Real-Time Documents RAG Chatbot

The project centres on building a Retrieval-Augmented Generation (RAG) chatbot that surfaces clear, accurate information from publicly available documents and keeps its answers current with real-time data updates. The conversational layer should feel natural to end-users while pulling the freshest possible facts the moment they ask.

What the finished solution needs to do
• Ingest a corpus of public documents—PDFs, web pages, data feeds or similar—and index them for semantic search.
• Retrieve the most relevant passages at run-time, feed them into an LLM, and stream answers back to the user in plain language, always citing the source material.
• Detect when underlying documents change and automatically refresh the index so responses never go stale.
• Expose an easy-to-embed web chat UI plus a simple REST/GraphQL endpoint so other apps can tap into the same knowledge base.
• Include deployment scripts (Docker or similar) so the stack can be spun up quickly on our cloud account.

Key technical notes
Python is preferred; popular RAG toolkits such as LangChain, LlamaIndex or Haystack are welcome as long as the code remains modular and well-commented. ElasticSearch, Pinecone or another vector store may be used for embeddings. The language model can be OpenAI, Anthropic or an open-source alternative—just document any API keys or weights required.

Acceptance criteria
1. Ask-answer cycle under three seconds for a standard query on a 1k-document set.
2. At least two citations returned with every answer.
3. One-command rebuild script successfully re-indexes fresh document versions and reflects updates in responses.

Hand-off deliverables
• Source code repository with README.
• Docker-compose (or Terraform/Kubernetes) files for deployment.
• Short Loom/video walkthrough showing configuration and a demo query session.