:
|
|
'This guardrail checks if specified keywords appear in the input text and can be configured to trigger tripwires based on keyword matches. Multiple keywords can be added separated by comma.',
|
|
displayOptions: {
|
|
show: {
|
|
'/operation': ['classify'],
|
|
},
|
|
},
|
|
},
|
|
{
|
|
displayName: 'Jailbreak',
|
|
name: 'jailbreak',
|
|
type: 'fixedCollection',
|
|
default: { value: { threshold: 0.7 } },
|
|
description: 'Detects attempts to jailbreak or bypass AI safety measures',
|
|
options: [wrapValue([THRESHOLD_OPTION, ...getPromptOption(JAILBREAK_PROMPT)])],
|
|
displayOptions: {
|
|
show: {
|
|
'/operation': ['classify'],
|
|
},
|
|
},
|
|
},
|
|
{
|
|
displayName: 'NSFW',
|
|
name: 'nsfw',
|
|
type: 'fixedCollection',
|
|
default: { value: { threshold: 0.7 } },
|
|
description: 'Detects attempts to generate NSFW content',
|
|
options: [wrapValue([THRESHOLD_OPTION, ...getPromptOption(NSFW_SYSTEM_PROMPT)])],
|
|
displayOptions: {
|
|
show: {
|
|
'/operation': ['classify'],
|
|
},
|
|
},
|
|
},
|
|
{
|
|
displayName: 'Personal Data (PII)',
|
|
name: 'pii',
|
|
type: 'fixedCollection',
|
|
default: { value: { type: 'all' } },
|
|
description: 'Detects attempts to use personal data content',
|
|
options: [
|
|
wrapValue([
|
|
{
|
|
displayName: 'Type',
|
|
name: 'type',
|
|
type: 'options',
|
|
default: '',
|
|
options: [
|
|
{ name: 'All', value: 'all' },
|
|
{ name: 'Selected', value: 'selected' },
|
|
],
|
|
},
|
|
{
|
|
displayName: 'Entities',
|
|
name: 'entities',
|
|
type: 'multiOptions',
|
|
default: [],
|
|
displayOptions: {
|
|
show: {
|
|
type: ['selected'],
|
|
},
|
|
},
|
|
options: Object.values(PIIEntity).map((entity) => ({
|
|
name: PII_NAME_MAP[entity],
|
|
value: entity,
|
|
})),
|
|
},
|
|
]),
|
|
],
|
|
},
|
|
{
|
|
displayName: 'Secret Keys',
|
|
name: 'secretKeys',
|
|
type: 'fixedCollection',
|
|
default: { value: { permissiveness: 'balanced' } },
|
|
description:
|
|
'Detects attempts to use secret keys in the input text. Scans text for common patterns, applies entropy analysis to detect random-looking strings.',
|
|
options: [
|
|
wrapValue([
|
|
{
|
|
displayName: 'Permissiveness',
|
|
name: 'permissiveness',
|
|
type: 'options',
|
|
default: '',
|
|
options: [
|
|
{
|
|
name: 'Strict',
|
|
value: 'strict',
|
|
description:
|
|
'Most sensitive, may have more false positives (commonly flag high entropy filenames or code)',
|
|
},
|
|
{
|
|
name: 'Balanced',
|
|
value: 'balanced',
|
|
description: 'Balanced between sensitivity and specificity',
|
|
},
|
|
{
|
|
name: 'Permissive',
|
|
value: 'permissive',
|
|
description:
|
|
'Least sensitive, may miss some secret keys (but also reduces false positives)',
|
|
},
|
|
],
|
|
},
|
|
]),
|
|
],
|
|
},
|
|
{
|
|
displayName: 'Topical Alignment',
|
|
name: 'topicalAlignment',
|
|
type: 'fixedCollection',
|
|
default: { value: { threshold: 0.7 } },
|
|
description: 'Detects attempts to stray from the business scope',
|
|
options: [
|
|
wrapValue([
|
|
THRESHOLD_OPTION,
|
|
...getPromptOption(
|
|
TOPICAL_ALIGNMENT_SYSTEM_PROMPT,
|
|
false,
|
|
'Make sure you replace the placeholder.',
|
|
),
|
|
]),
|
|
],
|
|
displayOptions: {
|
|
show: {
|
|
'/operation': ['classify'],
|
|
},
|
|
},
|
|
},
|
|
{
|
|
displayName: 'URLs',
|
|
name: 'urls',
|
|
type: 'fixedCollection',
|
|
default: { value: { allowedSchemes: ['https'], allowedUrls: '' } },
|
|
description: 'Blocks URLs that are not in the allowed list',
|
|
options: [
|
|
wrapValue([
|
|
{
|
|
displayName: 'Block All URLs Except',
|
|
name: 'allowedUrls',
|
|
type: 'string',
|
|
// keep placeholder to avoid limitation that removes collections with unchanged default values
|
|
default: 'PLACEHOLDER',
|
|
description:
|
|
'Multiple URLs can be added separated by comma. Leave empty to block all URLs.',
|
|
},
|
|
{
|
|
displayName: 'Allowed Schemes',
|
|
name: 'allowedSchemes',
|
|
type: 'multiOptions',
|
|
default: ['https'],
|
|
