Skip to main content
Back to top
Ctrl
+
K
PyRIT
Cookbooks
1. Sending a Million Prompts
2. Precomputing Turns for Attacks
3. Probing for copyright violations with FirstLetterConverter and PlagiarismScorer
4. Benchmarking models + Testing For Fairness/Bias
5: Automating and Evaluating Psychosocial Harms: - Optional
Install PyRIT Library
Install PyRIT with Docker
Install PyRIT with Conda
What can I do if Jupyter cannot find PyRIT?
Populating Secrets - Quick Start Guide
Azure SQL Database Setup, Authentication and User Permissions
Contributing
Setting up PyRIT Development Environment with uv (Windows)
DevContainers Setup in Visual Studio Code
Local Installation with Conda/Python
2. Contribute with Git
3. Incorporating Research
4. Style Guide
5. Running Tests
6. Unit Tests
7. Integration Tests
8. Notebooks
9. Pre-Commit Hooks
10. Exception Handling Guidelines
11. Releasing PyRIT
Architecture
User guide
Datasets
1. Loading Built-in Datasets
2. Creating Seeds Programmatically and with YAML
3. Writing Your Own Datasets
4. Contributing Datasets to PyRIT
Executor
Attack
1. Prompt Sending Attack (Single-Turn)
2. Red Teaming Attack (Multi-Turn)
3. Crescendo Attack (Multi-Turn)
Chunked Request Attack
Context Compliance Attack (Single-Turn) - optional
Flip Attack (Single-Turn) - optional
Many-Shot Jailbreak Attack (Single-Turn) - optional
Multi-Prompt Sending Attack - optional
Role Play Attack (Single-Turn) - optional
Skeleton Key Attack (Single-Turn) - optional
Tree of Attacks with Pruning (Multi-Turn) - optional
Violent Durian Attack (Multi-Turn) - optional
Workflow
1. Cross-domain Prompt Injection Attack (XPIA) via a website
2. Cross-domain Prompt Injection (XPIA) - AI Recruiter RAG Vulnerability
Benchmark
1. Q&A Benchmark
Prompt Generator
1. Anecdoctor Prompt Generator
GPTFuzzer Prompt Generator - optional
Prompt Targets
1. OpenAI Chat Target
2. Creating Custom Targets
3. AML Chat Targets
4. Azure Blob Storage Targets
5. Multi-Modal Targets
6. Rate Limit (RPM) Threshold
7. HTTP Target
8. OpenAI Responses Target
9. MessageNormalizer
OpenAI Completions - optional
Playwright Target - optional
Playwright Copilot Target - optional
Prompt Shield Target - optional
HuggingFace Chat Target - optional
Realtime Target - optional
Converters
1. Text-to-Text Converters
2. Audio Converters
3. Image Converters
4. Video Converters
5. File Converters
6. Selectively Converting
7. Human in the Loop Converter
Scoring
1. Float Scale Scoring using Azure Content Safety API
2. True False Scoring
3. Classification Scoring
4. Float Scale Scoring using Likert Scale
5. Human in the Loop Scoring with Gradio
6. Refusal Scorer
7. Batch Scoring
Insecure Code Scorer - optional
Conversation History Scorer - optional
Prompt Shield Scorer - optional
Generic Self-Ask Scorer - optional
8. Scorer Metrics
Memory
1. SQLite Memory
2. Basic Memory Programming Usage
3. Memory Types
4. Working with Memory Manually
5. Resending Prompts Using Memory Labels Example
6. Azure SQL Memory
7. PromptSendingAttack with Azure SQL Memory
8. Seed Database Management
9. Exporting Data Example
10. Memory Schema Diagram
Embeddings - optional
Setup
1. Configuration
2. Resiliency and Retry
Default Values
PyRIT Initializers
Auxiliary Attacks
1. Generating GCG Suffixes Using Azure Machine Learning
Scenarios
1. Configuring Scenarios
Registry
Listing Available Classes
Why Instance Registries?
PyRIT Command-Line Frontends
1. PyRIT Scan
2. PyRIT Shell - Interactive Command Line
Deployments
Deploying Hugging Face Models into Azure ML Managed Online Endpoint
Importing and Registering Hugging Face Models into Azure ML
Hugging Face LLMs on Azure ML: Endpoint Interaction Guide
Score Azure ML Managed Online Endpoint
Troubleshooting Guide for HF Azure ML Models
API Reference
Blog
When External Data Becomes a Trojan Horse
A More Generalized OpenAIChatTarget
Datasets and Seed Prompts
Proxying PyRIT
Using PyRIT as a Bug Hunter
Multi-Turn orchestrators
Repository
Suggest edit
Open issue
.rst
.pdf
pyrit.models.SeedType
Contents
SeedType
pyrit.models.SeedType
#
SeedType
#
alias of
Literal
[‘prompt’, ‘objective’, ‘simulated_conversation’]
Contents
SeedType
