feat: ollama integration

[gamedevCloudy]

Integrated Ollama to run the repo locally.

Usage

Run ollama list in the terminal to get a list of models you have:

Running the repo:

• Add prefix ollama: to the model name you want to use via ollama like this ollama:model_name
• Example:

python ai_lab_repo.py --llm-backend "ollama:deepseek-r1:1.5b" --research-topic "YOUR RESEARCH IDEA"

Changes Made:

• ollama integration in inference.py
• moved token and cost computation to a separate function, please let me know if this is not intended.
• updated README and requirements likewise.

Notes:

• I have skipped token computation for models run via ollama. This is becuase users can have multiple models installed and can have different encodings. This may not be supported by tiktoken.
  Please let me know if I am mistaken here. I will make changes accordingly. Thanks.

[LigetiCJ]

The only issue I've ran into is you still need to provide an API key. Please add a bypass for the API key and then you're golden!

[gamedevCloudy]

The only issue I've ran into is you still need to provide an API key. Please add a bypass for the API key and then you're golden!

I'll add a bypass.
I had been testing with OpenAI's key to ensure I did not break the existing implementation, that's why I didn't notice this.

Thanks for pointing it out - I'll make this change.

[gamedevCloudy]

patch: Bypassed API Key check when using Ollama.