feat: add LiteLLM as provider for 100+ LLM backends

[RheagalFire]

cc @msoedov

Summary

Adds LiteLLM as a provider, enabling vulnerability scanning against any of LiteLLM's 100+ supported LLM backends (OpenAI, Anthropic, Google, Groq, Together AI, AWS Bedrock, Azure, Mistral, etc.) through a single provider class.

Changes

File	What
agentic_security/llm_providers/litellm_provider.py	New LiteLLMProvider(BaseLLMProvider) with async/sync generate() and chat(). Uses drop_params=True for cross-provider compat. Maps litellm.exceptions.RateLimitError to LLMRateLimitError.
agentic_security/llm_providers/factory.py	Register "litellm" in _PROVIDERS dict.
agentic_security/llm_providers/__init__.py	Import and export LiteLLMProvider.
tests/unit/llm_providers/test_litellm_provider.py	29 unit tests covering init, kwargs, sync/async generate/chat, error handling, factory registration.

Tests

LiteLLM provider tests -- 29 passed:

tests/unit/llm_providers/test_litellm_provider.py::TestLiteLLMProviderInit          5 passed
tests/unit/llm_providers/test_litellm_provider.py::TestLiteLLMProviderCallKwargs    6 passed
tests/unit/llm_providers/test_litellm_provider.py::TestLiteLLMProviderMethods       5 passed
tests/unit/llm_providers/test_litellm_provider.py::TestLiteLLMProviderSync          3 passed
tests/unit/llm_providers/test_litellm_provider.py::TestLiteLLMProviderAsync         2 passed
tests/unit/llm_providers/test_litellm_provider.py::TestLiteLLMProviderErrors        5 passed
tests/unit/llm_providers/test_litellm_provider.py::TestLiteLLMProviderFactory       3 passed
29 passed in 1.04s

Full unit suite -- 317 passed, 0 failures:

$ python -m pytest tests/unit/ -q
317 passed in 16.30s

Lint:

$ ruff check agentic_security/llm_providers/litellm_provider.py tests/unit/llm_providers/test_litellm_provider.py
All checks passed!

Edge cases covered:

• drop_params=True always set
• API key/base forwarded when set, omitted when None (env var fallback)
• Null content returns empty string
• No usage returns None
• Rate limit error maps to LLMRateLimitError
• Auth error, timeout, model-not-found map to LLMProviderError
• Factory creates provider by name, lists it, uses default model
• System prompt forwarded correctly

Example usage

from agentic_security.llm_providers import create_provider

# Via factory
provider = create_provider(
    "litellm",
    model="anthropic/claude-sonnet-4-6",
    # api_key omitted -- reads from ANTHROPIC_API_KEY env var
)

# Direct instantiation
from agentic_security.llm_providers import LiteLLMProvider

provider = LiteLLMProvider(
    model="groq/llama-3.3-70b-versatile",
    api_key="gsk-...",
)

# Async
response = await provider.generate("Test prompt for vulnerability scan")

# Sync
response = provider.sync_generate("Test prompt for vulnerability scan")

Any LiteLLM model string works. See LiteLLM provider docs for the full list.

[Copilot]

Pull request overview

Adds a new LiteLLM-backed LLM provider intended to route scans through many LiteLLM-supported backends via the existing provider abstraction and factory.

Changes:

• Introduces LiteLLMProvider with sync/async generate and chat implementations.
• Registers and exports the provider through the LLM provider package and factory.
• Adds unit tests for initialization, request kwargs, response parsing, errors, and factory behavior.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

File	Description
agentic_security/llm_providers/litellm_provider.py	Adds the LiteLLM provider implementation.
agentic_security/llm_providers/factory.py	Registers litellm as a built-in provider.
agentic_security/llm_providers/__init__.py	Exports LiteLLMProvider from the provider package.
tests/unit/llm_providers/test_litellm_provider.py	Adds unit coverage for the new provider.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.