langwatch/langwatch
Support
Dashboard
langwatch/langwatch

API Endpoints

Content Safety

Documentation

LangEvals

Open Dashboard

GitHub Repo

Documentation

Introduction
Evaluators
Unit Tests
API Example
How to Choose Your Evaluator Guide
Modular Architecture

Tutorials

Extensive Unit Testing
RAG Evaluation
Evaluate on CI/CD Pipeline

API Endpoints

POST
Language Detection
POST
OpenAI Moderation
POST
PII Detection
POST
Content Safety
POST
Jailbreak Detection
POST
Prompt Injection Detection
POST
Competitor Detection with LLM
POST
Prompt Injection Detection
POST
Off-Topic Detection
POST
Answer Relevancy
POST
Context Precision
POST
Context Recall
POST
Context Relevancy
POST
Context Utilization
POST
Faithfulness
POST
Haystack Faithfulness
POST
LLM Boolean Evaluator
POST
LLM Score Evaluator
POST
LLM Basic Evaluator
POST
LLM Similarity Evaluator
POST
Llama Guard

API Endpoints

Content Safety

This evaluator detects potentially unsafe content in text, including hate speech, self-harm, sexual content, and violence. It allows customization of the severity threshold and the specific categories to check.

Env vars: AZURE_CONTENT_SAFETY_ENDPOINT, AZURE_CONTENT_SAFETY_KEY

Docs: https://learn.microsoft.com/en-us/azure/ai-services/content-safety/quickstart-text

POST

azure

content_safety

evaluate

object

data

array

data

Required

array

List of entries to be evaluated, check the field type for the necessary keys

settings

object

settings

object

Evaluator settings, check the field type for what settings this evaluator supports

severity_threshold

enum<integer>

severity_threshold

enum<integer>

The minimum severity level to consider content as unsafe, from 1 to 7.

Body

application/json

data

object[]

required

List of entries to be evaluated, check the field type for the necessary keys

data.input

string | null

data.output

string | null

settings

object | null

Evaluator settings, check the field type for what settings this evaluator supports

settings.severity_threshold

enum<integer>

default: 1

The minimum severity level to consider content as unsafe, from 1 to 7.

Available options:

1,

2,

3,

4,

5,

6,

7

settings.categories

object

The categories of moderation to check for.

settings.categories.Hate

boolean

default: true

settings.categories.SelfHarm

boolean

default: true

settings.categories.Sexual

boolean

default: true

settings.categories.Violence

boolean

default: true

settings.output_type

enum<string>

default: FourSeverityLevels

The type of severity levels to return on the full 0-7 severity scale, it can be either the trimmed version with four values (0, 2, 4, 6 scores) or the whole range.

Available options:

FourSeverityLevels,

EightSeverityLevels

env

object | null

Optional environment variables to override the server ones

env.{key}

string

Response

200 - application/json

status

enum<string>

default: processed

Available options:

processed

score

number

required

The severity level of the detected content from 0 to 7. A higher score indicates higher severity.

passed

boolean | null

details

string | null

Short human-readable description of the result

cost

object | null

cost.currency

string

required

cost.amount

number

required

PII Detection Jailbreak Detection

github linkedin