Fix: LLM API Key Error Handling (#21)

* Fix 'NoneType' Object Error in Tool Loading

## What Changed
1. **Local Tool Loading**: Added logic to check and load tools from the local directory first.
2. **Fixed NoneType Error**: Resolved the issue where the version was `None`, causing a `'NoneType' object has no attribute 'replace'` error.
3. **Improved Error Messages**: Enhanced error messages for better clarity and debugging.

## How It Works Now
- **Local First**: The system first checks the local tools folder.
- **Use Local Tool**: If a tool is found locally, it uses that version.
- **Remote Fallback**: If not found locally, it downloads the tool.
- **Version Handling**: Retrieves version from local `config.json` to ensure correct versioning.

## Testing
- Tested with `math_agent` using local `wikipedia` tool.
- Tested with `story_teller` agent.
- Confirmed no NoneType errors occur.

* Fix 'NoneType' Object Error in Tool Loading

## What Changed
1. **Local Tool Loading**: Added logic to check and load tools from the local directory first.
2. **Fixed NoneType Error**: Resolved the issue where the version was `None`, causing a `'NoneType' object has no attribute 'replace'` error.
3. **Improved Error Messages**: Enhanced error messages for better clarity and debugging.

## How It Works Now
- **Local First**: The system first checks the local tools folder.
- **Use Local Tool**: If a tool is found locally, it uses that version.
- **Remote Fallback**: If not found locally, it downloads the tool.
- **Version Handling**: Retrieves version from local `config.json` to ensure correct versioning.

* Fix 'NoneType' Object Error in Tool Loading

## What Changed
1. **Local Tool Loading**: Added logic to check and load tools from the local directory first.
2. **Fixed NoneType Error**: Resolved the issue where the version was `None`, causing a `'NoneType' object has no attribute 'replace'` error.
3. **Improved Error Messages**: Enhanced error messages for better clarity and debugging.

## How It Works Now
- **Local First**: The system first checks the local tools folder.
- **Use Local Tool**: If a tool is found locally, it uses that version.
- **Remote Fallback**: If not found locally, it downloads the tool.
- **Version Handling**: Retrieves version from local `config.json` to ensure correct versioning.

* Fix 'NoneType' Object Error in Tool Loading

## What Changed
1. **Local Tool Loading**: Added logic to check and load tools from the local directory first.
2. **Fixed NoneType Error**: Resolved the issue where the version was `None`, causing a `'NoneType' object has no attribute 'replace'` error.
3. **Improved Error Messages**: Enhanced error messages for better clarity and debugging.

## How It Works Now
- **Local First**: The system first checks the local tools folder.
- **Use Local Tool**: If a tool is found locally, it uses that version.
- **Remote Fallback**: If not found locally, it downloads the tool.
- **Version Handling**: Retrieves version from local `config.json` to ensure correct versioning.

* Fix

* Create python-package-conda.yml

* Update python-package-conda.yml

* Update python-package-conda.yml

* Update python-package-conda.yml

* Update python-package-conda.yml

* Update python-package-conda.yml

* Update python-package-conda.yml

* Update python-package-conda.yml

* Update python-package-conda.yml

* fix: improve agent loading and error handling

- Add debug logs to track agent loading
- Fix config handling in agent manager
- Update seeact_demo_agent config format

The code now:
- Shows better error messages
- Loads agents more reliably
- Keeps full config when packaging"

* Add Local Agent Support to run-agent

## What's Changed
- Added `--local_agent` flag to support loading agents from local filesystem
- Added path validation for local agent loading
- Simplified agent execution logic

* Refactor Configuration Management for URLs

This PR refactors the configuration management for Cerebrum by moving the URLs for the Agent and Tool Managers into a centralized configuration file (default.yaml). This change allows for easier management and customization of these URLs across different environments.
Changes:
Added agent_hub_url and tool_hub_url to default.yaml.
2. Updated AutoAgent and AutoTool classes to use URLs from the configuration file.

* Revert "Add Local Agent Support to run-agent"

This reverts commit 1324ed1.

* Reapply "Add Local Agent Support to run-agent"

* Revert "Refactor Configuration Management for URLs"

This reverts commit 53fd503.

* Update run_agent.py

* Update run_agent.py

* fix: LLM API Key Error Handling

## What's Changed
- Added proper error handling for invalid API keys
- Fixed response parsing for LiteLLM completion calls
- Added HTTP status codes for different error types:
  - 402 for API key issues
  - 500 for other errors

* Update

cerebrum/llm/communication.py

@@ -31,13 +31,17 @@ class Response(BaseModel):
     Response class represents the output structure after performing actions.
     
     Attributes:
-        response_message (Optional[str]): The generated response message. Default is None.
-        tool_calls (Optional[List[Dict[str, Any]]]): An optional list of JSON-like objects (dictionaries) 
-            representing the tool calls made during processing. Default is None.
+        response_message (Optional[str]): The generated response message.
+        tool_calls (Optional[List[Dict[str, Any]]]): Tool calls made during processing.
+        finished (bool): Whether the processing is complete.
+        error (Optional[str]): Error message if any.
+        status_code (int): HTTP status code of the response.
     """
-    response_message: Optional[str] = None  # The generated response message, default is None.
-    tool_calls: Optional[List[Dict[str, Any]]] = None  # List of JSON-like objects representing tool calls, default is None.
-    finished: bool
+    response_message: Optional[str] = None
+    tool_calls: Optional[List[Dict[str, Any]]] = None
+    finished: bool = False
+    error: Optional[str] = None
+    status_code: int = 200
 
     class Config:
-        arbitrary_types_allowed = True  # Allows arbitrary types in validation.
+        arbitrary_types_allowed = True