feat: Add Ollama integration and improve CLI usability and documentation

- Added full support for Ollama via official Python client (`ollama>=0.4.8`)
- New agent class `OllamaAgent` for local LLMs like LLaMA3 via Ollama
- Refactored `AgentBuilder` to support multiple backends (phi, ollama)
- Improved CLI with:
  - `--ollama` and `--model` flags
  - Host configuration via `--ollama-host`
  - Structured prompt help and error handling
- Updated README with new usage examples, clearer feature list, and Ollama setup instructions
- Renamed `phi_4_mini.py` → `phi_4_mini_agent.py` for consistency
- Enhanced logging and feedback for model loading and errors
- Improved UX: response formatting, keypress actions (copy/execute/abort) with better feedback

Author: codingmoh

README.md

@@ -8,40 +8,56 @@
 
 ---
 
-**Open Codex** is a fully open-source command-line AI assistant inspired by OpenAI Codex, supporting local language models like `phi-4-mini`.
+**Open Codex** is a fully open-source command-line AI assistant inspired by OpenAI Codex, supporting local language models like `phi-4-mini` and **full integration with Ollama**.
 
-No API key is required. Everything runs locally.
+🧠 **Runs 100% locally** – no OpenAI API key required. Everything works offline.
 
-Supports:
-- **One-shot mode**: `open-codex "list all folders"` -> returns shell command
-- 🧠 Local-only execution using supported OS models (currently `phi-4-mini`)
+---
+
+## Supports
+
+* **One-shot mode**: `open-codex "list all folders"` -> returns shell command
+* **Ollama integration** for (e.g., LLaMA3, Mistral)
+* Native execution on **macOS, Linux, and Windows**
 
 ---
 ## ✨ Features
 
-- Natural Language to Shell Command (via local models)
-- Works on macOS, Linux, and Windows (Python-based)
-- Confirmation before execution
-- Add to clipboard / abort / execute prompt
-- One-shot interaction mode (interactive and function-calling coming soon)
+- Natural Language → Shell Command (via local or Ollama-hosted LLMs)
+- Local-only execution: no data sent to the cloud
+- Confirmation before running any command
+- Option to copy to clipboard / abort / execute
 - Colored terminal output for better readability
+- Ollama support: use advanced LLMs with `--ollama --model llama3`
+
+### 🔍 Example with Ollama:
+
+```bash
+open-codex --ollama --model llama3 "find all JPEGs larger than 10MB"
+```
+
+Codex will:
+
+1. Send your prompt to the Ollama API (local server, e.g. on `localhost:11434`)
+2. Return a shell command suggestion (e.g., `find . -name "*.jpg" -size +10M`)
+3. Prompt you to execute, copy, or abort
+
+> 🛠️ You must have [Ollama](https://ollama.com) installed and running locally to use this feature.
 
 ---
 
 ## 🧱 Future Plans
 
-- Interactive, context aware mode
+- Interactive, context-aware mode
 - Fancy TUI with `textual` or `rich`
-- Add support for additional OSS Models
 - Full interactive chat mode
 - Function-calling support
-- Voice input via Whisper
-- Command history and undo
+- Whisper-based voice input
+- Command history & undo
 - Plugin system for workflows
 
 ---
 
-
 ## 📦 Installation
 
 
@@ -53,42 +69,46 @@ brew install open-codex
 ```
 
 
-### 🔹 Option 2: Install via pipx (cross-platform)
+### 🔹 Option 2: Install via pipx (Cross-platform)
 
 ```bash
 pipx install open-codex
 ```
 
-### 🔹 Option 3: Clone & Install locally
+### 🔹 Option 3: Clone & install locally
 
 ```bash
 git clone https://github.com/codingmoh/open-codex.git
 cd open_codex
 pip install .
 ```
 
-
-Once installed, you can use the `open-codex` CLI globally.
+Once installed, use the `open-codex` CLI globally.
 
 ---
 
-## 🚀 Usage
+## 🚀 Usage Examples
 
-### One-shot mode
+### ▶️ One-shot mode
 
