feat(guardrails): added guardrails block/tools and docs

[waleedlatif1]

Summary

added guardrails block/tools and docs

Type of Change

• New feature

Testing

Tested manually.

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Preview	Comments	Updated (UTC)
docs	Skipped			Oct 12, 2025 2:39am

[greptile-apps]

Greptile Overview

Summary

Added comprehensive guardrails block with four validation types: JSON, Regex, Hallucination Check (RAG + LLM scoring), and PII Detection (Microsoft Presidio). Integrated with workflow blocks system, added new grouped checkbox UI component for PII type selection, updated Docker setup for Python dependencies, and included full documentation.

Key Changes:

• New guardrails block configuration with conditional UI based on validation type
• API route at /api/guardrails/validate with input validation and error handling
• TypeScript validators for JSON, regex, and hallucination detection
• Python-based PII detection using Microsoft Presidio with TypeScript wrapper
• New grouped-checkbox-list sub-block component for multi-select PII types
• Docker configuration updated to install Python 3, pip, and run setup script
• Comprehensive documentation with examples and best practices

Critical Issue:

• validate_pii.ts:82 - Promise constructor missing reject parameter, causing runtime errors on lines 100 and 178

Confidence Score: 2/5

• This PR has a critical bug that will cause runtime failures in PII validation
• Score reflects a critical logical error in validate_pii.ts where the Promise constructor is missing the reject parameter (line 82), but reject is called on lines 100 and 178. This will cause immediate runtime errors when PII validation times out or encounters spawn errors. The rest of the implementation is well-structured with good error handling, comprehensive validation logic, and proper documentation.
• Critical attention required for apps/sim/lib/guardrails/validate_pii.ts - must fix Promise constructor before merge

Important Files Changed

File Analysis

Filename	Score	Overview
apps/sim/lib/guardrails/validate_pii.ts	1/5	PII validation with Python integration - critical bug: Promise missing reject parameter (lines 100, 178)
apps/sim/lib/guardrails/validate_hallucination.ts	4/5	Hallucination detection using RAG + LLM scoring - well-structured with proper error handling
apps/sim/blocks/blocks/guardrails.ts	5/5	Block configuration for guardrails with comprehensive PII types and conditional UI logic
apps/sim/app/api/guardrails/validate/route.ts	5/5	API route with input validation, type checking, and proper error handling
apps/sim/lib/guardrails/validate_pii.py	5/5	Python PII detection using Presidio with proper error handling and JSON output
docker/app.Dockerfile	5/5	Added Python 3 and pip to runner stage, copies guardrails setup files and runs setup script

Sequence Diagram

sequenceDiagram
    participant User
    participant Block as Guardrails Block
    participant Tool as guardrails_validate Tool
    participant API as /api/guardrails/validate
    participant Validator as Validation Logic
    participant KB as Knowledge Base API
    participant LLM as LLM Provider
    participant Python as Python Script (PII)

    User->>Block: Configure validation type & parameters
    User->>Block: Wire input from previous block
    
    Block->>Tool: Execute with input & config
    Tool->>API: POST /api/guardrails/validate
    
    alt JSON Validation
        API->>Validator: validateJson(input)
        Validator->>API: {passed, error?}
    else Regex Validation
        API->>Validator: validateRegex(input, pattern)
        Validator->>API: {passed, error?}
    else Hallucination Check
        API->>Validator: validateHallucination(input, config)
        Validator->>KB: Query knowledge base (RAG)
        KB->>Validator: Return relevant chunks
        Validator->>LLM: Score confidence (0-10)
        LLM->>Validator: {score, reasoning}
        Validator->>API: {passed, score, reasoning}
    else PII Detection
        API->>Validator: validatePII(input, config)
        Validator->>Python: Spawn validate_pii.py
        Python->>Python: Presidio analyze & anonymize
        Python->>Validator: {passed, detectedEntities, maskedText?}
        Validator->>API: {passed, detectedEntities, maskedText?}
    end
    
    API->>Tool: {success, output}
    Tool->>Block: Return validation result
    Block->>User: Display result (passed/failed)

Loading

_{25 files reviewed, 1 comment}

_{Edit Code Review Agent Settings | Greptile}