GPT-4o
⚙️ Technical
Intermediate
Local Knowledge Base Builder
Plan and build a private, local RAG-powered knowledge base from your organization internal documents.
The Prompt
# Local Knowledge Base Builder You are a RAG (Retrieval-Augmented Generation) architect. Design a local knowledge base system for the documents described below. ## Knowledge Base Scope - Document types: [LIST TYPES, e.g., "PDF policies", "Word SOPs", "Notion exports", "Slack exports"] - Total estimated size: [APPROXIMATE SIZE, e.g., "500 documents / 2GB"] - Primary users: [WHO WILL QUERY IT] - Query types: [EXAMPLES OF QUESTIONS USERS WILL ASK] - Must stay offline: [YES / NO] ## System Design ### Ingestion Pipeline Step-by-step process for loading, chunking, and embedding documents. Specify chunk size and overlap values for this document type. ### Embedding Model Recommend a local embedding model (with Hugging Face path) suited to the content language and size. ### Vector Store Choose between ChromaDB, Qdrant, Weaviate, or FAISS for this use case and explain the trade-off. ### Retrieval Strategy Describe the retrieval approach: semantic search, hybrid (BM25 + semantic), or re-ranking. Justify the choice for [QUERY TYPES]. ### LLM Layer Recommend the local generative model and the prompt template for the Q&A step. ### Minimum Hardware State the minimum RAM, VRAM, and storage for this setup to run comfortably.
📝 Fill in the blanks
Replace these placeholders with your own content:
[LIST TYPES, e.g., "PDF policies", "Word SOPs", "Notion exports", "Slack exports"]
[APPROXIMATE SIZE, e.g., "500 documents / 2GB"]
[WHO WILL QUERY IT]
[EXAMPLES OF QUESTIONS USERS WILL ASK]
[YES / NO]
[QUERY TYPES]
How to use this prompt
1
Copy the prompt
Click "Copy Prompt" above to copy the full prompt text to your clipboard.
2
Replace the placeholders
Swap out anything in [BRACKETS] with your specific details.
3
Paste into GPT-4o
Open your preferred AI assistant and paste the prompt to get started.