In this article, we will explore how Flow Wise can be used to load multiple internal documents or sources. Flow Wise is an open-source platform for building question-answering bots. It provides various options for selecting agents and types of chains based on your requirements.
To load multiple internal documents, Flow Wise offers different document loaders such as API loader, web scraper, docx file loader, and PDF file loader. These loaders allow you to fetch data from various sources and formats.
For embeddings, Flow Wise supports OpenAI embeddings and other available embeddings in the market. It also provides options for language models like OpenAI’s GPT-3, Azure, and Google Vertex AI.
To handle text chunking, Flow Wise offers different splitters such as HTML to markdown text splitter for web pages and recursive character text splitter for documents.
Flow Wise also provides tools for score retrievers, prompts, and store retrievers. These tools help in retrieving data from vector stores and connecting them to the appropriate chains.
In this article, we demonstrated how to load two documents, one in docx format and another in PDF format, and connect them to a Pine Cone Vector store. We created an index on Pine Cone and connected the document loaders to this index. We then used a vector store retriever to retrieve data from the vector store.
Finally, we connected the vector store retriever to a multi-retriever QA chain, which automatically selects the appropriate vector store based on the question asked. The QA chain is connected to a language model, such as ChatGPT, for answering questions.
While Flow Wise provides a powerful framework for loading and processing multiple internal documents, it may require further customization and fine-tuning for specific use cases. Overall, Flow Wise offers a flexible and efficient solution for handling document loading and retrieval in question-answering systems.
