[examples] Finalise the Ollama example application

[eolivelli]

Summary:

• use OpenAI Embeddings, they perform better
• switch to llama2:13b
• make the prompt more effective for the llama2:13b model