fix: resolve --generate_autodan CLI feature bugs

[sinhaabhiraj0]

Fixes #1520

This PR fixes the broken `--generate_autodan` CLI feature by updating the AutoDAN code to match the current generator API.

Changes

cli.py

• Use `parse_cli_plugin_config("probe", args)` to properly parse `--probe_options` instead of referencing non-existent `_config.probe_options`
• Add `SystemExit(1)` after error message to prevent execution with undefined variables

autodan.py

• Import `Conversation`, `Turn`, `Message` from `garak.attempt`
• Wrap string prompt in `Conversation` object to match updated `generator.generate()` API

genetic.py

• Import `Conversation`, `Turn`, `Message` from `garak.attempt`
• Convert old OpenAI-style dict format to `Conversation` object for mutation generator calls

Test plan

• Run `python -m garak --target_type huggingface.Model --target_name gpt2 --generate_autodan --probe_options '{"prompt": "Test prompt", "target": "Test target"}'`
• Verify AutoDAN generation starts without errors
• Verify mutation generator (gpt-3.5-turbo) works correctly"

[jmartin-tech]

This is a great find, this cli option is very much an edge case that needs some attention.

I have added some specific technical concern, and also offered that there may be feature or use case decision to be made here.

@erickgalinkin may want to weight in on the suggestions I have made on the cli.py changes. I suspect this cli option should be deprecated.

If this option is retained additional documentation of the configuration values and patterns for setting then are needed.

[jmartin-tech]

Suggested change

	raise SystemExit(1)
	raise

The exception handler should simply re-raise.

[jmartin-tech]

generator.generate must return a list[Message|None| this mean if this is not Message type it will be NoneType and casting that as a string is not a useful action. Another possible condition is for gen_output.text to be None.

Consider:

Suggested change

	gen_str = gen_output.text if isinstance(gen_output, Message) else str(gen_output)
	success = check_for_attack_success(gen_str, REJECTION_STRINGS)
	success = False
	if gen_output and gen_output.text:
	success = check_for_attack_success(gen_output.text, REJECTION_STRINGS)

[jmartin-tech]

Similar to other comment:

Suggested change

	response = mutation_generator.generate(prompt=conv)
	response_text = response[0].text if isinstance(response[0], Message) else str(response[0])
	revised_sentence = response_text.replace("\n", "")
	received = True
	response = mutation_generator.generate(prompt=conv)[0]
	if response and response.text:
	revised_sentence = response.text.replace("\n", "")
	received = True

[jmartin-tech]

This is inconsistent with other cli options, the help details mention prompt_options and the exception below mentions probe_options, this PR needs to expanded to ensure consistent messaging.

This also looks like something of a divergence from how configuration is done generally in the tooling as this is only possible as a cli option as coded.

At a minimum rely on the general configuration object that has already been processed and merged with file based configuration:

Suggested change

	probe_options = parse_cli_plugin_config("probe", args)
	if probe_options is None:
	raise ValueError("probe_options is None")
	prompt = probe_options["prompt"]
	target = probe_options["target"]
	probe_options = config_plugin_type.get("probe", None)
	if probe_options is None:
	raise ValueError("probe_options is None")
	prompt = probe_options["prompt"]
	target = probe_options["target"]

As an alternative maybe the generate_autodan option should be remove and the expectation should be adjusted to execute the dan.AutoDAN with exposed DEFAULT_PARAMS for goal_str and target which map to prompt and target here.