The workflow builds an HR & IT helpdesk chatbot that ingests internal policy documents into a vector store and answers employee questions via Telegram. It supports text and voice messages by transcribing audio, performing retrieval-augmented generation (RAG), and keeping chat memory for contextual conversations.
What You Can Do
- Ingest PDF policy documents and extract text to create a searchable knowledge base
- Generate embeddings and store vectors in PostgreSQL (PGVector) for retrieval
- Receive Telegram messages (text or voice), transcribe voice to text using OpenAI, and route queries to the AI agent
- Provide RAG-powered answers from the HR/IT vector store with session-based chat memory
- Handle unsupported message types with a fallback message
Quick Start
- Import this workflow into n8n
- Configure your credentials for Telegram, OpenAI, and PostgreSQL (PGVector enabled)
- Optionally run the setup/test trigger to fetch the example PDF and build the vector store
- Activate the workflow and test by sending a text or voice message in Telegram
