Add LiteLLM provider for pentestgpt-legacy

[RheagalFire]

Add LiteLLM provider for pentestgpt-legacy

LiteLLM is an open-source gateway that provides a unified completion() interface for 100+ LLM providers. Adding it as a provider gives pentestgpt-legacy users access to Azure, AWS Bedrock, Google Vertex, HuggingFace, vLLM, and any OpenAI-compatible endpoint, all through a single litellm:<model> prefix, without per-provider code.

Changes

File	Change
llm/providers/litellm_provider.py	New LiteLLMProvider(BaseProvider) using litellm.acompletion() with drop_params=True
llm/registry.py	Added litellm ProviderInfo + litellm: dynamic prefix resolver
llm/factory.py	Registered LiteLLMProvider in _PROVIDER_CLASSES
llm/providers/__init__.py	Export LiteLLMProvider
llm/config.py	Added LITELLM_API_KEY and LITELLM_BASE_URL settings
pyproject.toml	Added litellm>=1.80.0 as optional dependency
.env.example	Documented LiteLLM configuration
tests/legacy/test_litellm_provider.py	6 unit tests covering registry, resolve, factory, and proxy config

Usage

# Install with the litellm extra
uv sync --extra litellm

# Use any LiteLLM-supported model
pentestgpt-legacy --reasoning-model litellm:azure/gpt-4 --parsing-model litellm:bedrock/anthropic.claude-v2

# With a LiteLLM proxy
LITELLM_API_KEY=sk-proxy LITELLM_BASE_URL=http://proxy:4000 \
  pentestgpt-legacy --reasoning-model litellm:gpt-4

Tests

Live E2E - full LLMClient stack via Azure Foundry (anthropic/claude-sonnet-4-6):

from pentestgpt_legacy.llm.factory import get_client

# Factory resolves litellm: prefix -> LiteLLMProvider -> litellm.acompletion()
client = get_client('litellm:anthropic/claude-sonnet-4-6')

# Test 1: New conversation
response, cid = client.send_new_message('Say hello in 5 words.')
print(f"Response: {response}")
print(f"Conversation ID: {cid}")
# Response: Hello there, how are you?
# Conversation ID: <uuid>

# Test 2: Multi-turn (same conversation)
response2 = client.send_message('Now say goodbye in 3 words.', cid)
print(f"Follow-up: {response2}")
# Follow-up: Goodbye for now!

Full chain verified: get_client('litellm:...') -> LLMClientFactory -> LiteLLMProvider.acompletion() -> litellm.acompletion(drop_params=True) -> Azure Foundry -> response parsed correctly. Multi-turn conversation state maintained across calls.

[gemini-code-assist]

Code Review

This pull request introduces integration with LiteLLM as an optional multi-provider gateway, adding configuration options, a dedicated LiteLLMProvider class, and corresponding unit tests. The feedback suggests two improvements: updating the installation error message to recommend installing the optional dependency via 'uv sync --extra litellm' instead of 'uv add litellm', and safely guarding against an empty response.choices list to prevent a potential IndexError.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

The error message suggests installing litellm using uv add litellm. However, litellm is defined as an optional dependency (extra) in pyproject.toml. Running uv add litellm would add it as a direct dependency. To keep it as an optional dependency, the error message should recommend using uv sync --extra litellm (consistent with the instructions in .env.example).

Suggested change

	raise RuntimeError(
	"The 'litellm' package is required. Install with: uv add litellm"
	) from exc
	raise RuntimeError(
	"The 'litellm' package is required. Install with: uv sync --extra litellm"
	) from exc

[gemini-code-assist]

If the API response is empty or blocked (e.g., due to content filtering), response.choices might be empty. Accessing response.choices[0] directly without checking can raise an IndexError. It is safer to guard against empty choices.

Suggested change

	response = await litellm.acompletion(**kwargs)
	return response.choices[0].message.content or ""
	response = await litellm.acompletion(**kwargs)
	if not response.choices:
	return ""
	return response.choices[0].message.content or ""