Parse the json output of the LLM for summarization in a more structured manner

[Meteord]

Is your feature request related to a problem? Please describe.
The summarization code uses a lot of error prone custom logik, to parse the result of the LLM to a custom json model.
Describe the solution you'd like
Use langchains features to parse the output to a given class. Utilize_json-mode_ if the LLM supports that.

See:

• summarization code: https://github.com/it-at-m/mucgpt/blob/main/app/backend/summarize/summarize.py
• langchain structrured output: https://python.langchain.com/v0.1/docs/modules/model_io/chat/structured_output/

[Meteord]

Done with d27814d