A chatbot that reads your docs, spreadsheets, and YouTube backlog

What it does

Robby-chatbot is a Streamlit interface that turns documents and YouTube videos into chat partners. It ingests PDFs, CSVs, and text files, indexes them with FAISS vector embeddings, and answers questions with conversational memory across sessions. A separate mode uses PandasAI to let you query tabular data in plain English, while another mode summarizes YouTube videos via OpenAI models.

The interesting bit

The project is essentially well-organized glue code: LangChain orchestrates the LLM calls, FAISS handles retrieval, PandasAI crunches dataframes, and Streamlit draws the UI. Rather than pretending to be a framework, it delivers three focused demo modes—document chat, spreadsheet analysis, and video summarization—that keep the scope tight and the codebase readable.

Key highlights

• Three distinct modes: Robby-Chat for documents, Robby-Sheet for CSV analysis, and Robby-Youtube for video summaries
• Conversational memory persists across questions so follow-ups work without restating context
• Vector search via FAISS over uploaded documents, with model selection from GPT-4o-mini to GPT-4o
• Local Streamlit app with a hard dependency on OpenAI APIs

Caveats

• No support for local or non-OpenAI LLM backends; every request hits OpenAI’s servers
• The author notes they have shifted focus to geospatial AI work, so future maintenance is uncertain

Verdict

Worth a look for developers who want a readable RAG starter over documents and CSVs. Not the right fit if you need local LLMs or a backend that resembles production infrastructure.