added ollama support

[codingmoh]

🚀 Add Ollama Support to Open Codex

Summary

This PR introduces native support for Ollama as an alternative LLM backend alongside the existing Phi-4-mini integration.

✨ Key Changes

• New Agent: OllamaAgent implemented using the official ollama Python client:

  • Supports system_prompt, configurable model_name, and temperature
  • Validates model presence and Ollama server availability
  • Uses ollama.chat with a one-shot response interface

• AgentBuilder Enhancements:

  • get_ollama_agent() for easy instantiation
  • get_system_prompt() extracted for reuse

• CLI Extensions (main.py):

  • Added --ollama flag to enable Ollama mode
  • Added --model flag to specify the model (e.g., llama3)
  • Updated one_shot_mode() to dynamically select the agent backend
  • Improved CLI help messages for clarity

• Refactored phi_4_mini.py:

  • Moved download_model() below __init__ for better structure (functionality unchanged)

🔪 Example Usage

# Use default Phi-4-mini
open-codex "list files in this directory"

# Use Ollama with default model
open-codex --ollama "summarize this Python file"

# Use Ollama with a specific model
open-codex --ollama --model llama3 "explain what this shell script does"

📝 Notes

• Ollama must be running locally (http://localhost:11434)
• Models must be preloaded via ollama pull <model_name>

[Copilot]

Pull Request Overview

This PR adds native support for Ollama as an LLM backend alongside the existing Phi-4-mini integration. Key changes include:

• Introducing a new OllamaAgent to interface with the official Ollama Python client.
• Enhancing the AgentBuilder and CLI to support an --ollama flag and a configurable model name.
• Refactoring the prompt reading and repositioning the download_model function in phi_4_mini.py.

Reviewed Changes

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

File	Description
src/open_codex/main.py	Added argparse enhancements for choosing agent backend and CLI improvements.
src/open_codex/agents/phi_4_mini.py	Repositioned and re-added download_model to improve code structure.
src/open_codex/agents/ollama_restapi.py	Introduces the new OllamaAgent with proper error handling and API integration.
src/open_codex/agent_builder.py	Added get_system_prompt and get_ollama_agent for improved agent instantiation.

[codingmoh]

The host URL for Ollama is hardcoded as http://localhost:11434.!

[codewithdark-git]

The issue occurs when I try to load the Ollama model without running Ollama locally. When I run the command open-codex --ollama --model llama3 "create a tarball of the src dir", it loads the default model and checks if Ollama is running.

[codingmoh]

thanks @codewithdark-git - I'll check

[codingmoh]

Thanks again! - I added proper error handling and merged it