 ```bash
 open-codex "untar file abc.tar"
 ```
-
 ✅ Codex suggests a shell command  
 ✅ Asks for confirmation / add to clipboard / abort  
 ✅ Executes if approved
 
+### ▶️ Using Ollama
+
+```bash
+open-codex --ollama --model llama3 "delete all .DS_Store files recursively"
+```
+
 ---
 
 ## 🛡️ Security Notice
 
-All models run locally. Commands are only executed after explicit approval.
+All models run **locally**. Commands are executed **only after your explicit confirmation**.
 
 ---
 
@@ -105,4 +125,3 @@ MIT
 ---
 
 ❤️ Built with love and caffeine by [codingmoh](https://github.com/codingmoh).
-

pyproject.toml

@@ -7,6 +7,7 @@ requires-python = ">=3.11"
 dependencies = [
     "huggingface-hub>=0.30.2",
     "llama-cpp-python>=0.3.8",
+    "ollama>=0.4.8",
     "prompt_toolkit",
     "pyinstaller>=6.13.0",
     "pyperclip>=1.9.0",

src/open_codex/agent_builder.py

@@ -1,15 +1,29 @@
 from importlib.resources import files
 
-from open_codex.agents.phi_4_mini import AgentPhi4Mini
 from open_codex.interfaces.llm_agent import LLMAgent
 
 class AgentBuilder:
+    
+    @staticmethod
+    def get_system_prompt() -> str:
+        return files("open_codex.resources") \
+            .joinpath("prompt.txt") \
+            .read_text(encoding="utf-8")
 
     @staticmethod
-    def get_agent() -> LLMAgent:
-        system_prompt = files("open_codex.resources").joinpath("prompt.txt").read_text(encoding="utf-8")
-        return AgentPhi4Mini(system_prompt=system_prompt)
+    def get_phi_agent() -> LLMAgent:
+        from open_codex.agents.phi_4_mini_agent import Phi4MiniAgent
+        system_prompt = AgentBuilder.get_system_prompt()
+        return Phi4MiniAgent(system_prompt=system_prompt)
 
+    @staticmethod
+    def get_ollama_agent(model: str, host: str) -> LLMAgent:
+        from open_codex.agents.ollama_agent import OllamaAgent
+        system_prompt = AgentBuilder.get_system_prompt()
+        return OllamaAgent(system_prompt=system_prompt, 
+                           model_name=model,
+                           host=host)
+
     @staticmethod
     def read_file(file_path: str) -> str:
         with open(file_path, 'r') as file:

src/open_codex/agents/ollama_agent.py

@@ -0,0 +1,110 @@
+from typing import List, Dict
+import logging
+import ollama
+
+from open_codex.interfaces.llm_agent import LLMAgent
+
+# Configure logger
+logger = logging.getLogger(__name__)
+
+class OllamaAgent(LLMAgent):
+    """
+    Agent that connects to Ollama to access local language models 
+    using the official Python client.
+    """
+    
+    def __init__(self, 
+                 system_prompt: str,
+                 model_name: str,
+                 host: str,
+                 temperature: float = 0.2,
+                 max_tokens: int = 500):
+        """
+        Initialize the Ollama agent.
+        
+        Args:
+            system_prompt: The system prompt to use for generating responses
+            model_name: The name of the model to use (default: "llama3")
+            host: The host URL of the Ollama API (default: None, uses OLLAMA_HOST env var or http://localhost:11434)
+            temperature: The temperature to use for generation (default: 0.2)
+            max_tokens: The maximum number of tokens to generate (default: 500)
+        """
+        self.system_prompt = system_prompt
+        self.model_name = model_name
+        self.host = host
+
+        self.temperature = temperature
+        self.max_tokens = max_tokens
+        self._ollama_client = ollama.Client(host=self.host)
+    
+    def _check_ollama_available(self) -> None:
+        """Check if Ollama server is available and the model exists."""
+        try:
+            # List models to check connection
+            models: ollama.ListResponse = self._ollama_client.list()
+            
+            available_models = [model.model for model in models.models if model.model is not None]
+            
+            if not available_models:
+                logger.error(f"No models found in Ollama. You may need to pull the model with: ollama pull {self.model_name}")
+            elif self.model_name not in available_models:
+                logger.error(f"Model '{self.model_name}' not found in Ollama. Available models: {', '.join(available_models)}")
+                logger.error(f"You can pull the model with: ollama pull {self.model_name}")
+                
+        except ConnectionError as e:
+            logger.error(f"Could not connect to Ollama server.")
+            logger.error(
+                f"Make sure Ollama is running at {self.host} or install it from https://ollama.com"
+            )
+            raise ConnectionError(
+                f"Could not connect to Ollama server. "
+                f"Make sure Ollama is running at {self.host} or install it from https://ollama.com"
+            )
+    
+    def one_shot_mode(self, user_input: str) -> str:
+        """
+        Generate a one-shot response to the user input.
+        
+        Args:
+            user_input: The user's input prompt
+            
+        Returns:
+            The generated response as a string
+        """
+        self._check_ollama_available()
+        messages = [
+            {"role": "system", "content": self.system_prompt},
+            {"role": "user", "content": user_input}
+        ]
+        
+        response = self._generate_completion(messages)
+        return response.strip()
+    
+    def _generate_completion(self, messages: List[Dict[str, str]]) -> str:
+        """
+        Generate a completion using the Ollama API.
+        
+        Args:
+            messages: The conversation history as a list of message dictionaries
+            
+        Returns:
+            The generated text response
+        """
+        try:
+            
+            response = ollama.chat(
+                model=self.model_name,
+                messages=messages,
+                options={
+                    "temperature": self.temperature,
+                    "num_predict": self.max_tokens,
+                }
+            )
+            
+            if "message" in response and "content" in response["message"]:
+                return response["message"]["content"]
+            else:
+                raise ValueError(f"Unexpected response format from Ollama API: {response}")
+                
+        except Exception as e:
+            raise ConnectionError(f"Error communicating with Ollama: {str(e)}")