Generate a response from the LLM.

Formats the documents for better readability. Args: docs (list): List of Document objects. Returns:

JSON dumps with safe encoding for SSE events.

Generate a streaming response from the LLM. Returns (generator, thread) where generator yields text chunks.

Return the path to the .env file used by this server.

Split text into chunks, respecting paragraph boundaries. Args: text: The text to split Retu

* Shared helper: store messages in the database after a chat interaction.

Initialize the text splitter based on the environment settings.

Handle user interaction with the RAG system.

Streaming version of handle_user_interaction. Yields SSE-formatted events for each pipeline step and then streams the LLM response to

Initialize the Language Model based on environment configurations.

Initialize the embeddings based on the CPU/GPU configuration.

Loads data from various file types and chunks it into an ensemble retriever.

Reinitialise the LLM client (and reranker if enabled) after environment variables have been refreshed. This is called from the /confi

Initializes the RAGHelper class and loads environment variables.

Initialize the ParagraphChunker. Args: max_chunk_size: The target maximum size for each chunk (can be exceeded f

Handle chat interactions with the RAG system. This endpoint processes the user's prompt, retrieves relevant documents, and returns the a

Streaming version of the chat endpoint using Server-Sent Events (SSE). Streams pipeline steps and LLM tokens as they are generated.

Delete a specific document from the data directory and the Milvus vector store. This endpoint expects a JSON payload containing the filename

Return all key-value pairs from the .env file.

Retrieve a specific document from the data directory. This endpoint expects a JSON payload containing the filename of the document to retrie