// eslint-disable-next-line n8n-nodes-base/node-param-multi-options-type-unsorted-items
|
|
options: [
|
|
// eslint-disable-next-line n8n-nodes-base/node-param-display-name-miscased
|
|
{ name: 'https', value: 'https' },
|
|
// eslint-disable-next-line n8n-nodes-base/node-param-display-name-miscased
|
|
{ name: 'http', value: 'http' },
|
|
// eslint-disable-next-line n8n-nodes-base/node-param-display-name-miscased
|
|
{ name: 'ftp', value: 'ftp' },
|
|
// eslint-disable-next-line n8n-nodes-base/node-param-display-name-miscased
|
|
{ name: 'data', value: 'data' },
|
|
// eslint-disable-next-line n8n-nodes-base/node-param-display-name-miscased
|
|
{ name: 'javascript', value: 'javascript' },
|
|
// eslint-disable-next-line n8n-nodes-base/node-param-display-name-miscased
|
|
{ name: 'vbscript', value: 'vbscript' },
|
|
// eslint-disable-next-line n8n-nodes-base/node-param-display-name-miscased
|
|
{ name: 'mailto', value: 'mailto' },
|
|
],
|
|
},
|
|
{
|
|
displayName: 'Block Userinfo',
|
|
name: 'blockUserinfo',
|
|
type: 'boolean',
|
|
default: true,
|
|
description:
|
|
'Whether to block URLs with userinfo (user:pass@domain) to prevent credential injection',
|
|
displayOptions: {
|
|
show: {
|
|
'/operation': ['classify'],
|
|
},
|
|
},
|
|
},
|
|
{
|
|
displayName: 'Sanitize Userinfo',
|
|
name: 'blockUserinfo',
|
|
type: 'boolean',
|
|
default: true,
|
|
description:
|
|
'Whether to sanitize URLs with userinfo (user:pass@domain) to prevent credential injection',
|
|
displayOptions: {
|
|
show: {
|
|
'/operation': ['sanitize'],
|
|
},
|
|
},
|
|
},
|
|
{
|
|
displayName: 'Allow Subdomains',
|
|
name: 'allowSubdomains',
|
|
type: 'boolean',
|
|
default: true,
|
|
description:
|
|
'Whether to allow subdomains (e.g. sub.domain.com if domain.com is allowed)',
|
|
},
|
|
]),
|
|
],
|
|
},
|
|
{
|
|
displayName: 'Custom',
|
|
name: 'custom',
|
|
type: 'fixedCollection',
|
|
typeOptions: {
|
|
sortable: true,
|
|
multipleValues: true,
|
|
},
|
|
placeholder: 'Add Custom Guardrail',
|
|
default: {
|
|
guardrail: [{ name: 'Custom Guardrail' }],
|
|
},
|
|
options: [
|
|
{
|
|
displayName: 'Guardrail',
|
|
name: 'guardrail',
|
|
values: [
|
|
{
|
|
displayName: 'Name',
|
|
name: 'name',
|
|
type: 'string',
|
|
default: '',
|
|
description: 'Name of the custom guardrail',
|
|
},
|
|
THRESHOLD_OPTION,
|
|
...getPromptOption('', false),
|
|
],
|
|
},
|
|
],
|
|
displayOptions: {
|
|
show: {
|
|
'/operation': ['classify'],
|
|
},
|
|
},
|
|
},
|
|
{
|
|
displayName: 'Custom Regex',
|
|
name: 'customRegex',
|
|
type: 'fixedCollection',
|
|
typeOptions: {
|
|
sortable: true,
|
|
multipleValues: true,
|
|
},
|
|
placeholder: 'Add Custom Regex',
|
|
default: {},
|
|
options: [
|
|
{
|
|
displayName: 'Regex',
|
|
name: 'regex',
|
|
values: [
|
|
{
|
|
displayName: 'Name',
|
|
name: 'name',
|
|
type: 'string',
|
|
default: '',
|
|
description:
|
|
'Name of the custom regex. Will be used for replacement when sanitizing.',
|
|
},
|
|
{
|
|
displayName: 'Regex',
|
|
name: 'value',
|
|
type: 'string',
|
|
default: '',
|
|
description: 'Regex to match the input text',
|
|
placeholder: '/text/gi',
|
|
},
|
|
],
|
|
},
|
|
],
|
|
},
|
|
],
|
|
},
|
|
{
|
|
displayName: 'Customize System Message',
|
|
name: 'customizeSystemMessage',
|
|
description:
|
|
'Whether to customize the system message used by the guardrail to specify the output format',
|
|
type: 'boolean',
|
|
default: false,
|
|
displayOptions: {
|
|
show: {
|
|
'/operation': ['classify'],
|
|
},
|
|
},
|
|
},
|
|
{
|
|
displayName: 'System Message',
|
|
name: 'systemMessage',
|
|
type: 'string',
|
|
description:
|
|
'The system message used by the guardrail to enforce thresholds and JSON output according to schema',
|
|
hint: 'This message is appended after prompts defined by guardrails',
|
|
default: LLM_SYSTEM_RULES,
|
|
typeOptions: {
|
|
rows: 6,
|
|
},
|
|
displayOptions: {
|
|
show: {
|
|
'/customizeSystemMessage': [true],
|
|
},
|
|
},
|
|
},
|
|
];
|