[FEATURE] Finalize Ollama model provider integration

[og-hayden]

Problem Statement

legion/providers/ollama.py should be updated to ensure the four primary modes of Agent operation are working with models running with Ollama, those being:

1. Text in -> text out
2. Text in -> text out (with tool calls)
3. Text in -> JSON out
4. Text in -> JSON out (with tool calls)

Proposed Solution

1. Use the current OpenAI provider as a starting point, as that is working correctly. The primary difference between the OpenAI provider and the Ollama provider is purely expectations of their respective APIs, so refer to the [Ollama API docs] (https://github.com/ollama/ollama-python).
2. Create a new test file in tests/providers modeled after tests/providers/test_openai.py with integration tests for the Ollama model provider.
3. Write a script defining an agent using an Ollama-supported model and testing it to ensure it's working within the Legion component system.

Alternative Solutions

n/a

Additional Context

Sometimes model providers have their own API docs, but also offer a different set of endpoints that are OpenAI compatible. If that is the case, then consider using those endpoints as OpenAI's API is the most simple and straightforward to use.

The Ollama python library was designed to be very similar to OpenAI's, so there should be limited differences.

Implementation Ideas

Benefits

Serves Legion's goal of being provider agnostic.

Potential Challenges

This ticket does require having Ollama installed on your system, as well as some hardware requirements for running a minimum of a 7b parameter model (16GB+ ram recommended). For installation instructions, visit [Ollama Install Instructions] (https://ollama.com/)

[djnatronic]

I've got Ollama and i've connected to it using Node

Are you thinking something like this but python?

class ModelService {
constructor(model = 'llama3:8b') {
this.model = model;
this.baseUrl = 'http://127.0.0.1:11434/api';
this.port = 3007;
this.app = express();
this.setupLogging();
this.setupRoutes();
}

setupLogging() {
    this.logger = winston.createLogger({
        level: 'info',
        format: winston.format.json(),
        transports: [
            new winston.transports.Console(),
            new winston.transports.File({ filename: 'model-service.log' })
        ]
    });
}

setupRoutes() {
    this.app.use(express.json());
    
    this.app.post('/generate', async (req, res) => {
        try {
            const { prompt, params = {} } = req.body;
            const response = await this.generate(prompt, params);
            res.json(response);
        } catch (error) {
            this.logger.error('Generation error:', error);
            res.status(500).json({ error: error.message });
        }
    });

[Tanner-Perham]

I'm going to take this on!

I've got the same suite of tests as the OpenAI Provider passing for the Ollama Provider now. Just implementing an example agent and troubleshooting.

[Tanner-Perham]

Pull Request now available for review!