Current System Overview
The Knowledge Base is critical for chatbot instruction. When you update documents, embeddings must be regenerated for the chatbots to use the latest information. This guide covers all methods for updating documents and ensuring proper embedding regeneration.
📊 Current Knowledge Base Status
Storage Structure: Firebase Storage knowledge-base/public/ (9 markdown files)
Firestore Collection: knowledge_documents (9 document records)
Embeddings: knowledge_chunks (62+ embedding chunks for chatbot RAG)
Status: ✅ All documents loaded and operational
- • README.md → SHELTR Platform Overview
- • blockchain.md → SHELTR Blockchain Technical Documentation
- • donor-guide.md → Donor User Guide
- • hacking_homelessness.md → Hacking Homelessness Strategy
- • participant-guide.md → Participant User Guide
- • shelter-admin-guide.md → Shelter Admin Guide
- • sheltr-tokenomics.md → SHELTR Tokenomics and SmartFund™ Model
- • system-design.md → SHELTR System Design and Architecture
- • whitepaper_final.md → SHELTR White Paper
Method 1: Script-Based Update (Recommended)
Use for bulk updates or when you have local markdown files
- • List available documents
- • Update single document
- • Update specific document by ID
- • Update all documents from directory
Method 2: UI-Based Update
Use for individual document updates through the dashboard
- • Login as Super Admin
- • Find document to update
- • Edit content in text editor
- • Save with automatic embedding regeneration
Method 3: Direct File Replacement
Advanced users only - requires manual embedding regeneration
Embedding Regeneration Process
Why Embeddings Matter
- • Chatbots use embeddings for semantic search
- • Enables RAG (Retrieval-Augmented Generation)
- • When document content changes, old embeddings become outdated
- • New embeddings ensure chatbots have access to latest information
Automatic vs Manual Regeneration
✅ Automatic Regeneration
- • Using the update script
- • Using UI update with file upload
- • Using the new API endpoint
❌ Manual Regeneration Needed
- • Directly editing files in Firebase Storage
- • Manually updating Firestore documents
- • Importing documents via other methods
🔧 Recommended Workflow
For Regular Updates: Edit markdown files locally → Run update script → Verify in dashboard → Test chatbot responses
For Single Document Updates: Edit specific file → Update using script → Check dashboard for updated chunk count
Always verify that embeddings were regenerated and chatbot responses reflect new information