Proposed Modification for Dynamically Managing Models in Langchain::LLM::Ollama

[Jhonnyr97]

Refactor of Langchain::LLM::Ollama

This refactoring was inspired by the implementation of Langchain::LLM::OpenAI, particularly in terms of parameter usage.

Solved Problems

• Exception Generated by llm.embed: Previously, invoking llm.embed(text: "hello world") caused a Faraday::ResourceNotFound type exception. This problem has now been identified and resolved.
• Non-Existent Streaming Support: Previously, streaming was not actually supported. The current modification introduces support for streaming.
• Parameter Management in the complete and embed Methods: The complete and embed methods did not offer a simple management of parameters. This pull request introduces the allowed parameters, in line with the specifications of the documentation.

[andreibondarev]

Does the embed endpoint really take the temperature parameter?

[Jhonnyr97]

It seems strange to me too but in the documentation it indicates that you can enter the parameters you find in the Modelfile

[andreibondarev]

Hmm... seems to be the case.

[andreibondarev]

@Jhonnyr97 Great job, thank you!