feat: add LiteLLM as LLM provider

[RheagalFire]

Summary

• Adds LiteLLM as a new model provider in ask_llm_anything(), enabling access to 100+ LLM providers (OpenAI, Anthropic, Google, Azure, Bedrock, Ollama, etc.) via a single unified SDK
• When model_provider="litellm", routes through litellm.completion() with drop_params=True instead of the OpenAI client
• No proxy server needed, uses litellm SDK directly

Changes

• tools/ask_llm_v2.py - added import litellm at top level, added litellm code path in ask_llm_anything(): when model_provider == "litellm", calls litellm.completion() with drop_params=True
• model_config.yaml - added litellm provider entry (uses "EMPTY" placeholder following repo convention, litellm falls back to provider env vars)
• requirements.txt - added litellm>=1.60.0,<2.0.0
• tests/test_litellm_provider.py - 12 unit tests (all passing)

Proof of work

Unit tests (12/12 passing):

$ python -m pytest tests/test_litellm_provider.py -v
test_litellm_branch_exists PASSED
test_uses_drop_params_true PASSED
test_uses_litellm_completion PASSED
test_imports_litellm_at_top PASSED                                                                                                                                                                                 
test_skips_empty_api_key PASSED
test_skips_empty_api_base PASSED                                                                                                                                                                                   
test_litellm_in_config PASSED
test_litellm_config_has_api_base PASSED
test_litellm_config_has_api_key PASSED                                                                                                                                                                             
test_completion_called_with_drop_params PASSED
test_completion_forwards_api_key PASSED                                                                                                                                                                            
test_completion_response_has_content PASSED
12 passed in 0.04s                                                                                                                                                                                                 

Live E2E (Azure Foundry, Claude):

Model: anthropic/claude-sonnet-4-6                                                                                                                                                                                 
Response: "4"   
Tokens: prompt=20, completion=5
E2E SUCCESS

Example usage

# model_config.yaml - set your key or leave EMPTY to use env vars
litellm:                                                                                                                                                                                                           
    api_base: "EMPTY"
    api_key: "EMPTY"                                                                                                                                                                                               

from tools.ask_llm_v2 import ask_llm_anything

# LiteLLM reads provider env vars automatically
# (ANTHROPIC_API_KEY, OPENAI_API_KEY, etc.)
response = ask_llm_anything(                                                                                                                                                                                       
    "litellm",
    "anthropic/claude-sonnet-4-20250514",                                                                                                                                                                          
    [{"role": "user", "content": "Describe the GUI elements on screen."}],
)                                                                                                                                                                                                                  

See https://docs.litellm.ai/docs/providers for 100+ supported model strings.

Impact

• Additive only, existing providers (local, stepfun) untouched
• litellm added to requirements.txt as a dependency
• drop_params=True silently drops provider-unsupported kwargs
• "EMPTY" api_key/api_base values are skipped so litellm falls back to provider env vars

[RheagalFire]

cc @softsweetengineer

[RheagalFire]

@softsweetengineer do you have any update on this PR?