adding ollama

[ohdearquant]

PR Description

This PR adds Ollama as a new provider integration, enabling local or self-hosted LLM usage under the same consistent interface as other providers. Below is a brief summary of changes:

1. Ollama Integration

   • Introduced a new OllamaChatCompletionEndPoint class to handle local chat completions from an Ollama server.
   • Added logic to pull or list local models from Ollama. If a requested model isn’t installed, the system automatically downloads it.
   • The new endpoint is registered under match_endpoint.py so that specifying provider="ollama" routes requests to the Ollama endpoint.

2. Configuration & Key Handling

   • Updated IMODEL constructor logic to accept provider="ollama" and skip the usual API-key verification (since Ollama is local).
   • Ensures simpler usage patterns, e.g. model="ollama/mymodel" or provider="ollama", model="myModelName".

3. Schema & Endpoint Adjustments

   • Minor refactoring in function_to_schema.py for better handling of function calls and JSON schema creation.
   • Several improvements in base.py and imodel.py to unify endpoint initialization logic, so that the same base structures can handle remote or local providers.

4. Version Bump

   • Bumped project version from 0.9.5 to 0.9.6 to reflect the new provider support.

Why Ollama?
Ollama provides a local LLM runtime for secure and faster experimentation, removing external dependencies. This integration makes it easier to switch between remote (OpenAI, etc.) and local LLMs while reusing the same Lion-AGI code patterns.

Testing

• Manually tested model pulls and local completions with ollama:latest.
• Verified that the fallback to other providers is unchanged.

Impacts

• No breaking changes for existing code.
• A new ollama provider is now recognized, letting users specify {"provider": "ollama", "model": "<modelName>"} to run local inference.

Thanks for reviewing!