PDF Learning Buddy Bot Using Telegram

Workflow Steps:

Telegram Trigger

• Starts when a user sends a message or PDF document via Telegram.
• Captures message text or document metadata.

Check If is a Document (If Node)

• Determines if the incoming message contains a PDF document.
• Routes the workflow based on message type (document vs question).

Telegram get File

• Downloads the PDF document from Telegram if present.

Change to application/pdf (Code Node)

• Ensures the file is properly formatted as a PDF.
• Updates metadata such as MIME type and file extension.

Default Data Loader

• Loads the content of the document for processing.

Recursive Character Text Splitter

• Splits the text into manageable chunks with overlap to preserve context.

Embeddings / Embeddings OpenAI

• Converts text chunks into vector embeddings for semantic search.

Pinecone Vector Store / Pinecone Vector Store1

• Stores embeddings in Pinecone for future retrieval.

Vector Store Retriever

• Searches the Pinecone database to find relevant information for user queries.

OpenAI Chat Model

• Processes queries with GPT-4.1-mini to generate AI responses.

Question and Answer Chain

• Combines retrieved context with AI model to answer user questions accurately.

Telegram Response

• Sends AI-generated answers back to the user.

Limit to 1

• Prevents duplicate messages by limiting output to a single item.

Telegram Response about Database

• Confirms the number of PDF pages saved to Pinecone.

Data Sources & Tools

• Telegram Bot → Receives messages and documents from users.
• OpenAI API → Generates embeddings and answers questions.
• Pinecone Vector Store → Stores and retrieves embeddings for AI queries.
• n8n → Orchestrates the workflow and node connections.

Key Benefits

• Automates learning support by answering questions and processing new documents.
• Efficiently manages PDF documents by splitting, embedding, and storing them for retrieval.
• Provides accurate, context-aware AI responses in Telegram chat.