ProbChain Hashnode blog

Flow

app.py

• Initialize the LLMChain with the specified local LLM model file and configuration.
• Initialize the PromptTemplate and HuggingFaceBgeEmbeddings with the specified template and model details.
• Initialize the Chroma vector store and RetrievalQA with the specified vector store and retriever.
• Perform semantic search using the retriever and the user's query.
• Initialize the Gradio interface with the specified input and output settings.
• Launch the Gradio interface. Outputs
• The semantic search results.
• The response from the question-answering bot in the Gradio interface.

Outputs

• The semantic search results.

Flow

ingest.py

• The code imports necessary modules and classes.
• It defines the model name, model kwargs, and encode kwargs for creating the HuggingFaceBgeEmbeddings instance.
• The code creates the embeddings using the specified model and kwargs.
• It initializes the PyPDFLoader with the PDF file name and loads the documents.
• The RecursiveCharacterTextSplitter is instantiated with the chunk size and overlap.
• The code splits the documents into smaller chunks of text using the text splitter.
• The Chroma algorithm is used to create a vector store from the text chunks and embeddings.
• The vector store is created with the specified collection metadata and persist directory.
• The code prints a message indicating that the vector store has been created.