Chat UI doesn't respect custom model/endpoint configuration

[eyalsela]

(Issue descrption generated by Claude)

Currently, there's an inconsistency between how model configuration is handled in the main DAILA operations versus the sidebar chat UI.

Current Behavior

• Main DAILA operations (like function summarization, variable renaming, etc.) properly use custom model and endpoint configurations from DAILAConfig
• The sidebar chat UI bypasses these configurations and directly uses litellm with just the model name, ignoring any custom endpoint settings

Expected Behavior

The sidebar chat should respect the same configuration as the main DAILA operations, including:

• Custom endpoints
• Custom model configurations
• Any other LiteLLM-related settings from DAILAConfig

Technical Details

The issue is in dailalib/llm_chat/llm_chat_ui.py where LLMThread.run() directly uses litellm without accessing the configuration from the parent LiteLLMAIAPI instance:

def run(self):
    import litellm
    litellm.modify_params = True
    response = litellm.completion(
        model=self.model_name,
        messages=self.chat_history,
        timeout=60,
    )

Suggested Fix

Modify the chat UI to use the same configuration mechanism as the main DAILA operations by:

Passing the full configuration from LiteLLMAIAPI to the chat UI
Using the same completion call pattern in LLMThread as used in the main operations
Ensuring configuration changes are immediately reflected in both the main operations and chat UI

[mahaloz]

Hey @eyalsela, what OS and what Decompiler are you on?

[eyalsela]

Windows, ida