Building a Personal Notes Assistant with RAG, Amazon Bedrock, and Pinecone

Описание и применение

Создана система персонального ассистента заметок, использующая RAG, Amazon Bedrock и Pinecone для быстрого и приватного поиска информации в текстовых файлах.

Задача

Users struggle to locate specific information within hundreds of personal notes, wasting time searching manually.

Результат

The system provides a private, accurate, and efficient way to query personal notes using RAG, eliminating manual search and ensuring answers are grounded in the user's own documents.

Рабочий процесс

• Upload .txt file to /ingest endpoint
• Validate file type, size, encoding, and content
• Store original file in private S3 bucket with unique UUID key
• Chunk text into overlapping snippets
• Generate 512-dim embeddings with Amazon Titan
• Index embeddings and metadata in Pinecone
• Receive question at /ask endpoint
• Embed question with Amazon Titan
• Retrieve top 3 matching snippets from Pinecone
• Assemble prompt with question and snippets
• Generate answer with Amazon Nova Lite
• Return answer as JSON

Доказательства

Project Goal: Upload plain text (.txt) notes and receive accurate answers strictly grounded in their content.

Источник

Amazon Nova Lite generating the response.

Источник