I need a Telegram-based AI assistant that can act as a true subject-matter expert. The core idea is simple: we will feed it a large, well-structured document set from whichever industry offers the richest material, then let the bot do the heavy lifting. While the sector itself can be decided once I see which corpus is most complete, the bot must speak fluently about Artificial Intelligence concepts in particular, as that will be a recurring theme in user queries. Once deployed the assistant should: • read and index every document I supply (PDFs, Word files, web links, internal wikis) • answer ad-hoc questions instantly, citing sources • provide actionable recommendations where appropriate • deliver concise summaries of lengthy texts on request I expect a smooth user experience inside Telegram: natural conversation, inline replies, and threaded follow-ups. Technically, you are free to use OpenAI, Claude, or another LLM, so long as retrieval-augmented generation keeps answers grounded in my data. Python or Node.js, LangChain, and a vector store such as Pinecone or Weaviate all fit well here, but I am open to equivalent stacks. Deliverables 1. Fully functional Telegram bot deployed to my server 2. Ingestion pipeline with clear instructions so I can add new documents myself 3. Brief user guide plus setup notes Acceptance is complete when the bot can handle a ten-document pilot set, return accurate answers, recommendations, and bullet-point summaries, and preserve source traces for every response.