Detail backend (api) for qa on text files

[mt7180]

• write TextDocument class to load text files and convert them into Llama-nodes
• write Custom ChatEngine class to use and customize VectorindexRetriever, load document nodes into it
• test main functionality in script with if name == "main"

[mt7180]

Implemented code for the Document and ChatEngine classes in script.py, functionality is tested in if _name_ == "_main_"

[mt7180]

I implemented LlamaIndex and created a CondensedQuestionChatEngine in combination with a RetrieverQueryEngine, which efficiently retrieve relevant context from a knowledge base when given a query. The RetrieverQueryEngine is based on a VectorIndecRetriever to retrieve the relevant nodes of a VectorIndex. The nodes are parsed with marvin ai_model as metadataExtractor, the metadata is used by the retriever to find the relevant nodes.

In general the Concepts "Decoupling Chunks Used for Retrieval" and "Chunks Used for Synthesis" are both used here:

• The CondenseQuestionChatEngine is responsible for synthesis
• while the RetrieverQueryEngine handles retrieval.
  This allows for more efficient and accurate retrieval of relevant documents before retrieving the specific chunks needed for synthesis.

[Zaubeerer]

Great! @mt7180, you nailed the issue commenting game :)

Such comments can be made either in the PR or the issue, depending on the company's preferences.