chore: Add TAP attack example with Azure OpenAI custom target

[rdheekonda]

Summary

Adds a new example notebook demonstrating TAP (Tree of Attacks with Pruning) attack against Azure OpenAI endpoints using CustomTarget.

Changes

• New notebook: examples/airt/tap_azure_openai_custom_target.ipynb
• Demonstrates wrapping Azure OpenAI API as a custom attack target
• Includes character join transform for obfuscation testing
• Uses Groq Llama 4 for attacker/evaluator models (LiteLLM supported)
• Adds Azure Foundry configuration instructions

Configuration

• Early stopping: 0.75 score
• Max trials: 50
• Transform: Character join with underscores
• Attack goal: Educational keylogger generation

Testing

Notebook tested successfully with Azure OpenAI gpt-5.2-chat deployment.

Generated Summary:

• Introduced a new Jupyter Notebook: tap_azure_openai_custom_target.ipynb.
• Demonstrates executing a TAP (Tree of Attacks with Pruning) attack using Azure OpenAI as a custom target.
• Includes configuration details for Azure OpenAI and Dreadnode.
• Features a step-by-step guide for querying Azure OpenAI endpoint, configuring the TAP attack, running the attack, and analyzing the results.
• Attack goal is to generate Python keylogger code for educational purposes with specific transformations applied.
• Throughout the notebook, code snippets are provided for practical implementation.
• Result output includes performance metrics of the TAP trials.

This summary was generated with ❤️ by rigging